For those wondering, previously I found 24H2 was 11% faster with the 9700X when compared to the 'bad install'. Compared to the 'good install' it's 7% faster on average, and if you remove the best 10 results from the 43 game sample it's just 4% faster. So the margins will depend heavily on the games used for testing. 7%: ibb.co/vHQ7mnx 4%: ibb.co/QX49twq
Administrative powershell or command prompt. Sfc /scannow. Windows found no integrity violations. Fresh installs always need this and usually after certain updates.
The system admin performance boost hasn't really been addressed I feel since the original video - would it be feasible to check if that makes a difference on any of these installs/versions?
Ok so, for future Windows installs, you should keep track of: - Room temperature - Humidity - Time of day - Moon phase - Complete list of all devices detected by Windows - Sequence of plugging in and out screens, keyboards and mice - Amount of computer restarts So yeah... good luck!
? That is it? The future sounds awesome! Today and the last 30 years have been a dark art of magic to install. So no blow into the cartridge whiles chanting the Devil's minion supposition Nr5? I mean XI. What actually very fond of that part. Humidity and Room temperature is like second nature. More sweat and higher fake FPS's in Online Service p2w games.
@@mikem9536 first, not really a fine wine issue. This is a sporadic degradation in performance, not a free boost. Many initial benchmarks will have been roughly at the good/windows 10 perf level, and this is just reclaiming (largely) what has been lost. Secondly, everyone is blaming this on windows, but wth is AMD doing not noticing this. It’s clear their performance is janky all over, and they are doing a terrible job benchmarking/validating their products. I’m not a AMD hater, I fully intend to upgrade with AMD next based on intels current issues. But nothing about this screams “fine” to me. Both companies have to get their acts together. Even still, I find no evidence some of these new patches won’t break/degrade later on leading to similar losses in perf which is unacceptable. Both Intel and AMD should be working to validate their product with the OS 90% of users install, which is, like it or not, windows.
So AMD's internal testing had variance of windows installs which may explain their "performance gains" from the 7700x to the 9700x. The uplift may just be the difference between a good vs bad install, or a patched 23h2 vs one that wasn't.
@@davidhiggins2804 I don't think anyone in the industry have discovered this until HW did such a thorough test. Who would have thought that a difference in install would create a 7% gain.
Or.. (And I say this as someone who have or have had the following AMD CPUs: 7950X3D, 5950X, 2700X, Phenom X6 1100T, Athlon 64 X2 6400+, K6-III 450Mhz)... perhaps.... AMD is just trying to ass-cover so they resort to spouting unverifiable nonsense.
This isn't too surprising. When I taught an CompTIA A+ class we had a room full of identical PCs, with identical HDDs, and CD-Rs of Windows 7, Server 2003/2008 , etc and inevitably one PC or more would just have a borked install. My colleague and I started tracking which CD-R, HDD, PC, etc was having the issue and we could never find any rhyme or reason. The issue didn't follow any particular device, and it appeared totally random. This would never happen in my Linux classes, only ever with Windows. Years later I remember I was installing MS Exchange (2013?) on a fresh server, and I ran across an error that according to MS KB article the fix was to reboot 4-6 times. I didn't believe it, but after rebooting about 6 times in total, the issue resolved itself. Since then, things like this with Windows are pretty much to be expected unfortunately, apparently even many years later.
Summary: So it sounds like 23H2 vanilla is often depressed (lower) than WIndows 10 pretty much across the board. Plus it can be even worse if your 23H2 gets into the "broken" state. Adding the KB patch/update brings 23H2 up to Windows 10 levels and sometimes even a bit better. 24H2 still seems to have the best performance overall, surpassing Windows 10, and often even better than 23H2 with the KB update. Of course 24H2 is new enough that we probably don't have enough data to know if you can get a "Bad install" of 24H2 also. So seems like sticking with Windows 10 is fine, or go to Win11 24H2 if you want a perf boost. Avoid 23H2 because there is enough variability that you might end up with worse than Win10 performance. What a headache.
Honestly for Ryzen Windows 11 has been bad from the start, I remember when I upgraded from 10 to 11 that I lost around 5-10% no matter what build of Windows 11 I had used, this is also true for my Intel based laptop. Will have to check with the new KB as I'm not currently home
How about 22H2? Having the TPM 2.0 disabled in the Bios can prevent Windows 11 updating to 23H2. I wonder if 22H2 could have the same inconsistencies or not.
I can confirm once I installed the Dev version of Windows Insider build 24H2 (currently 26120.1542), the system is tangibly faster (7700X). And not just gaming. It's just overall snappier. 🤷♂
@@Ilestun could be putting threads on better cpus, leaving threads on cpus longer to not lose what is in cache, some code path related to making system calls that is optimized for zen4+? I'm super curious too
Warning: *Do not install the KB5041587 preview update if you dual-boot with Linux* This is the one that breaks GRUB. It's listed as a known issue in the KB page.
I have two SSDs, one for Windows and another for Linux.. two separated EFI partitions, one in each drive for each OS. No problems here. I press F8 during system boot to choose the OS to boot to.
It broke mine, but if you remove all other drives from fstab then Linux will boot. You can remount the drives once you are booted into Linux again though, I think the drive UID changed and thats why it didnt boot.
Yes I noticed that too, when I did a clean install of Win11 23H2 on my gaming rig. Before I was getting 168 FPS in Horizon Forbidden West on average, but now I'm getting 188 FPS on average. That's crazy. Still running a 5800X3D
Yeah i tested 5800X3D pretty much as soon as the patch dropped. There were some minor average FPS losses but my biggest gains were from 0.1% lows aside from two games that saw a regression. I mainly tested older games. None of them using UE5. The newest game that i had installed and that i tested was Ghost of Tsushima and this saw on of the biggest 0.1% gains. Also im no Steve and i dont have the patience for 3 runs. Aside from RotTR all were live gameplay's and not canned benchmarks. I did not lower the settings to 1080p low or similar to create an artificial CPU bottleneck. All my tests were at 1440p max settings or close to it since the goal was to see of i could see/feel gains on the settings i actually use. Due to the limited numbers of runs, variability of these runs being live gameplays and GPU limited settings there are some oddities. Also when fps is low then even a 5fps gain can mean massive percentages - like in RotTR for example or even GoT. if i remove the outliers 8both positive and negative) that equates to -1.75% drop in average fps and +5,5% rise in 0.1% low fps I did see +13-14% often while other number were more rare. Here are the games and percentages: Ace Combat 7: -2% average. +14% 0.1% Ghost of Tsushima: -2% average. +38% 0.1% Hitman 2016: -7% average. -25% 0.1% Mass Effect 3: -1% average. -14% 0.1% Quake Champions: 0% average. +13% 0.1% Rise of the Tomb Raider: 0% average. +60% 0.1% Star Wars Battlefront II: -2% average. +9% 0.1%
@@AnomieDomine Well in this specific game yes, but I haven't tested anymore games. just haven't had any time lately. But I'll test more games when there's time.
Na. And it is not 'just' the OS holding the hardware back. Regression in performance. Improvement in performance. If a bug or whatever reason is fixed, and causes performance loss? To bad. If a update is bad and regress performance then that is a problem. Intel and AMD have pushed security updates that loses performance. But Intel is far worse even telling people to disable hyperthreading. Literately making your I7 CPU turn into a I5... A know 'good' performance should be the baseline. Any regression should be taken as a bug or mismanagement. But we are talking about Microsoft and Winblows here. Do not expect anything from a company asking Home: $139, Pro: $199 for Windows 11.
Microsoft needs to be telling us why there's so much variance and giving us real solutions. As it stands, nobody can know if he or she is losing performance to a "bad" OS installation. That's not acceptable.
I suupose it's because they don't even know their own OS. With every update (except maybe for this one) Windows becomes worse and worse. I've given up on it, but still I understand it isn't an ultimate solution, and some people cannot leave Windows or even prefer it with all its weaknesses and problems. Sometimes there's no other options, and I feel for those people. Hope Microsoft do at lease something about their main product.
If games just work on linux i would switch. Try to do it last month but one day everything was working then the next my games would not boot. Spent my limited gaming time troubleshooting linux just to give up go back to window. I spend enough time troubleshooting problems at work and don't want to do it in my free time. One day Linux will catch up I hope.
@@joshhelberg I'm sorry you had that experience. Mine is solid for almost two years on one of the most "unreliable" distributions. Though I like tweeking and modifying my setup, so to be fair it should be considered too.
You could take a registry dump from your good and bad windows installs… and then I can’t remember the name of the software, but you can get a delta of differences between the two. Might have a clue.
lol Imagine actually being thanked by a hardware or software company for doing unpaid beta testing of their buggy junk. These days buying hardware or software anywhere near it's launch day is basically a guarantee you'll overpay for a bad experience, it's a joke.
@@amurtigress_mobile365 He is navigating a labyrinth of bs from these companies. He is releasing videos with as much of more information as anyone else has. He reports again and again as he discovered new pieces of info. It is not a correction as if a previous video was flawed. It is an update based of 100 hours of testing. If he didn't release anything until he had every piece of the puzzle then the video would be worthless as it would all be solved and/or covered by other channels. Your claims are conjecture as there is no evidence of any wrongdoing or mistakes.
The fact that Callisto can run at nearly 300FPS with that level of fidelity is kinda crazy. It stuttered like hell on day 1 but the optimization is pretty good after the patches.
I played it recently, because it was free on Epic Store and knowing of the bad reviews, I was surprised how fun it ended being, I liked it, is rare for me to finish a game this days but I did that one. But I'm pretty sure what made me like it more, was the fact I could disable many of the gameplay bs, like the quick time events and turn the combate mostly into a single button press for some stuff, meant it became very fluid and that made it fun for me. I still add I few pet peeves with it but the rest made up for it. Hope it starts getting better reviews.
Callistro still stutters if you lean hard enough on the GPU. It's like it can't figure out which shader code to compile when and seemingly disregards untapped CPU resources.
It's refreshing to watch someone present just the salient information calmly and clearly rather than breathlessly racing through vast quantities of information like they're commentating on a horse race like some other tech TH-cam platforms.
Point of note, it’s unlikely, but the specific SSD is a variance between those good and bad installs. For the sake of complete troubleshooting, it might be worth spot checking that by cloning that drive to another one.
Yes. Steve have you checked the SSD the bad 23H2 is on ? I believe there are checkers that will also test various locations of the ssd. Agree with @joeynessily that a clone to a brand new ssd would rule this out.
Earlier this year, I bought a new SSD and cloned my Windows drive to it. Over the next few days, I got a few blue screens and found a few games that crashed on launch when they worked fine before. At first, I worried that the new SSD was bad, but running chkdsk, dism and sfc and re-installing the broken games fixed all of the issues and I haven't had any since. So, it seems like the SSD is good and the problem was with the cloning. I'm just throwing that out there because it might be a good idea to not just assume that a clone is a 100% match and, instead, verify it after the process completes or at least not wipe the old drive until the clone has been used for a while.
