Yes. Anecdotal means not necessarily true or reliable and completely subjective and not based in fact. Thanks. I’m an AMD user and I love this channel.
The problem is that 95% of gamers are not interested in spending 3 days tuning a cpu and possibly finding out its still unstable after. Reviewers are not wrong testing the out of the box experience as thats 95% of all users. Tuning and channels that revolve around tuning are an extremely niche population. Reviews need to be geared to the setup most people are going to be using. In that case the 7800x3d still offers the best out of the box experience for 95%+ of gamers.
Chaser does really underestimate how much a average Joe wants to configure / optimise their system. Mainstream could do a better job of a dirty overclock section in their reviews .
@@mushbeck Well, yeah, but when I was first trying to OC I found a program called !xSpeed. This OCs Windows and Windows software, but NOT hardware. Don't even ask me how it does; all I know is that it worked on most software I tried. Anyway, that whole saga didn't end well. I eventually got convinced to remove dibeng.dll, lol. I think most people learn how to OC when they learn the other stuff first, and tbh this not-aggressive gatekeeping is probably for the best. Anyone who wants to find it will.
I immediately had to look out my window to see if there were any pigs flying by. Much respect for just calling it how it is, Literally the ONLY channel who does this and it's the main reason why I've been subscribed for a cpl years now. I always know that I can count on Frame Chasers for raw facts W/O any BS
that was also my criticism of all the reviewers that they never showed whether the AVX-512 values have improved, because one of the main advertisements for the 9000 series from AMD is exactly that point. never saw a PS3 Emulator (RPCS3) speed test, like u always said, most of them not are gamers Zen 5 improvements compare to Zen 4 The L1 cache per core is increased from 64 KB to 80 KB per core. A ~16% IPC uplift on average, claimed by AMD. Memory speeds up to DDR5-5600 and LPDDR5X-7500 are officially supported. Infinity Fabric clock (FCLK) has been increased to 2400 MHz. New 2-ahead branch predictor AVX-512 capabilities have been expanded with Zen 5 with a doubling of the floating point pipe width to 512-bit. Additionally, there is greater bfloat16 throughput which is beneficial for AI workloads.
Listening to Jufes talk about how much he likes an AMD CPU makes me feel like I'm in Bizarro World. He's always said he tries to be an unbiased channel that offers real info, but he ripped AMD so much people just figured he's an Intel shill no matter what he says...this video seems to prove otherwise.
He likes what works and the moments when he feels he is being fairly treated by the company. But yeah, many AMD fans are so religious about it, that I think Jufes just loves to rub in their cognitive dissonances a bit. The x3d chips were weird topic here. Mostly because most of techtube tests things at stock settings and x3d shines with those. Then most tech beginners get the idea that x3d is the only way to go. Their OC and tuning potential is very low though (for mostly the same reason why they are good out of the box) and that is why they do not compare as favorably to tuned systems. That difference has been very hard for the AMD community to accept and a lot of semireligious discussion has risen from it. Prime grounds for a somewhat heated discussions.
So this is why 9950X requires a scheduler, compared to 7950X that does not. AMD is covering up the fact that they are putting one bad CCD in that thing. So lame...
Yup.. This was the same for 5900x and 5950x... One good die and one Med or shit CCD. Some people also get really screwed over when there was the covid chip shortage and both CCD's on a 5950x was only slightly faster then 5900x.. :(
The big reason for that is that the cross-CCD latency is 2.5x as much as previous gens. If games use cores on both chips, it definitely hurts performance.
no really amd has had issues windows scheduler for a while now. but all the cries that windows users are having are non-existent on linux. but I am not saying peps need to switch to linux but windows has always had issues with cpu scheduling for a long time now. but then again intels mixed cores were a tire fire when they released it and they worked w/ MS to make the Ecore&Pcore thing work poorly at minimum.
bro most people are not going to adjust 50 settings if its 3% faster in default settings vs old gen thats what reviewers will say and their not wrong its not on the user
if you're leaving that much to be acquired over 23% in tuning comparing to tech jesus himself, is insane your opinion is wrong but let me tell you WHY it's wrong. You get what you pay for and if you paid for the best you should get the best and if you buy the best tool for the job it better do its fucking job.
Well, the idea is that configured right, this thing is genuinely great for workflow and gaming, without X3D dual CCD problems from a 7950X3D. That more performance is genuinely achievable for those who seek it. Again, this is a channel directed at people who are going to tune when they get their mits on something.