I have fixed bugs and other performance issues in the past by running the Check Disk Utility (CHKDSK), Deployment Image Servicing and Management (DISM), System File Checker (SFC) commands. These have to be run with as an elevated PowerShell prompt. These fix any corruptions on the drive and system image, which can happen even on a fresh install and with updates. It may have nothing to do with this specific issue, but it is a silent variable not a lot of people know exist.
@@opbush5272 it depends on what is actually broken/corrupted, it might fix something you don't even touch, some people get bluescreens on games due to something being corrupted due to a update somehow going "wrong" might not be the right word, but either way they have had those not happen after running those cmd checks and fixes
Somebody mentioned running sfc. And also, let me quote "sometimes needing dism". I know that sfc stands for system file checker which tests and restores the system files, in case they got corrupted. No idea what dism is.
@@Winnetou17 probably refering to this command: Dism /Online /Cleanup-Image /RestoreHealth and a periodic DISM.exe /online /Cleanup-Image /StartComponentCleanup can free up diskspace on occasion too (can optionally stick /resetbase on the end of that one but read up on what it does)
@@Winnetou17 Dism is almost the same thing, but (long story short): it tests a few other files. The way to go is always: DISM first (as it redownload the image from MS, so if you have a faulty SFC, it is most likely fixing it) then SFC.
Steve, this is bloody amazing work. I shudder to think how much testing this would have been, and you've presented dense, complex information in a super easy to digest and see the difference format. Thank you for all you do for the PC hardware community, we are blessed to have you
Windows can perform inconsistently even between boot-ups/reboots. You cold-start a PC in the morning and find out Windows is performing very sluggish, a simple file copy operation is extremely slow or web pages are not loading up properly. You reboot the PC and it performs normally like none of it ever happened. A real hair puller.
This can be a result of the fast boot setting in windows - it defaults to on, and makes it so a total shutdown is actually not, but a restart will always do a proper shutdown before starting again and as a result clears problems a default shutdown won’t
Good vs Bad install is intriguing. Export the Windows registry for each install to a USB and compare all 4 registries for non-duplicated data before the patch and after the patch. How to do this, I think its possible to export to a TXT file, then open that in Notepad++ to do a diff between the TXT files. Recommend exporting first folders within the registry into separate TXT files instead of just exporting from the root into 1 TXT file.
The Windows registry has always been a pile of crap. Everytime a registry key gets deleted it is only 'marked' as deleted the key and its data is left where it was, even if an identicle key and data are re-created it allocates new space to store it. This is one of the main reasons Windows slows down over the years, each little addition on its own only adds a few ns every access, but do that thousands of times over a few years and it becomes noticeable. I found this out in the late nineties writing installation scripts for software I had written, installing/uninstalling testing the same software dozens of times over a year for upgrades made my developement machine noticeably slower. This was made worse due to the rather limited RAM on most PC's back then, having the registry file grow from about 2MB when new to 6-8MB when you only had 64 or 128 total. Re-installing Windows (it was NT4 and Windows 2000 back then) resulted in a much nicer responsive machine. I suspect the reason it has never been fixed in the almost 30 years is that it makes people buy new machines with new Windows licenses. When all they often need is a new/clean install of these poorly designed/implimented OS's.
@@vadnegru Who all run with Virtualization-Based Security on (which is default) Meaning they are still slower than Windows 10. But people now wrongfully think Windows 11 is faster. HWU testing with VBS disabled is bad. Disabling anti-malware scanner during testing as that can disrupt runs when it decides to scan or update, sure. Disabling default security features that run passively all the time and therefore produce consistent results, no. That's bad to the point of being misleading. People, such as yourself, now think Windows 11 is faster while it's really not. You have to disable security features to make it 2% faster and even HWU does not recommends doing that.
@@torvasdh yupp. During my early experience in TH-cam my comments got deleted due to some reasons. Got into a practice of censoring stuff like this and it's now automatic. 😁
Windows installs have been broken since about 2020. Literally every fresh install should run sfc. And it'll be faulty every time Sometimes needing dism. It's wild
its not always broken, but i do recommend after a system update that people should run sfc /scannow and a dism scan to rule out a bad update/install that could possibly happen.
I also tested this exhaustively and found no difference. The culprit is in exploit mitigations. If you disable exploit mitigations, you can see these same performance gains on any version of Windows 10 or 11.
From personal experience after installing 24H2, I had an increase in my Valorant and CS2 FPS by about 12% and 16%. Which for me, is the difference between being able to max out my refresh rate of my 520hz monitor and not. Before I was getting around 480-550fps in Valorant, but with constant drops to 350-400. With the 24H2 the lows never drop below 500 now, so the lows have gained A TON for me. Which means that it's so much more smooth on my monitor. Def recommend installing this for people. I have a 7800X3D. CS2 lows improvements are even more drastic for me.
@@diktaf No AA, though that depends on your graphics card. If I use 4x my lows drop below 520 and it's noticeable with this monitor, but before when i was using a 360hz, I used AA, because I didn't care if it dropped to 400fps. AA is all GPU, and since I only have a 3070, I've found that turning it on gives me random GPU spikes that drop my lows. But if you have a 4070 or AMD equivalent, you can def run AA just fine.
Is it possible that 24H2 could also suffer from the "good and bad" situation, considering there were times it had only matched KB5041587 on the bad Win11 install, and the good install running the patched 23H2 pulled a few percent ahead of everything else? It could also just be run to run variance I guess, but still.
@@Hardwareunboxed Yeah guys not getting the point like with OLD windows it used to be bunch of files which used to get installed and used to be consistent. Now that AI is in the play it does as it wishes and installs the way it wishes depending on it's mood. Before people start giving mea earful about how AI doesn't play a role during installation. I want to make it clear that this was a sarcastic joke on how AI is pushed into every dumb stuff nowadays where it doesn't even make tiny bit of sense.
@@KING_DRANZER AI is also pushed when almost nobody is asking for it, while things that people are asking are not done. Well, this patch is one exception. Also, here I'm saying it again, since I can't abstain myself: if AI is so good, then why isn't Microsoft using it to finish the Control Panel to System App transition ? It's been "only" 10 years (literally) since that started, and it's still not done and it's a huge mess (best example I have is the power profiles which you still need to dig for into the Control Panel for exact tuning, which is very frustrating for laptop users, especially those trying to test battery life) and extremely stupid.
@@Winnetou17 I remember seeing a reddit post a few years ago where people talked about Windows 11 and generally how slow Windows development is, and why it has so many inconsistent UI elements, and allegedly there was a leak of an internal chat between developers where they freaked out and got angry how they have no idea how the old code works in control panel and that it took them days to change like 1 UI element in control panel because the smallest change broke a bunch of other things in Windows and they had no idea why. It was agreed on that post that this is probably the case, because original Windows devs from decades ago left so new devs can't really deal with the decades old codes that are there because of backward compatibility and bad optimization (they keep adding more and more new code to not touch and destroy the old code) making development super slow.
@@canyongoat2096 Oh, I can totally understand why it's slow, I'm a programmer myself. Still, 10 years is A LOT. And the way they are doing this is also annoying, as settings that are still important are much harder to reach now. Really, there's no excuse why Windows 11 hasn't launched with everything ported to the Settings App, so the Control Panel could be removed completely. Or at least, still keep it easy to use, if you don't remove it. Right now it's the worst of both worlds in plenty (thankfully not all) of cases.
Very much appreciate the Windows 10 testing as well, not many (if any) larger channels cover Windows 10 anymore, and these issues would be lost in the shuffle. Can’t believe it took Microsoft this long (W11 24H2 official release will be less than a year away from end of W10 support) for them to reach performance parity with W10. Can finally feel like I’m not downgrading my machine when W10 sunsets and I move to W11
High fps still matters for input latency even if your monitor can't display all those frames. The faster the game renders, the newer and most recent the frame will be shown by your monitor when it refreshes. Watch 3kliksphilip's video on it.
@@xFluing Input latency is meaningless for anything above the 200-300 fps range. We're talking fractions of a millisecond here. For reference, the average person response time is 200 milliseconds.
@@mikehawk6918the reaction time doesn’t matter since latency adds up. But you’re right over 200 to 300 it’s diminishing returns. Unless you’re buying a 360 or 500hz monitor
Thank You Steve. Incredible work as always. Here's the difference in percentage. 22 Games Tested. Windows 11 24H2 ----109%---- Windows 11 23H2 Good Install KB5041587 ----107%---- Windows 11 23H2 Bad Install KB5041587 ----107%---- Windows 10 22H2 ----105%---- Windows 11 23H2 Good Install ----100%---- Windows 11 23H2 Bad Install ----95%---- Can't wait for 7800x3d & 5800x3d 24h2 test video. 😊😊
So KB5041587 is a "cumulative" update, so unless you can narrow it down to one or a couple individual patches it's hard to tell what the culprit is. I'd expect some security mitigation patch though. Your "broken" install could be either missing one of the KB's that the "unbroken" has - or vice versa. Microsoft likes to sometimes test some patches on a limited run off machines, so if you run any windows updates, that sounds like the most likely situation to me. There is a way to just print all installed KBs, would be the best place to start. Anyway the KB5041587 package would then install the missing patch.
Would love to see a video on the difference between the upgrade to 24H2 or a fresh install of 24H2. Upgrade a clone of the bad and good 23H2 installs to compare maybe.
Reminds me of how back around 2000 I found that, for reasons I could not figure out, fresh installing Windows 98 SE on a hard drive laying flat gave me worse FPS in Crimson Skies than when I installed Windows on the same hard drive standing on its side. Everything installed identically, I got noticeably more FPS from one install than the other and I had no clue why, but I installed over and over to the point I could recite the Windows key and the results were replicable.
It could be that the hard drive was overheating when laying flat. I have faint memories of having issues when I mounted internal drives right on top of one another, with barely any airflow between them. Also, much more recently, I found that my portable hard drives overheat and give me abysmal performance (like
When support for W10 runs out i'll be switching to Linux. I've had it up to here with MS. Every few updates my second monitor stops working for no reason, and a few days later all is fine again. Things that should be simple, like inverting left and right channel audio is not possible in Windows without 3th party software. No way i'm getting on the W11 or W12 train! I've lost so much time troubleshooting MS these past few years that i might just as well install linux.
@@kiamppand the new explorer/shell is laggy as shit. Every folder you open loads for a quick moment, it’s like the Win10 (or Win11 22H2) explorer on a >10 year old machine
@@TheSometimeAfter Normally you would stress Twenty 'Four' H Two to note the change from Twenty 'Three' H2. But in these videos Steve has often been stressing Twenty Four 'H' Two, 'Twenty' Four H Two, etc. I imagine its because hes said it so many times the words have just about lost meaning. Just a mildly amusing side note.
Have you tried doing a file-by-file comparison between the bad install and the good install? Just calculate a SHA256Sum for every file on both installs and stuff the results into a database. Then you can select the results where a given file path has more than one checksum. This would tell you what files are different between the two installs.
Yep, this is the way. Got to be something that hasn’t updated on the bad install and updated on the good install or it could be the other way, but I lean towards the first one.
Just take both drives, put them in a motherboard or enclosure running another install of Windows and do a "File Content" comparison of the two drives using FreeFileSync. Much easier...
@@tim3172 perhaps its easier if you already know FreeFileSync. Personally, I've never heard of it and never have used it. I do however have a database running on my NAS and several programming languages already installed so for me it's easier to spend a few minutes writing a program than to properly vet and install an unknown 3rd-party file utility.
9:00 As it has been noted on the forums, this is a problem in the new Nvidia driver, and I have this problem myself, the performance in the 537.58 driver is much better than in the 560 and newer
and that's why it is a good idea to stay on previous windows until it dies. if you are on Win10 stay until 2025. and hopefully until then win11 will be good
Just a small nitpick about wording - you're disabling _virtualization_ in bios, not _virtualization based security_ . The latter _uses_ virtualization.
I run VMs for work on Oracle VirtualBox. I think I need virtualization turned on in the BIOS. My Windows SAYS VBS is disabled when I query via powershell and none of the HyperV stuff is installed at all (Windows Hypervisor Platform, Virtual Machine Platform, Hyper-V are all unchecked in Windows Features). It would be nice to know if my Windows is performing up to snuff but I am not really sure how to tell. 5800x3D.
@@samhunter6813 benchmark with your current settings and then benchmark with virtualization turned off in BIOS to rule out anything if you really wanted to
@@samhunter6813 There is no issue with having virtualisation by itself enabled. There is no performance regression with it (ofc, unless VMs are running in the background). And even with VMs running, I found that the issues I thought were VM related had other causes (for some reason my mouse started introducing stutter in some games at higher than 250Hz polling rate). To make sure, you could enable VBS, benchmark your system, disable VBS, benchmark again. There should be a decent performance gain when disabling VBS.
@@james2042 bought mine 135 euros (and sold my 3700x for 85 euros), it clowns even more on everything for a 50 euros upgrade. Definitely the best performance/price ratio. AMD and their AM4 socket is definitely goated
microshits bs is pathetic. As a windows modder and a gamer, I tend to reinstall windows quite often, and I've felt that some installs just perform worse in games. I've had that feeling all the way back when windows 10 came out. I'm glad I'm not going crazy, but I'm infuriated at M$
It's baffling to me that every youtuber seems to keep overlooking the fact that 24h2 is a complete OS replacement update. Unlike the updates we are used to this 24h2 update replaces your current version of windows and doesn't actually updating anything. It's more like a fresh install than it is applying anything over your existing os. It's clearly stated in the update description and should easily explain why the 24h2 update performs better than the os the update is applied to and why it can also be similar to clean installs of older versions of windows when they are clean installed. There's also performance impacting security features that will automatically be enabled when they weren't in older versions as well as some settings being kept if 24h2 is applied as an update. Performance impacting things that may have been turned off by the user previously will still be off while an actual clean install of 24h2 will have them enabled by default. Considering how many times simple things are over looked I do have some doubt in things just being completely random like claimed here.
No wonder this video took so long to make, the amount of effort put into this test is staggering. On my case, I have been using Windows 24H2 for over a week with very good results. There are some bugs, but performance is really good. I always ran SVM disabled. Most UE4 games got a 10-20 fps increase. With Remnant 2 I got a 10% increase. Subnautica went from 70-90 fps, to a locked 144 fps. Horizon Zero Dawn and Serious Sam 4 had close to identical results.
Unity and Unreal are majorly impacted by this? That's like 80% of the games released today.. this should be a (literal) game changer for everyone on 11. All 30% of them. At least those of us still on 10 aren't suffering as bad, but having improvements too would be welcome. Always expect you're not getting the performance you could be, simply because software optimization went the way of the dodo when people stopped needing to program their own engines and everyone switched to bloated high level libraries and visual scripting.
@@AceStrife I also tested with My time at Portia, and that game also got a huge increase in performance. UE4 and 5 games seem to get a good boost. But Unity game seem to benefit even more. Also tested Wo Long: Fallen Dynasty, but this one got no improvements. It uses the Katana engine.
Here's what you do. You do a file compare of those install drives to see what's different between the installs. You get a 3rd SSD. Install Linux on it. Install the other two SSDs as well, but don't boot to either of those, boot to the Linux. In Linux mount both drives with the Windows installs and start using tools to compare the two drives to see if they're different. Get a Linux scripter to build you some scripts or Google some tools. Then you can see if there are files missing or added between both. You'll also be able to see if some files are sized differently. Any suspect files, you Google those and see what they're used for and then track down what's different between the two installs that way. I only recommend Linux to do this rather than a 3rd Windows install since Windows might put random files on both drives
It’s kind of wild that the hardware market has been so lackluster lately that one of the most significant upgrades you can make to your PC is a fresh install of Windows
I'm looking forward to my 7700X getting some free performance... not that it needs it, lol. Some thoughts on the variance between Windows 11 installs: * Windows 11 is very network chatty onto the internet; are these tests done offline, and if not, it might be worth investigating? * Was the "bad" install before the "good" install? Modern AMD Zen processors "learn" things at a low hardware level that would survive any kind of OS and drive change. This unfortunately means that it could even result in slightly different data depending on the order of games test (or rather that changing the order could change results) * Do those SSD's have MLC or TLC NAND? MLC would give the most repeatable results. * Run the following in an administrative Command Prompt to ensure Windows is as healthy as possible (this can be a problem even on fresh installs and especially after installing any kind of Windows Updates): sfc /scannow DISM /Online /Cleanup-Image /ScanHealth If the DISM line results in something like "the component store is repairable," run: DISM /Online /Cleanup-Image /RestoreHealth Restart the machine for the corrupted system files to effectively be repaired, and then the testing can begin. Either way, all this means that: * Game testing on Windows is probably only going to get less and less repeatable (less precise) as it gets more sophisticated * Same for CPUs and probably GPUs as well. Branch prediction alone already introduces some slight variability. * This will then require even more games thrown out if they exhibit too much run-to-run variance and install-to-install variance, or just mentioned for transparency but not included in actual performance metrics since the data is "untrusted." * All this variance / lack of precision will mean that larger sampling is needed to ensure that a single or a few outliers don't heavily skew the data. Of course, as already said, the new patch results in a wide range of results depending on the game. At least by increasing the game testing array, it improves the odds that a real-world user's experiences are reflected. * It probably needs to be verbally stated more often by tech reviewers not to take any results as gospel; to research performance based on data from multiple different sources; and to try and put the data and results in perspective in the real world. For example, while disabling MS Defender is wise to improve consistency, this shouldn't be done in the real world under normal circumstances. The goal of reviewers is to eliminate as many variables as possible with the caveat that not EVERY variable can be removed.
I have a theory - Not sure if you will see this as there are many comments. For good vs bad install, it could be the polling rate for USB devices connected to the PC. As someone who has been playing around with polling rate for years, this is something I am very well aware of, but it might not be the case here, just throwing it out there for perhaps further testing (by unplugging before tests, or using VERY OLD mice and keyboard for the tests. Sometimes the driver being used for the keyboard and the mouse randomly uses a different polling rate. If a mouse or keyboard runs at 8000Hz or 4000Hz polling rate, it will impact CPU performance.
But, if everything is the same every time with the exception of the bad install then how can that be it? I would expect the SSD if that hasn't been ruled out. It could just be something random happening on occasion causing this. I remember windows 95 installing differently every time with the interrupts being assigned seemingly at random.
I though 98 wasn't that good but 98SE was good. The next one, ME, was bad while its contemporary NT based version 2000 is good and was used on a lot of business and embedded systems even into the 2010s in some cases.
Calling Win2k bad is hilarious. It was the most stable one yet. I even use it for my old Pentium II Laptop, distraction free ms word machine regularly.
I watched your video and immediately installed a fresh windows install from the 24h2 iso on a RUFUS usb drive. FOR ME... everything runs better. FPS on average increases between 5-8% playing on my 4090. I do play at 4k so performance improvements may even be better at lower resolutions.
Steve, don't wanna open the Pandora's box, buuut, can't help but wondering what would win11 vs some of the linux gaming versions look like! Also, standing up gives another fps boost.😅
@@mikem9536 Eh, less compatible isn't really in question as there are still games and rootkits that don't work well or at all. But slower? That's very variable and can't really said as a blanket statement. Most of the games I play, RPGs and strategy titles generally, tend to actually run better. The DXVK translation is a boon for some titles and CPU performance seem better under Linux at least for me.
@@mikem9536 Linux tends to be faster, even with the needed translation layers, but it depends on the distro and kernel. there are a few tests from the German outlet computerbase and pcgh, results are really impressive
I mean, both zen 4 and 5 get an uplift from the patch, so there's still poor gaming increase for zen 5. I'm looking forward to getting some zen 5 CPUs in the server racks, but I won't be running out to upgrade my gaming rig anytime soon.
@justsomepandawithinternet It doesn't suck for server work or productivity, which a LOT of people do. It's a good upgrade for them. It's got a small increase for gaming, but it's an increase, so it doesn't really suck from that aspect. It's the current pricing vs a 7000 series processor that sucks.
Dear goodness, I'm impressed and in awe of the amount of testing you've done for this one. Thank you so much for providing this attempt at clarity for as far as possible. Much respect. Have a good one 👍.
I installed the patch just because, knowing anything about what it does and I noticed CP2077 was running a lot smoother, even in DogPoopTown and I'm on Zen3.
@@ericbanker1285 I looked this up on Microsoft, the description page says nothing about fixing the scheduler for AMD or anything similar. Did AMD say this update contained their new code?
@@bakakafka4428 I tried replying earlier, but my post seems to have been blocked, possibly because I tried to directly link benchmark results. A Reddit user posted his benchmarks on the AMD subredddit that found the similar ~3% boost in a specific benchmark from both Window 11 KB5041587 and Windows 10 KB5041582. The Microsoft page for Windows 11 KB5041587 likewise says nothing about fixing scheduler for AMD, despite benchmarks showing improvements.