I think the buyers of this CPU are people who are akin to tuning and getting max performance. If you are buying products that cost this much, you are the kind of customer who love max performance irregardless of parts cost in time and money. These are for enthusiasts and PC builders. I love watching this channel because FC only caters to the top 0.00001% of pc builders. Almost all of my friends who subscribe to FC and his Patreon are buying all the products he tests. I have bought every version of 5x00x3d 7800x3d 13900k/kf/ks and 14900k. Dude everything that Dufus has reviewed. If you listen to his content, you would realize that tuning is the real deal when working with these high end products. It’s so extreme that even tuning ram is important. buy the FC ram and you’ll see how tuning really changes your system and 1%lows. I hope this helps and also makes you excited for tuning as well.
Actually the 2 CCDs are not an issue for me cuz I like to VM it on Proxmox. If you have a partner, you can cut the 9950X in HALF making it a 9700X for both of you. Now, all you need is 2 monitors, 2 GPUs and have fun playing on a single machine. And another way is, you can still run the 1CCD for gaming, and the other slower CCD, just run a local LLM like Llama3 or idk... maybe a video encoder or something... You got 16 cores with fully enabled AVX512, so push the multitasking to the limit! That CPU should be excellent for RPCS3...
RPCS3 only primarily uses 128bit AVX512 instructions so, as demonstrated by TechPowerUp which benchmarked RPCS3, it's within margin of error relative to the 7950X (though interestingly the 9700X was a bit faster than the 7700X to the point of not being within margin of error?) For reference you can read more info on this from the recent blog post from the developer of y-cruncher about Zen5's AVX512 (a simple ctrl+f for "rpcs3" on the post will take you to the corresponding text).
Pbo +150. -33 Curve x5 very cool, low watts and great performance on my 9900x I also have approached the 6ghz for fun although dialled it back for daily use and longevity. I notice the "mainstream" often tell people to avoid the CPUs I buy.... In-fact you can achieve benchmark scores (Cinebench etc) only attainable using liquid nitrogen or water chillers with 7950/7900x with just a well put together custom WC loop with this 9000 series. A very fun CPU for those that like to tweak :)
Always the first manufactured Dies with different process like zen 5 its on 4nm with dual ccds have much bigger gap between 2 ccds as nod improve with the time this gap in Mhz will be closed a lot, that's happen in all AMD cpus with 2 ccds as release it, also the first ccd will improve with all core boost close to max speed in a future, but is much better than intel as 14900K have only 3 cores of 24c to boost at max speed (6000 mhz)
They are doing it 100%. bought 3 9600x and all max out between 5450 and 5550 all core. Then again, zen 5 at 4.9ghz matches zen 4 at 5.625ghz so there's more to the story than "AMD using crappy ccds"
The very principal behind the CCDs works out in how workloads end up being handled. Your light loads will shuffle around your highest clockimg cores. As the load increases to more threads those clocks collapse anyway. There have always been inter-CCD issues in games, because the other is basically empty in those moments when the thread jumps.
Remember in the late 90s hitting 300 MHz was a significant technological achievement for cpus which costs only a few hundred bucks. So just think about it that way
I'm certain all the best chips go to the server market. Reminds me when the consumer market had a larger slice of the pie and Xeon and Opteron chips fit on consumer boards which also had the best OC.
Zen 5 is only legit if up coming arrow lake is crap. I just don’t want to regret after 2 months when arrow lake releases. At this moment, I would say just wait. BTW, I can live without 6GHz @1.55v.
14900k 5.8 at 1.34 lol. 6.0 won't change anything for gaming lol. Test it. Arrow Lake Leaks are already out. You only need a 10900k. My 4090 is my bottleneck on my 7800x3d and 14900k maxed systems. I'm nowhere near cpu bottlenecked, and would be fine on a 10900 lol....
@@Hardware_Neutral That's irrelevant. CPUs not only need to not be a bottleneck, they need a lot of compute in reserve to take care of potential stutter in intense scenarios that are barely tested by benchmarks.
I get it that you don't want to share the secrets, but you should've mentioned more of the reasons why the mainstream techtubers are wrong. Basically everything you said boils down to scheduling = bad. Which is a total non-issue for 9700X and 9600X, which also had subpar results in their reviews. So that can't be the only reason why Zen 5 is .... weird.