Hey Steve, I don't mean to be unappreciative or anything, but we really need you to add both the 7800X3D and the 7950X3D to these charts, because 1) the 7800X3D is the best CPU a gamer can buy, and is also affordable, so a really strong recommendation on top of it already being very popular in sales, and 2) the 7950X3D is a CPU that is incredibly vulnerable to the whims of Windows, so your work is undoubtedly valuable in these videos where you're holding Microsoft to account for us gamers who needed to get a 7950X3D for the non-gaming workloads. 🙏
Correct, Am the outlier as I've been telling my friend helped him build a 5800X build in 2020 to stay on 10. Idk how many 11 installs I went through (still have a 10 install around), 11 drove me back into the arms of Linux for gaming and HTPC the abuse got bad enough. 0 issues from that friend on 10 though he's been stable for years (is a poweruser).
According to these results Win 11 24H2 should be the best gaming OS - find it hard to believe it beats Win 10 but the data is the data. If you're going 24H2 might as well use the LTSC version, which you can install MS Store on too should you need it.
@@Osprey850 I have to look this up. of the gains are good then I might just go back to windows 10 for the rest of the year lol. changing builds every other month is very frustrating 🥲
It should not be up to the consumer to have to roll an "installation lottery". Microsoft sells Windows 11 as a product, the consistency of that product is the their full responsibility. Hopefully though, it seems like these patches have basically eliminated this lottery behavior, but only time will tell. Still shouldn't have happened in the first place.
What a hot mess! Who knew I was going to be looking at benchmarks based on how good or bad an OS install was? Wonder if the PC is in a black case vs a pink case will make a difference.
Of note: I don't think there's any level of attention-grabbing thumbnails or titles with vitriol thrown at windows that could upset the community. Windows 11 does, in fact, suck and I think even the handful of people who say they don't agree do still actually agree.
Back in Xp days, Some of us would strip the Xp Installed down to 200-500Mb, removing all the fats and bundled bloat. Now this was when only HDD's existed and Ram at most for the avg person was 2GB most people where 256MB/500MB or maybe 1GB By stripping Xp down, one was able to make it run on less Ram and faster, and also the HDD access/read/write was faster. I would not be surprised one bit if the same is for Win 7-11 What we really need is to have most components optional add-ons that install during setup there should be 1 Version and everything else is just profiles (Home/ Pro / Gamer/ Education/Enterprise etc etc) of what add-ons get installed. And while each has a default list of add-ons select you can unselect them and not have them installed
@@scarletspidernz a little bit of both. I start off with NTlite since you are basically able to do alot of tweaks that you would normally do while windows is running. Then I finish all my other tweaks while windows is actually up & running. The trick is to not disable critical services since it can brick your windows. another thing to take into account is how you lose performance over time from updating windows often. correct me if I’m wrong but my take is that windows gets corrupted from all the left over updates. I turn off windows updates now lol. It’s getting to a point where I need to just have 1 copy of a back up registry to save even more time
In no way an expert or knowledgeable but... The recent discussions about downgrade attacks (see Alon Leviev) show that windows does not have an overarching integrity system with updates merely using predefined steps. If a "step" was missed, hit an error, etc. and did not have a custom integrity check within the update (the catalog does have some rudimentary checks that trigger rollback steps) 2 systems may behave wildly different after applying updates. It seems likely that an upgrade to 24 would either do a massively expanded integrity check or just touch enough files that problematic system components are likely to be overwritten. Try comparing the checksums of all files in the Windows folder between the good and bad install. Rudimentary check but it's a simple one to do. There are built-in windows utilities that could be used without a reference filesystem but they are incomplete and Levievs results show they make many assumptions favoring speed over accuracy.
For those wondering, previously I found 24H2 was 11% faster with the 9700X when compared to the 'bad install'. Compared to the 'good install' it's 7% faster on average, and if you remove the best 10 results from the 43 game sample it's just 4% faster. So the margins will depend heavily on the games used for testing.
7%: ibb.co/vHQ7mnx
4%: ibb.co/QX49twq
it's incredible to see How AMD got nerfed by Microsoft....
Administrative powershell or command prompt.
Sfc /scannow.
Windows found no integrity violations.
Fresh installs always need this and usually after certain updates.
will 25h2 will reduce performance back to 23h2?
and if remove the 10 worst results?
The system admin performance boost hasn't really been addressed I feel since the original video - would it be feasible to check if that makes a difference on any of these installs/versions?
Disappointed. I thought steve actually went full lex luthor.
This must be his villain arc.
Same haha I was like welcome to the BIC Club and disappointed now 😂
same , it actually suits him not many people can pull that
Is Steve actually bald and has been wearing a hair system all this time? Everything needs to be questioned now.
🤣
Ok so, for future Windows installs, you should keep track of:
- Room temperature
- Humidity
- Time of day
- Moon phase
- Complete list of all devices detected by Windows
- Sequence of plugging in and out screens, keyboards and mice
- Amount of computer restarts
So yeah... good luck!
"My boot sequence is questionable, but 60% of the time it works every time!" th-cam.com/video/yvBTCS3Ky4w/w-d-xo.html
So it's as sensitive as Intel's original 10nm.
Alternative approach: don't cry and whine about minor performance differences and just enjoy the game.
? That is it? The future sounds awesome! Today and the last 30 years have been a dark art of magic to install.
So no blow into the cartridge whiles chanting the Devil's minion supposition Nr5? I mean XI.
What actually very fond of that part. Humidity and Room temperature is like second nature. More sweat and higher fake FPS's in Online Service p2w games.
@@tim3172 Nah, I didn't buy the highest end pc possible to not utilize it to it's maximum.
The silicone lottery is dead, in comes the windows install lottery
*Silicon
i think you meant the silicon lottery; the silicone lottery is your chances of looking the way you want after plastic surgery (lol)
The silicone lottery is when you finally test the watertightness of your newly installed bathtub
@@fonsuimuch worse odds there than the silicon lottery i’d wager
Silicone lottery is whether or not your upcoming date has real airbags.
Awesome work as usual! We haven’t got many reviewers left willing to put this much effort to get the data right.
DEFCON 5: Steve is sitting
DEFCON 4: Steve is standing
DEFCON 3: Steve is bald
DEFCON 2: Steve is bald AND standing
DEFCON 1: Steve is [REDACTED]
DEFCON 0: Steve
DEFCON 1: Steve and Steve Burke from Gamersnexus are in the same video.
@@antiheldd.3081Apocalyptic levels of bad: Both Steves are bald.
@@oldmanmarsh2528 the universe is ending: they're both standing
When does this lame joke die?
A windows update being equivalent to a generational upgrade is just insane and frankly insulting.
But it's not. Did y'all not watch the video, ffs?
AMD fine wine, you must be young. This is not new for AMD.
@@mikem9536 This has nothing todo with AMD "fine wine" bs, and I'm 43 years old and a AMD user for ~15 years. Oh and CPU's have no drivers.
@@Argoon1981 You kids make me laugh...
@@mikem9536 first, not really a fine wine issue. This is a sporadic degradation in performance, not a free boost. Many initial benchmarks will have been roughly at the good/windows 10 perf level, and this is just reclaiming (largely) what has been lost.
Secondly, everyone is blaming this on windows, but wth is AMD doing not noticing this. It’s clear their performance is janky all over, and they are doing a terrible job benchmarking/validating their products. I’m not a AMD hater, I fully intend to upgrade with AMD next based on intels current issues. But nothing about this screams “fine” to me. Both companies have to get their acts together. Even still, I find no evidence some of these new patches won’t break/degrade later on leading to similar losses in perf which is unacceptable. Both Intel and AMD should be working to validate their product with the OS 90% of users install, which is, like it or not, windows.
When 24H2 is officially released, could you make a comparison of a clean 24H2 install vs an upgrade from 23H2 with KB5041587 to 24H2?
Officially?
@@azamatshaimyerdyen6037 currently it's Release Preview. so not yet released to every one.
@@azamatshaimyerdyen6037 Been for a while, lol.
Using Windows 11 IoT Enterprise LTSC here
There will be a new iso @@azamatshaimyerdyen6037
@@azamatshaimyerdyen6037Currently is a preview only
Honestly I don’t know how you can keep up this level of benchmarking on such a regular basis. You must be benchmarking 10 hours a day 😂
haha yeah it's about that, I don't really have a life :D
@@Hardwareunboxed god bless ya mate! You must dream in benchmarks. We appreciate it!
@@Hardwareunboxed please add a Windows 24H2 versus Windows 10 too in your vid please you forget that one Graph.
@@Hardwareunboxedno this is ur life 😅
@@Hardwareunboxed how many fps does that "no life" app/game you reference get using Ultra settings/max RT at 4k.
So AMD's internal testing had variance of windows installs which may explain their "performance gains" from the 7700x to the 9700x. The uplift may just be the difference between a good vs bad install, or a patched 23h2 vs one that wasn't.
Yeah, it's possible, but they should have double-checked the performance with a few different setups
@@davidhiggins2804 I don't think anyone in the industry have discovered this until HW did such a thorough test. Who would have thought that a difference in install would create a 7% gain.
@@davidhiggins2804 if they just use automated testing there is no one verifying consistency.
Or.. (And I say this as someone who have or have had the following AMD CPUs: 7950X3D, 5950X, 2700X, Phenom X6 1100T, Athlon 64 X2 6400+, K6-III 450Mhz)... perhaps.... AMD is just trying to ass-cover so they resort to spouting unverifiable nonsense.
@@Singuy888 Considering the KB fixes the bad install.... I think Microsoft has known this for a good long while.
regarding bigger differences between 24H2 and 23H2 (KB5041587), I noticed these are all Unreal Engine games
Yeah, I was noticing engine is a really key part of this.
Yes. fortnite hogwarts legacy all performed significantly better on 24h2.
Unreal is a monster for unoptimized games by devs with close to zero experience.
I was just gonna add a comment, asking if they could show a graph with game engine vs performance differences.
@@p_mouse8676 That would be pointless, unless it was 2 different engines from the same dev team.
Steve is sitting but also he's bald now. I don't know what to think anymore
It's obvious: standing made him lose his hair, so he sat down.
LOL😂😂
Stressing over the CPU's took a toll on Steve.
steve used his scott herkelman form
It does not render when he is standing. Some aspect ratio bug
So Bald Steve isnt real??
Unsubscribed
Give it time mate.... and not much time :)
Everyone is bald under their hair
@@krjal3038 Everyone is a skeleton
@@al6r725 everyone is a star child
@@HardwareunboxedHey Steve, what do you mean by a bad windows 11 install in the video?
This isn't too surprising. When I taught an CompTIA A+ class we had a room full of identical PCs, with identical HDDs, and CD-Rs of Windows 7, Server 2003/2008 , etc and inevitably one PC or more would just have a borked install. My colleague and I started tracking which CD-R, HDD, PC, etc was having the issue and we could never find any rhyme or reason. The issue didn't follow any particular device, and it appeared totally random. This would never happen in my Linux classes, only ever with Windows. Years later I remember I was installing MS Exchange (2013?) on a fresh server, and I ran across an error that according to MS KB article the fix was to reboot 4-6 times. I didn't believe it, but after rebooting about 6 times in total, the issue resolved itself. Since then, things like this with Windows are pretty much to be expected unfortunately, apparently even many years later.
Honestly, I think much of these issues boils down to race conditions during hardware initialization and driver/registry changes.
That's Windows for ya 😅
Summary: So it sounds like 23H2 vanilla is often depressed (lower) than WIndows 10 pretty much across the board. Plus it can be even worse if your 23H2 gets into the "broken" state. Adding the KB patch/update brings 23H2 up to Windows 10 levels and sometimes even a bit better. 24H2 still seems to have the best performance overall, surpassing Windows 10, and often even better than 23H2 with the KB update.
Of course 24H2 is new enough that we probably don't have enough data to know if you can get a "Bad install" of 24H2 also. So seems like sticking with Windows 10 is fine, or go to Win11 24H2 if you want a perf boost. Avoid 23H2 because there is enough variability that you might end up with worse than Win10 performance.
What a headache.
Honestly for Ryzen Windows 11 has been bad from the start, I remember when I upgraded from 10 to 11 that I lost around 5-10% no matter what build of Windows 11 I had used, this is also true for my Intel based laptop. Will have to check with the new KB as I'm not currently home
How about 22H2? Having the TPM 2.0 disabled in the Bios can prevent Windows 11 updating to 23H2. I wonder if 22H2 could have the same inconsistencies or not.
@@lorsch. Not really, 23H2 could be updated thru an "enablement package" as it was already baked in. I am on 23H2 with TPM and Secure Boot disabled
@@AeiKei can I get this without downloads from third party sites?
Headache started with asymmetric cpus. And this is not going to go away any time soon.
I can confirm once I installed the Dev version of Windows Insider build 24H2 (currently 26120.1542), the system is tangibly faster (7700X).
And not just gaming. It's just overall snappier. 🤷♂
what can be the actual reasons : Better management of cpu priorities ?
@@Ilestun could be putting threads on better cpus, leaving threads on cpus longer to not lose what is in cache, some code path related to making system calls that is optimized for zen4+? I'm super curious too
Because they made changes to the UI. They mentioned it in their blogpost.
Why did he say it sucks?
Not faster, your system was just artificially slower for over a year
Warning: *Do not install the KB5041587 preview update if you dual-boot with Linux*
This is the one that breaks GRUB. It's listed as a known issue in the KB page.
Wow, thanks for the headsup. I assume a manual GRUB restore fixes it?
I have two SSDs, one for Windows and another for Linux.. two separated EFI partitions, one in each drive for each OS.
No problems here.
I press F8 during system boot to choose the OS to boot to.
It broke mine, but if you remove all other drives from fstab then Linux will boot. You can remount the drives once you are booted into Linux again though, I think the drive UID changed and thats why it didnt boot.
Supposedly this only breaks grub when using an outdated version for security reasons. Maybe don't use Debian stable or a similar outdated distro.
@@JohnDoe-ip3oq I'm running nobara, should i just try it and see if you're right? I haven't booted into windows for a while
Yes I noticed that too, when I did a clean install of Win11 23H2 on my gaming rig. Before I was getting 168 FPS in Horizon Forbidden West on average, but now I'm getting 188 FPS on average. That's crazy. Still running a 5800X3D
Would be crazy if Intel CPUs have a totally different result compared to what this video shows
Yeah i tested 5800X3D pretty much as soon as the patch dropped. There were some minor average FPS losses but my biggest gains were from 0.1% lows aside from two games that saw a regression. I mainly tested older games. None of them using UE5. The newest game that i had installed and that i tested was Ghost of Tsushima and this saw on of the biggest 0.1% gains.
Also im no Steve and i dont have the patience for 3 runs. Aside from RotTR all were live gameplay's and not canned benchmarks. I did not lower the settings to 1080p low or similar to create an artificial CPU bottleneck. All my tests were at 1440p max settings or close to it since the goal was to see of i could see/feel gains on the settings i actually use.
Due to the limited numbers of runs, variability of these runs being live gameplays and GPU limited settings there are some oddities. Also when fps is low then even a 5fps gain can mean massive percentages - like in RotTR for example or even GoT.
if i remove the outliers 8both positive and negative) that equates to -1.75% drop in average fps and +5,5% rise in 0.1% low fps
I did see +13-14% often while other number were more rare.
Here are the games and percentages:
Ace Combat 7: -2% average. +14% 0.1%
Ghost of Tsushima: -2% average. +38% 0.1%
Hitman 2016: -7% average. -25% 0.1%
Mass Effect 3: -1% average. -14% 0.1%
Quake Champions: 0% average. +13% 0.1%
Rise of the Tomb Raider: 0% average. +60% 0.1%
Star Wars Battlefront II: -2% average. +9% 0.1%
That's a significant improvement, that raise from hardware would be expensive
@@AnomieDomine Well in this specific game yes, but I haven't tested anymore games. just haven't had any time lately. But I'll test more games when there's time.
As an aspiring ultra budget 1st time pc builder, Am4 system or intel 12400 series seems better?
I would call it performance hindrance rather than performance improvement. Specially since its just the OS holding back the hardware.
Na. And it is not 'just' the OS holding the hardware back. Regression in performance. Improvement in performance.
If a bug or whatever reason is fixed, and causes performance loss? To bad. If a update is bad and regress performance then that is a problem.
Intel and AMD have pushed security updates that loses performance. But Intel is far worse even telling people to disable hyperthreading. Literately making your I7 CPU turn into a I5...
A know 'good' performance should be the baseline. Any regression should be taken as a bug or mismanagement. But we are talking about Microsoft and Winblows here.
Do not expect anything from a company asking Home: $139, Pro: $199 for Windows 11.
Things are starting to make sense now.
LTT's results and your results really can both be accurate, when Windows is behaving randomly like this.
Microsoft needs to be telling us why there's so much variance and giving us real solutions. As it stands, nobody can know if he or she is losing performance to a "bad" OS installation. That's not acceptable.
I suupose it's because they don't even know their own OS. With every update (except maybe for this one) Windows becomes worse and worse. I've given up on it, but still I understand it isn't an ultimate solution, and some people cannot leave Windows or even prefer it with all its weaknesses and problems. Sometimes there's no other options, and I feel for those people. Hope Microsoft do at lease something about their main product.
If games just work on linux i would switch. Try to do it last month but one day everything was working then the next my games would not boot. Spent my limited gaming time troubleshooting linux just to give up go back to window. I spend enough time troubleshooting problems at work and don't want to do it in my free time. One day Linux will catch up I hope.
Microsoft would just blame the SSD vendor. "Nothing to do with us!"
@@joshhelberg I'm sorry you had that experience. Mine is solid for almost two years on one of the most "unreliable" distributions. Though I like tweeking and modifying my setup, so to be fair it should be considered too.
@@joshhelberg Yep, as much as the linux trolls think otherwise, it's still not ready for gaming.
You could take a registry dump from your good and bad windows installs… and then I can’t remember the name of the software, but you can get a delta of differences between the two. Might have a clue.
regshot
@@pingumania690 Poor Reg, i knew him well...
@@malphadourAlas, poor Reg, I knew him well. A man of infinite jest……….
that would be a lot of differences to look at, and the registry might not be the issue. Worth a try? Sure.
Microsoft: Once again, we'd like to thank all of our Windows 12 beta testers. 😉
lol Imagine actually being thanked by a hardware or software company for doing unpaid beta testing of their buggy junk. These days buying hardware or software anywhere near it's launch day is basically a guarantee you'll overpay for a bad experience, it's a joke.
@@K31TH3R It's only a joke, because people keep paying for it.
@@mikem9536 Exactly.
Lol this happens all the time. @@K31TH3R
@@mikem9536imagine paying for a windows key lol
My Windows install is 11 years old. Started life as 8.1. 6 CPU's and many updates over the years. Still runs smooth as butter.
Dare you to benchmark it and then compare to a fresh install.
That was a epic tale Steve! Much appreciated. In place of "Jason and the Argonauts" we have "Steve and the Windows Versions".
Holy crap dude. How long did this take seriously? Top-Tier investigative journalism sir top tier
Tippy-top-tier
Top tier journalism would mean he would slow down with his vids until he knew wtf he was talking about instead of all these "correction video's".
@@mikem9536 I suspect is HAS been MADE acceptable by the shere fact that millions and millions of users haven't fully noticed. Until now.
@@amurtigress_mobile365 He is navigating a labyrinth of bs from these companies. He is releasing videos with as much of more information as anyone else has. He reports again and again as he discovered new pieces of info. It is not a correction as if a previous video was flawed. It is an update based of 100 hours of testing. If he didn't release anything until he had every piece of the puzzle then the video would be worthless as it would all be solved and/or covered by other channels. Your claims are conjecture as there is no evidence of any wrongdoing or mistakes.
34 minutes of "we have no bloody idea what's going on!". I like that one!
The fact that Callisto can run at nearly 300FPS with that level of fidelity is kinda crazy. It stuttered like hell on day 1 but the optimization is pretty good after the patches.
I played it recently, because it was free on Epic Store and knowing of the bad reviews, I was surprised how fun it ended being, I liked it, is rare for me to finish a game this days but I did that one.
But I'm pretty sure what made me like it more, was the fact I could disable many of the gameplay bs, like the quick time events and turn the combate mostly into a single button press for some stuff, meant it became very fluid and that made it fun for me. I still add I few pet peeves with it but the rest made up for it.
Hope it starts getting better reviews.
It was super rushed, you can thank the execs for that. They almost killed the game
@@djixi98 Yeah I know. I've heard that recent Interview with Glen :/
they removed denuvo
Callistro still stutters if you lean hard enough on the GPU. It's like it can't figure out which shader code to compile when and seemingly disregards untapped CPU resources.
It's refreshing to watch someone present just the salient information calmly and clearly rather than breathlessly racing through vast quantities of information like they're commentating on a horse race like some other tech TH-cam platforms.
Point of note, it’s unlikely, but the specific SSD is a variance between those good and bad installs. For the sake of complete troubleshooting, it might be worth spot checking that by cloning that drive to another one.
Yes. Steve have you checked the SSD the bad 23H2 is on ? I believe there are checkers that will also test various locations of the ssd. Agree with @joeynessily that a clone to a brand new ssd would rule this out.
@@Sanbo1119 Good point I wonder the same.