9950x also costs ~ $200 dollars more than the 7800x3d and MOST people can't be bothered to spend hours upon hours of fine tuning a CPU that SHOULD give the performance experience you were able to message out of it, out of the box (in a perfect world right). Anyways the 9950x is a really interesting CPU but after a 6-year hiatus from pc building I just built a new pc and went with the 7800x3d for all of the obvious reasons (wanted easy ;) and couldn't be happier with it. I do like to tinker though and with minimal effort in the oh so scary bios 😨 it's incredible how efficient even the 7800x3d is/can be. Max load, yes water cooled (not direct die f that), I NEVER exceed 83C. Just ran a single clip CR23 while this vid was playing with 10 tabs open in Edge and COD minimized, 5050 all core, 1.07v average for cores, 83c Max TDie, 78w package power max; 18357 was my score, not too shabby for a little 8 core (considering everything else that's running). But yeah, the 9950x is a great CPU and if video editing and workstation tasks are your use case and the price doesn't bother you absolutely get it; but the 7800x3d is equally great for its use case and it DOES NOT SUCK at everything accept gaming. 😒"Terrified and scared shitless of a bios screen." 🤣🤣 "What's the goal?" To not support Intel in its current state, AMD is MORE SO on the right track. Intel will no doubt get there shit together but until then
It is very easy to tune 9950x.1.SMT OFF 2.Curve optimizer -20 all cores.3.Tune memory .READY.OR:1 SMT off,2 OC at max the good ccd-Oc at max the bad CCd.3.Tune memory.READY.The same you can do with 7950x.And the best for everyone is the 12 core.1SMT off,Curve optimizer,2.Memory tuning READY
Do your old tuning guides eventually get posted elsewhere "out in the open" so to say, or do they remain exclusive forever to "god-tier" supporters on the discord? This came to mind when I was tuning an i7-4770K today (which probably predates your channel but you get the idea).
@@fefsefefesfsefsef2733 When I was using that chip it was idling around 20-25 watts. He asked if it's fixed and you act like a child instead of answering the damn question or providing actual idle wattages. No surprise your brain has melted when you put anime footfetish porn in your benchmark thumbnails.
@@fefsefefesfsefsef2733 It does idle at 20+ watts, I can confirm since I own the damn thing and a multitude of posts all over the web confirming 20-30 watt idle draw. In your benchmark afterburner says you run it at 4800(default), most people will run 6000 mt/s or more on a 7800x3d. That will lower your idle draw. It's sad you call misinfo on something that is just a fact. Next time close the animefootfetish porn and enable expo, or maybe copy buildzoids "ez ryzen timings" for some free performance.
About the 2 CCD binning, it is normal and has been in place for at least 2 gens. First CCD always clocks higher, at least since Zen 3, by at least 100MHz. Also, jumping from CCD0 to CCD1 would have even worse perf than just the top frequency difference, due to the extremely high core to core latencies between CCDs. The reason it shouldnt make a difference in multi-core workloads, just like the previous series, is that threads are fully applied to CCD0 first, and due to power constraints, by the time you hit CCD 1, you'll be way below 5.4 anyway for that many active cores.
I finally realized who you remind me of: Bob Odenkirk! Btw, I'm wondering how close is the gap between 7950X and 9950X when both CPUs get the FC special treatment? That's a video I'd like to see.
One chiplet and one shitlet is pretty standard for AMD. I ran into this issues on my 3900x. I would typically tune mine by adjusting a negative curve in PBO (once the curve optimizer was added), figuring out what my max VF Curve point was. Then manually locking in that voltage, slightly lowering until I found a good stable point, and then locking in the multi. Then turn PBO totally off. Looks to me, at least from all these reviewers, that about 1.35 at 5.7 seems to be the high end "cap" of the good chiplet curve and 5.4 at 1.35 for the bad... but the bad one hits that 1.35 VF Curve at like 5.1 ghz.. which is aids. Full discloser... I have not tested this chip yet so this is all speculation based on my previous experience. They say they are going to unlock the new X3D chips and let us change voltages on those now... That might be interesting with direct die cooling.
CCD inequality is the reason that 9950X3D has insignificant chance to be true - expensive to manufacture Both ccd being binned enough to maintain 3d requirements would be rare, they better off making 2 9950x instead The good thing is at the end of the manuf node they could release a sweep the floor low clock 3d still at profit.
They are charging extra for these good CCDs since being used in non-consumer skus. Nothing is left behind in AMD factory, every piece of silicone that doesn't meet the requirements goes to the next lower level product. It's cost effective and great for them
They are making so much profits in data center right now with Q1 to Q2 earnings doubled this year so two high quality ccds on 9900x and up is too expensive when they could sell this one high ccd for epyc instead. we really need intel to be competitive again especially in server as amd is trying to have their cake and eat it too
The other great AMD product was Zen 3 mobile stuff. I have it as 5600H it's just great at everything for it's purpose. Monolithic cores with decent iGPU for 1080p out there and hours of battery efficiency. It's not throttling on battery, compared to my last laptop with best 10gen intel and 2070.