Speaking of ssd’s I love how no one ever turns off power saving for their drives in bios, it makes a difference
Earlier this year, I bought a new SSD and cloned my Windows drive to it. Over the next few days, I got a few blue screens and found a few games that crashed on launch when they worked fine before. At first, I worried that the new SSD was bad, but running chkdsk, dism and sfc and re-installing the broken games fixed all of the issues and I haven't had any since. So, it seems like the SSD is good and the problem was with the cloning. I'm just throwing that out there because it might be a good idea to not just assume that a clone is a 100% match and, instead, verify it after the process completes or at least not wipe the old drive until the clone has been used for a while.
yeah, my first thought as well, sounds like the SSD could be faulty in some way
I have fixed bugs and other performance issues in the past by running the Check Disk Utility (CHKDSK), Deployment Image Servicing and Management (DISM), System File Checker (SFC) commands. These have to be run with as an elevated PowerShell prompt. These fix any corruptions on the drive and system image, which can happen even on a fresh install and with updates. It may have nothing to do with this specific issue, but it is a silent variable not a lot of people know exist.
I have a problem with Windows. DISM and SFC comments: Allow us to introduce ourselves.
Good info, thanks. Commenting to easier find it in my comment history later.
Usually, those commands don't do anything noticeable for the times I've used it. I'm surprised that it actually helps some people.
Blehhhh blehhh
@@opbush5272 it depends on what is actually broken/corrupted, it might fix something you don't even touch, some people get bluescreens on games due to something being corrupted due to a update somehow going "wrong" might not be the right word, but either way they have had those not happen after running those cmd checks and fixes
Sooooo... how the hell do we even establish if our own machines are in a good or bad state? What a faff...
You'd have to make multiple windows installs and test multiple games
I would install the patch and see what it does for your performance.
Somebody mentioned running sfc. And also, let me quote "sometimes needing dism".
I know that sfc stands for system file checker which tests and restores the system files, in case they got corrupted. No idea what dism is.
@@Winnetou17 probably refering to this command: Dism /Online /Cleanup-Image /RestoreHealth and a periodic DISM.exe /online /Cleanup-Image /StartComponentCleanup can free up diskspace on occasion too (can optionally stick /resetbase on the end of that one but read up on what it does)
@@Winnetou17 Dism is almost the same thing, but (long story short): it tests a few other files. The way to go is always: DISM first (as it redownload the image from MS, so if you have a faulty SFC, it is most likely fixing it) then SFC.
Steve, this is bloody amazing work. I shudder to think how much testing this would have been, and you've presented dense, complex information in a super easy to digest and see the difference format. Thank you for all you do for the PC hardware community, we are blessed to have you
Windows can perform inconsistently even between boot-ups/reboots. You cold-start a PC in the morning and find out Windows is performing very sluggish, a simple file copy operation is extremely slow or web pages are not loading up properly. You reboot the PC and it performs normally like none of it ever happened. A real hair puller.
This can be a result of the fast boot setting in windows - it defaults to on, and makes it so a total shutdown is actually not, but a restart will always do a proper shutdown before starting again and as a result clears problems a default shutdown won’t
Good vs Bad install is intriguing. Export the Windows registry for each install to a USB and compare all 4 registries for non-duplicated data before the patch and after the patch. How to do this, I think its possible to export to a TXT file, then open that in Notepad++ to do a diff between the TXT files. Recommend exporting first folders within the registry into separate TXT files instead of just exporting from the root into 1 TXT file.
And do this AFTER running the benchmarks, in case the good/bad settings is something that gets locked down by the first runs of graphic software.
The Windows registry has always been a pile of crap.
Everytime a registry key gets deleted it is only 'marked' as deleted the key and its data is left where it was, even if an identicle key and data are re-created it allocates new space to store it. This is one of the main reasons Windows slows down over the years, each little addition on its own only adds a few ns every access, but do that thousands of times over a few years and it becomes noticeable.
I found this out in the late nineties writing installation scripts for software I had written, installing/uninstalling testing the same software dozens of times over a year for upgrades made my developement machine noticeably slower. This was made worse due to the rather limited RAM on most PC's back then, having the registry file grow from about 2MB when new to 6-8MB when you only had 64 or 128 total. Re-installing Windows (it was NT4 and Windows 2000 back then) resulted in a much nicer responsive machine.
I suspect the reason it has never been fixed in the almost 30 years is that it makes people buy new machines with new Windows licenses. When all they often need is a new/clean install of these poorly designed/implimented OS's.
Now there's atleast one windows 11 build which is faster than windows 10.
So finally windows 11 is faster than windows 10
And has more that 50 % on steam survey.
Both are still slower than Linux
@@ultimate9056 ???? 98% of the games out there u can not even play on linux lol so who cares about linux except some nerds?
@@ultimate9056Irrelevant
@@vadnegru Who all run with Virtualization-Based Security on (which is default) Meaning they are still slower than Windows 10. But people now wrongfully think Windows 11 is faster. HWU testing with VBS disabled is bad. Disabling anti-malware scanner during testing as that can disrupt runs when it decides to scan or update, sure. Disabling default security features that run passively all the time and therefore produce consistent results, no. That's bad to the point of being misleading. People, such as yourself, now think Windows 11 is faster while it's really not. You have to disable security features to make it 2% faster and even HWU does not recommends doing that.
Bald Steve = Steve given up and tore his hair out with this cr*p
Bro censored crap
@@torvasdh yupp. During my early experience in TH-cam my comments got deleted due to some reasons. Got into a practice of censoring stuff like this and it's now automatic.
😁
Windows installs have been broken since about 2020.
Literally every fresh install should run sfc. And it'll be faulty every time
Sometimes needing dism.
It's wild
you will use a wrong version it never happened to me in the entire machine park
its not always broken, but i do recommend after a system update that people should run sfc /scannow and a dism scan to rule out a bad update/install that could possibly happen.
@@CoCo.-_-that's exactly what I've gotten in the habit of myself. Too many faulty updates recently.
I also tested this exhaustively and found no difference. The culprit is in exploit mitigations. If you disable exploit mitigations, you can see these same performance gains on any version of Windows 10 or 11.
From personal experience after installing 24H2, I had an increase in my Valorant and CS2 FPS by about 12% and 16%. Which for me, is the difference between being able to max out my refresh rate of my 520hz monitor and not. Before I was getting around 480-550fps in Valorant, but with constant drops to 350-400. With the 24H2 the lows never drop below 500 now, so the lows have gained A TON for me. Which means that it's so much more smooth on my monitor. Def recommend installing this for people. I have a 7800X3D. CS2 lows improvements are even more drastic for me.
Did you have 23h2 with the update before going to24h2?
What are your settings like for cs2? Would you be able to share a screenshot? Cheers!
@@diktaf Everything low except textures (high) and shadows (medium)
@@z_wulf sweet. Thank you for such a prompt response!
No AA? Man that's tough to get used to.
@@diktaf No AA, though that depends on your graphics card. If I use 4x my lows drop below 520 and it's noticeable with this monitor, but before when i was using a 360hz, I used AA, because I didn't care if it dropped to 400fps. AA is all GPU, and since I only have a 3070, I've found that turning it on gives me random GPU spikes that drop my lows. But if you have a 4070 or AMD equivalent, you can def run AA just fine.
Quiet funny that if the new processors performed better this problem may never have been picked up...
That's a good point!
Or... the new processors could always perform better but Wintel Windows was holding them back.
Great work Steve. Explains how an outlet could show much bigger gains for Zen 5. There Zen 4's had "bad" installs and their Zen 5's had "good" ones.
Is it possible that 24H2 could also suffer from the "good and bad" situation, considering there were times it had only matched KB5041587 on the bad Win11 install, and the good install running the patched 23H2 pulled a few percent ahead of everything else? It could also just be run to run variance I guess, but still.
I'm not sure. We will check once it's released.
@@Hardwareunboxed Yeah guys not getting the point like with OLD windows it used to be bunch of files which used to get installed and used to be consistent. Now that AI is in the play it does as it wishes and installs the way it wishes depending on it's mood.
Before people start giving mea earful about how AI doesn't play a role during installation. I want to make it clear that this was a sarcastic joke on how AI is pushed into every dumb stuff nowadays where it doesn't even make tiny bit of sense.
@@KING_DRANZER AI is also pushed when almost nobody is asking for it, while things that people are asking are not done. Well, this patch is one exception.
Also, here I'm saying it again, since I can't abstain myself: if AI is so good, then why isn't Microsoft using it to finish the Control Panel to System App transition ? It's been "only" 10 years (literally) since that started, and it's still not done and it's a huge mess (best example I have is the power profiles which you still need to dig for into the Control Panel for exact tuning, which is very frustrating for laptop users, especially those trying to test battery life) and extremely stupid.
@@Winnetou17 I remember seeing a reddit post a few years ago where people talked about Windows 11 and generally how slow Windows development is, and why it has so many inconsistent UI elements, and allegedly there was a leak of an internal chat between developers where they freaked out and got angry how they have no idea how the old code works in control panel and that it took them days to change like 1 UI element in control panel because the smallest change broke a bunch of other things in Windows and they had no idea why. It was agreed on that post that this is probably the case, because original Windows devs from decades ago left so new devs can't really deal with the decades old codes that are there because of backward compatibility and bad optimization (they keep adding more and more new code to not touch and destroy the old code) making development super slow.
@@canyongoat2096 Oh, I can totally understand why it's slow, I'm a programmer myself.
Still, 10 years is A LOT. And the way they are doing this is also annoying, as settings that are still important are much harder to reach now.
Really, there's no excuse why Windows 11 hasn't launched with everything ported to the Settings App, so the Control Panel could be removed completely. Or at least, still keep it easy to use, if you don't remove it. Right now it's the worst of both worlds in plenty (thankfully not all) of cases.
Very much appreciate the Windows 10 testing as well, not many (if any) larger channels cover Windows 10 anymore, and these issues would be lost in the shuffle.
Can’t believe it took Microsoft this long (W11 24H2 official release will be less than a year away from end of W10 support) for them to reach performance parity with W10. Can finally feel like I’m not downgrading my machine when W10 sunsets and I move to W11
Welcome to the wonderful world of PC's where random shit happens for random reasons
That’s not random, they just don’t have enough “ ocd “ lol
Pretty much like sometimes I was able to fix most error with just pure luck
So, that takeaway is Rainbow 6 can now run more consistently at 500+ fps on monitors that cannot.
High fps still matters for input latency even if your monitor can't display all those frames. The faster the game renders, the newer and most recent the frame will be shown by your monitor when it refreshes. Watch 3kliksphilip's video on it.
Yeah but input latency isn't a major concern for those that aren't into competitive e-sports gaming
@@MitchellTheMitchtrue
@@xFluing Input latency is meaningless for anything above the 200-300 fps range. We're talking fractions of a millisecond here. For reference, the average person response time is 200 milliseconds.
@@mikehawk6918the reaction time doesn’t matter since latency adds up. But you’re right over 200 to 300 it’s diminishing returns. Unless you’re buying a 360 or 500hz monitor
I thought we were getting Bald Steve Arc based on the thumbnail right after the Standing Up Arc 😂
No, that's just a teaser for the Intel Arc.
Thank You Steve. Incredible work as always.
Here's the difference in percentage.
22 Games Tested.
Windows 11 24H2 ----109%----
Windows 11 23H2 Good Install KB5041587 ----107%----
Windows 11 23H2 Bad Install KB5041587 ----107%----
Windows 10 22H2 ----105%----
Windows 11 23H2 Good Install ----100%----
Windows 11 23H2 Bad Install ----95%----
Can't wait for 7800x3d & 5800x3d 24h2 test video. 😊😊
So KB5041587 is a "cumulative" update, so unless you can narrow it down to one or a couple individual patches it's hard to tell what the culprit is. I'd expect some security mitigation patch though.
Your "broken" install could be either missing one of the KB's that the "unbroken" has - or vice versa. Microsoft likes to sometimes test some patches on a limited run off machines, so if you run any windows updates, that sounds like the most likely situation to me. There is a way to just print all installed KBs, would be the best place to start. Anyway the KB5041587 package would then install the missing patch.
Would love to see a video on the difference between the upgrade to 24H2 or a fresh install of 24H2. Upgrade a clone of the bad and good 23H2 installs to compare maybe.
Reminds me of how back around 2000 I found that, for reasons I could not figure out, fresh installing Windows 98 SE on a hard drive laying flat gave me worse FPS in Crimson Skies than when I installed Windows on the same hard drive standing on its side. Everything installed identically, I got noticeably more FPS from one install than the other and I had no clue why, but I installed over and over to the point I could recite the Windows key and the results were replicable.
Hmm, maybe the drive was writing more clearly in one position, leading to less error correction being needed upon reading the data back later.
It could be that the hard drive was overheating when laying flat. I have faint memories of having issues when I mounted internal drives right on top of one another, with barely any airflow between them. Also, much more recently, I found that my portable hard drives overheat and give me abysmal performance (like
I thought Crimson Skies was an OG Xbox exclusive. Had no idea it was available on windows.
@@levygaming3133 I'm pretty sure that it was a Microsoft published game, which explains the Xbox exclusivity and it being on Windows.
Great work! Now my conclusion: stay on Windows 10 as long as possible, until Microsoft forces me to switch 👍
yet the new win11 24H2 upcoming update seems to get you more fps then windows 10 on most of these games
When support for W10 runs out i'll be switching to Linux. I've had it up to here with MS. Every few updates my second monitor stops working for no reason, and a few days later all is fine again. Things that should be simple, like inverting left and right channel audio is not possible in Windows without 3th party software. No way i'm getting on the W11 or W12 train! I've lost so much time troubleshooting MS these past few years that i might just as well install linux.
@@kiamppand the new explorer/shell is laggy as shit. Every folder you open loads for a quick moment, it’s like the Win10 (or Win11 22H2) explorer on a >10 year old machine
23H2 and 24H2 performance is almost as unpredictable as which syllable of 23H2 and 24H2 Steve is going to emphasise.
Wot
@@TheSometimeAfter Normally you would stress Twenty 'Four' H Two to note the change from Twenty 'Three' H2. But in these videos Steve has often been stressing Twenty Four 'H' Two, 'Twenty' Four H Two, etc.
I imagine its because hes said it so many times the words have just about lost meaning. Just a mildly amusing side note.
It's never the part that actually changes 😅
Never crossed my mind we would be benchmarking windows on the same hardware. Mad...
At this point, with all the benchmarking of the past few weeks, Steve is not making coffee in the morning, he's benchmarking the coffee machine.
Have you tried doing a file-by-file comparison between the bad install and the good install? Just calculate a SHA256Sum for every file on both installs and stuff the results into a database. Then you can select the results where a given file path has more than one checksum. This would tell you what files are different between the two installs.
It's likely some registry settings that isn't getting applied correctly.
Yep, this is the way. Got to be something that hasn’t updated on the bad install and updated on the good install or it could be the other way, but I lean towards the first one.
Just take both drives, put them in a motherboard or enclosure running another install of Windows and do a "File Content" comparison of the two drives using FreeFileSync.
Much easier...
@@tim3172 perhaps its easier if you already know FreeFileSync. Personally, I've never heard of it and never have used it. I do however have a database running on my NAS and several programming languages already installed so for me it's easier to spend a few minutes writing a program than to properly vet and install an unknown 3rd-party file utility.
So, the lesson is to stick to Win10, which is exactly what I have !!!
Or leave Windows for good, if your usecases and preferences allow it!
Windows 11 is slower that Windows 10 up to version 24H2.
So it would be worth it when 24H2 would drop ;-)
9:00 As it has been noted on the forums, this is a problem in the new Nvidia driver, and I have this problem myself, the performance in the 537.58 driver is much better than in the 560 and newer
So many data points. How do you stay sane looking at that Excel spreadsheet?
and that's why it is a good idea to stay on previous windows until it dies.
if you are on Win10 stay until 2025.
and hopefully until then win11 will be good
No, then you Jump to 12
Or Steam OS 3 (if valve bothers to release it) or Linux (Pop or whatever)
Just a small nitpick about wording - you're disabling _virtualization_ in bios, not _virtualization based security_ . The latter _uses_ virtualization.
I run VMs for work on Oracle VirtualBox. I think I need virtualization turned on in the BIOS. My Windows SAYS VBS is disabled when I query via powershell and none of the HyperV stuff is installed at all (Windows Hypervisor Platform, Virtual Machine Platform, Hyper-V are all unchecked in Windows Features).
It would be nice to know if my Windows is performing up to snuff but I am not really sure how to tell. 5800x3D.
@@samhunter6813 benchmark with your current settings and then benchmark with virtualization turned off in BIOS to rule out anything if you really wanted to
@@samhunter6813 There is no issue with having virtualisation by itself enabled. There is no performance regression with it (ofc, unless VMs are running in the background). And even with VMs running, I found that the issues I thought were VM related had other causes (for some reason my mouse started introducing stutter in some games at higher than 250Hz polling rate).
To make sure, you could enable VBS, benchmark your system, disable VBS, benchmark again. There should be a decent performance gain when disabling VBS.
@@CoCo.-_- Good news, SVM in my BIOS makes no difference.
Would be cool to see a 5800x3D vs 7700x vs 9700x....just to see how stagnant that price segment is....
Thr 5700x3d clowns on everything for being 95% of the 5800x3d and being 200 dollars
@@james2042 my 5700x3D undervolted + 3080 is on par with a 7700x and 4070 according to timespy. I aint changing til am6
That seems like the sort of thing that would be better as part of the 7600X3D review.
@@james2042 bought mine 135 euros (and sold my 3700x for 85 euros), it clowns even more on everything for a 50 euros upgrade. Definitely the best performance/price ratio.
AMD and their AM4 socket is definitely goated
I think the idea is that bald Steve was him just pulling his hair out in frustration. Thanks Steve.
lol
microshits bs is pathetic. As a windows modder and a gamer, I tend to reinstall windows quite often, and I've felt that some installs just perform worse in games. I've had that feeling all the way back when windows 10 came out. I'm glad I'm not going crazy, but I'm infuriated at M$
It makes you not like gaming as much because there are so many loop holes lol
It's baffling to me that every youtuber seems to keep overlooking the fact that 24h2 is a complete OS replacement update. Unlike the updates we are used to this 24h2 update replaces your current version of windows and doesn't actually updating anything. It's more like a fresh install than it is applying anything over your existing os. It's clearly stated in the update description and should easily explain why the 24h2 update performs better than the os the update is applied to and why it can also be similar to clean installs of older versions of windows when they are clean installed. There's also performance impacting security features that will automatically be enabled when they weren't in older versions as well as some settings being kept if 24h2 is applied as an update. Performance impacting things that may have been turned off by the user previously will still be off while an actual clean install of 24h2 will have them enabled by default. Considering how many times simple things are over looked I do have some doubt in things just being completely random like claimed here.
No wonder this video took so long to make, the amount of effort put into this test is staggering.
On my case, I have been using Windows 24H2 for over a week with very good results. There are some bugs, but performance is really good. I always ran SVM disabled.
Most UE4 games got a 10-20 fps increase.
With Remnant 2 I got a 10% increase.
Subnautica went from 70-90 fps, to a locked 144 fps.
Horizon Zero Dawn and Serious Sam 4 had close to identical results.
Unity and Unreal are majorly impacted by this?
That's like 80% of the games released today.. this should be a (literal) game changer for everyone on 11. All 30% of them.
At least those of us still on 10 aren't suffering as bad, but having improvements too would be welcome.
Always expect you're not getting the performance you could be, simply because software optimization went the way of the dodo when people stopped needing to program their own engines and everyone switched to bloated high level libraries and visual scripting.
@@AceStrife I also tested with My time at Portia, and that game also got a huge increase in performance.
UE4 and 5 games seem to get a good boost. But Unity game seem to benefit even more.
Also tested Wo Long: Fallen Dynasty, but this one got no improvements. It uses the Katana engine.
Here's what you do. You do a file compare of those install drives to see what's different between the installs. You get a 3rd SSD. Install Linux on it. Install the other two SSDs as well, but don't boot to either of those, boot to the Linux. In Linux mount both drives with the Windows installs and start using tools to compare the two drives to see if they're different. Get a Linux scripter to build you some scripts or Google some tools. Then you can see if there are files missing or added between both. You'll also be able to see if some files are sized differently. Any suspect files, you Google those and see what they're used for and then track down what's different between the two installs that way. I only recommend Linux to do this rather than a 3rd Windows install since Windows might put random files on both drives
Looking foward to windows 10 getting the patch and take the crown once again.
It’s kind of wild that the hardware market has been so lackluster lately that one of the most significant upgrades you can make to your PC is a fresh install of Windows
I'm looking forward to my 7700X getting some free performance... not that it needs it, lol.
Some thoughts on the variance between Windows 11 installs:
* Windows 11 is very network chatty onto the internet; are these tests done offline, and if not, it might be worth investigating?
* Was the "bad" install before the "good" install? Modern AMD Zen processors "learn" things at a low hardware level that would survive any kind of OS and drive change. This unfortunately means that it could even result in slightly different data depending on the order of games test (or rather that changing the order could change results)
* Do those SSD's have MLC or TLC NAND? MLC would give the most repeatable results.
* Run the following in an administrative Command Prompt to ensure Windows is as healthy as possible (this can be a problem even on fresh installs and especially after installing any kind of Windows Updates):
sfc /scannow
DISM /Online /Cleanup-Image /ScanHealth
If the DISM line results in something like "the component store is repairable," run:
DISM /Online /Cleanup-Image /RestoreHealth
Restart the machine for the corrupted system files to effectively be repaired, and then the testing can begin.
Either way, all this means that:
* Game testing on Windows is probably only going to get less and less repeatable (less precise) as it gets more sophisticated
* Same for CPUs and probably GPUs as well. Branch prediction alone already introduces some slight variability.
* This will then require even more games thrown out if they exhibit too much run-to-run variance and install-to-install variance, or just mentioned for transparency but not included in actual performance metrics since the data is "untrusted."
* All this variance / lack of precision will mean that larger sampling is needed to ensure that a single or a few outliers don't heavily skew the data. Of course, as already said, the new patch results in a wide range of results depending on the game. At least by increasing the game testing array, it improves the odds that a real-world user's experiences are reflected.
* It probably needs to be verbally stated more often by tech reviewers not to take any results as gospel; to research performance based on data from multiple different sources; and to try and put the data and results in perspective in the real world. For example, while disabling MS Defender is wise to improve consistency, this shouldn't be done in the real world under normal circumstances. The goal of reviewers is to eliminate as many variables as possible with the caveat that not EVERY variable can be removed.
I have a theory - Not sure if you will see this as there are many comments. For good vs bad install, it could be the polling rate for USB devices connected to the PC. As someone who has been playing around with polling rate for years, this is something I am very well aware of, but it might not be the case here, just throwing it out there for perhaps further testing (by unplugging before tests, or using VERY OLD mice and keyboard for the tests. Sometimes the driver being used for the keyboard and the mouse randomly uses a different polling rate. If a mouse or keyboard runs at 8000Hz or 4000Hz polling rate, it will impact CPU performance.
This sound eerily familiar to me. My mouse introduces stuttering in some games once I set the polling rate higher than 250Hz. Incredeable!
But, if everything is the same every time with the exception of the bad install then how can that be it? I would expect the SSD if that hasn't been ruled out. It could just be something random happening on occasion causing this. I remember windows 95 installing differently every time with the interrupts being assigned seemingly at random.
I'm on the Windows 10 LTSC till 2032 team. By then maybe I'll try Linux. Excellent work here! Thank you!!!
That's MS windows since the 90s. One system version is good next one is to be evaded. 95/98 good, 2000 bad, XP great and so on till today.
2000 was good, ME was bad
I though 98 wasn't that good but 98SE was good. The next one, ME, was bad while its contemporary NT based version 2000 is good and was used on a lot of business and embedded systems even into the 2010s in some cases.
every other article i see about win 11 is a regression
Calling Win2k bad is hilarious. It was the most stable one yet.
I even use it for my old Pentium II Laptop, distraction free ms word machine regularly.
Windows 2000 (WinNT based) was great…. Windows Millennium (ME, 9X based) was very bad.
I watched your video and immediately installed a fresh windows install from the 24h2 iso on a RUFUS usb drive. FOR ME... everything runs better. FPS on average increases between 5-8% playing on my 4090. I do play at 4k so performance improvements may even be better at lower resolutions.
NOTE Ryzen 9 5950x user here
While 5% difference is hardly noticeable , it's so frustrating how bad windows has become
Steve, don't wanna open the Pandora's box, buuut, can't help but wondering what would win11 vs some of the linux gaming versions look like! Also, standing up gives another fps boost.😅
Linux is slower and less compatible with games, not sure why that needs to be tested.
@@mikem9536 Eh, less compatible isn't really in question as there are still games and rootkits that don't work well or at all.
But slower?
That's very variable and can't really said as a blanket statement. Most of the games I play, RPGs and strategy titles generally, tend to actually run better.
The DXVK translation is a boon for some titles and CPU performance seem better under Linux at least for me.
@@mikem9536 Linux tends to be faster, even with the needed translation layers, but it depends on the distro and kernel. there are a few tests from the German outlet computerbase and pcgh, results are really impressive
@@mikem9536 i mean, you could say similar things about intel GPUs but we all agree those are worth testing
@@mikem9536 Level1tech would beg to disagree.
Steve is sitting guysss
i bet he is sitting because of a hemorrhoids pillow
When hes sitting, everything is okay. It is the way.
From Zen 5 sucks to Windows 11 sucks
interesting....
I mean, both zen 4 and 5 get an uplift from the patch, so there's still poor gaming increase for zen 5. I'm looking forward to getting some zen 5 CPUs in the server racks, but I won't be running out to upgrade my gaming rig anytime soon.
Zen5 still sucks for gaming
Zen 5 still sucks
@justsomepandawithinternet It doesn't suck for server work or productivity, which a LOT of people do. It's a good upgrade for them. It's got a small increase for gaming, but it's an increase, so it doesn't really suck from that aspect. It's the current pricing vs a 7000 series processor that sucks.
Windows 11 always sucked. That's not new.
Dear goodness, I'm impressed and in awe of the amount of testing you've done for this one. Thank you so much for providing this attempt at clarity for as far as possible. Much respect. Have a good one 👍.
I installed the patch just because, knowing anything about what it does and I noticed CP2077 was running a lot smoother, even in DogPoopTown and I'm on Zen3.
13:45 Steve says it has been suggested Win 10 will also get patched. May I ask, by who? Official Microsoft or AMD source or just rumours?
It already did. It is Windows 10 optional update KB5041582. However, I am unsure if Steve's tests here include this update.
@@ericbanker1285 I looked this up on Microsoft, the description page says nothing about fixing the scheduler for AMD or anything similar. Did AMD say this update contained their new code?
@@bakakafka4428 I tried replying earlier, but my post seems to have been blocked, possibly because I tried to directly link benchmark results.
A Reddit user posted his benchmarks on the AMD subredddit that found the similar ~3% boost in a specific benchmark from both Window 11 KB5041587 and Windows 10 KB5041582.
The Microsoft page for Windows 11 KB5041587 likewise says nothing about fixing scheduler for AMD, despite benchmarks showing improvements.
Hey Steve, I don't mean to be unappreciative or anything, but we really need you to add both the 7800X3D and the 7950X3D to these charts, because 1) the 7800X3D is the best CPU a gamer can buy, and is also affordable, so a really strong recommendation on top of it already being very popular in sales, and 2) the 7950X3D is a CPU that is incredibly vulnerable to the whims of Windows, so your work is undoubtedly valuable in these videos where you're holding Microsoft to account for us gamers who needed to get a 7950X3D for the non-gaming workloads. 🙏
So, everyone telling me to move to Windows 11 for the past couple of years because "it's just better" and "what are you, a caveman" was wrong? Neat!
Correct, Am the outlier as I've been telling my friend helped him build a 5800X build in 2020 to stay on 10. Idk how many 11 installs I went through (still have a 10 install around), 11 drove me back into the arms of Linux for gaming and HTPC the abuse got bad enough. 0 issues from that friend on 10 though he's been stable for years (is a poweruser).
and yet in just 1 year Microsoft will stop supporting Windows 10
According to these results Win 11 24H2 should be the best gaming OS - find it hard to believe it beats Win 10 but the data is the data. If you're going 24H2 might as well use the LTSC version, which you can install MS Store on too should you need it.
Thanks Steve. I cannot believe the headache this has been, but that's some nice data.
I'd love to see a comparasion between 24h2 and linux
Windows spyware and bloatware requires lots of resources as well. Youre expecting too much.
You can install KB5041587 update also on WIN 10.
if I’m not mistaken the lowest is windows 11 22h2
I don't know about that, but I'm reading that there's a KB5041582 for Win10 that provides similar gains. Maybe that's what you're thinking of.
@@Osprey850 I have to look this up. of the gains are good then I might just go back to windows 10 for the rest of the year lol.
changing builds every other month is very frustrating 🥲
It should not be up to the consumer to have to roll an "installation lottery". Microsoft sells Windows 11 as a product, the consistency of that product is the their full responsibility.
Hopefully though, it seems like these patches have basically eliminated this lottery behavior, but only time will tell. Still shouldn't have happened in the first place.
What a hot mess! Who knew I was going to be looking at benchmarks based on how good or bad an OS install was? Wonder if the PC is in a black case vs a pink case will make a difference.
Of note: I don't think there's any level of attention-grabbing thumbnails or titles with vitriol thrown at windows that could upset the community. Windows 11 does, in fact, suck and I think even the handful of people who say they don't agree do still actually agree.
Zen3 is not dead, give it some love..
It's dead, homie. Pour some liquor out for our boy.
@@blue-pi2kt kkkkk, good one
@@mcury85ah essa risada kkkkk... br detectado 🤣
@@BrunoUtech detectou corretamente :)
Zen 3 may not be dead, but it smells funny!
😂😂😂
We should get Steve a standing desk.
HUB is absolutely deserving the Subscriber Growth they are receiving. They are class leading in their reviews.
Back in Xp days, Some of us would strip the Xp Installed down to 200-500Mb, removing all the fats and bundled bloat.
Now this was when only HDD's existed and Ram at most for the avg person was 2GB most people where 256MB/500MB or maybe 1GB
By stripping Xp down, one was able to make it run on less Ram and faster, and also the HDD access/read/write was faster.
I would not be surprised one bit if the same is for Win 7-11
What we really need is to have most components optional add-ons that install during setup there should be 1 Version and everything else is just profiles (Home/ Pro / Gamer/ Education/Enterprise etc etc) of what add-ons get installed. And while each has a default list of add-ons select you can unselect them and not have them installed
It’s legit still the same, I had to do this same process 3 days ago lol
@@silfrido1768 Are you using NTLite or manually stripping it with CMD+Scripts?
@@scarletspidernz a little bit of both.
I start off with NTlite since you are basically able to do alot of tweaks that you would normally do while windows is running. Then I finish all my other tweaks while windows is actually up & running.
The trick is to not disable critical services since it can brick your windows.
another thing to take into account is how you lose performance over time from updating windows often. correct me if I’m wrong but my take is that windows gets corrupted from all the left over updates.
I turn off windows updates now lol. It’s getting to a point where I need to just have 1 copy of a back up registry to save even more time
Downloading 24H2 right this second
Gonna do it right when I get back home
Gonna wait for official release but looking forward to it
Glad to still be rocking my trusty windows 10 till it totally losing support, no win 11 any time soon for me!
honestly the support is a non issue. people rode out vista and 7 on xp with no issue
I didnt skip, thanks for the hard work
In no way an expert or knowledgeable but...
The recent discussions about downgrade attacks (see Alon Leviev) show that windows does not have an overarching integrity system with updates merely using predefined steps. If a "step" was missed, hit an error, etc. and did not have a custom integrity check within the update (the catalog does have some rudimentary checks that trigger rollback steps) 2 systems may behave wildly different after applying updates. It seems likely that an upgrade to 24 would either do a massively expanded integrity check or just touch enough files that problematic system components are likely to be overwritten.
Try comparing the checksums of all files in the Windows folder between the good and bad install. Rudimentary check but it's a simple one to do. There are built-in windows utilities that could be used without a reference filesystem but they are incomplete and Levievs results show they make many assumptions favoring speed over accuracy.
Been looking for a 24H2 vs Win 10 comparison to decide whether to side-grade from Win 10. This is exactly what I needed!