Ryzen 7950x vs. Ryzen 9950x (both delided), using the integrated graphics, CPU cooled with Kraken 360, running Handbrake (2 jobs simultaneously), x265, 1080p, 2pass. The FPS numbers are for 2nd pass and the numbers depends on the source that is being worked with. - 7950x with 142000 PPT, 110000 TDC, 170000 EDC, Curve optimizer -20 => 20-22FPS (No asm=avx512, same FPS with asm=avx512) => Temperature: 65C-75C (fans audible) - 9950x with 142000 PPT, 110000 TDC, 170000 EDC, Curve optimizer -30 => 20-22FPS (No asm=avx512) => 28-31FPS (With asm=avx512) => Temperature: 50C-63C (fans silent) - 9950x with 230000 PPT, 160000 TDC, 225000 EDC, Curve optimizer -30 => 34-38FPS (With asm=avx512) => Temperature: 70C-85C (fans audible) The 9950x with 142000 PPT etc. runs a couple of hundred MHz slower than 7950x with 142000 PPT etc. but manages to get the same work done and is way cooler. I guess the extra 2.1 billion (or how many there are) transistors per CCD come pretty handy. So this is pretty cool, but when asm=avx512 is enabled... sweeet lord. This performance is exactly what I have expected from this CPU. I didn't play much around with BIOS (curve shaper, etc), but I guess I could get another 5-10% out of the CPU since I don't see 100% utilization when running Handbrake. Probably -20 would maintain a more stable Voltage flow to the CPU and it would run more smooth. Noticed, if I do simple tasks, once CCD is parked, don't see any sings of life in it :D. Didn't reinstall Windows, uninstalled the drivers (chipset and graphics), cleared registry, removed all tracks, reinstalled drivers and that's it. No game bar or anything else, Power profile was set to performance since the dawn of time.
guys, should i wait for 9800x3d or just get 9950x ? i want the best cpu for highest fps possible in games (ultra with 4K res) paired with rtx 5090 when it comes out.
So, does this mean that the 9700x will all be good dies, bad dies, or a random mix? EDIT: Just saw on AMD's page that it says "up to 5.5 ghz" compared to the 9950x's "5.7 ghz."
If Intel wasn't giving me a full $600 refund for my Nov 2022 13900k, I would definitely use it over my 7800x3d. 7800x3d is seriously overrated. Thankfully, Intel is paying for my ArrowLake upgrade :)
@@jjlw2378True the 7800x3d is amazing in games that benefit extra cache. In games that care more for IPC, clock speed, more cores, faster memory it’s mediocre. In workstation tasks it’s not a great CPU.
@@ZackSNetwork The Raptor Lake has also nice cache, it was improved significantly over the Alder Lake. IMO, the best mid-range CPU is the 7700 or the 7700X with fast memory. As for the 13600K, the E-cores definitely make rendering and the whole Windows experience faster, especially the boot up time, but I had to disable because the VRM on the motherboard couldn't handle it. Intel went from making 90W i5s to making 180W i5s in just a couple of years. The montherboard manufacturers changed nothing, so the cheaper boards have exactly the same VRM as 5-10 years ago and can't handle 100+W CPUs. The problem is intel doesn't care. And also, HP are scamming their enterprise customers by shipping workstation PCs with fixed 135W/65W PL1/PL2 limits on i7s and i9s, no VRM heatsinks on their pathetic boards and completely inadequate cooling, somehow worse than the intel's stock cooler. Hence, 135W can be held for like 10 seconds under load, then you have a 65W CPU (they're 220W - 250W CPUs btw).
Are both ccds clocked the same on the 7950x or are they at different frequencies like the 9950x? If they’re both at the same frequency on the 7950x then that would make the 7950x a better choice for me.
@@eliteleaf5305yes. He is doing per-CCD overclock. I should try that again also on my 5950x. I briefly tested it ages ago and only saw 100MHz difference then. It would be worth it if I could get at least 200MHz.
A little misleading.. mainstream is serving the everyday PC user who isn't going to spend a ton of time tinkering with the hardware they purchased. Gamers Nexus and you have entirely different consumers or end users in mind. I am in between both types of end user currently which is why I watch both. Also not doubting your skills at fine tuning hardware, Steve and yourself are just creating content for different markets.
We got it, it's great in the only game you care about (COD) with tuned memory I suppose, but what about other games? Also, not knowing how you tested this stuff won't help the average users nor prove the mainstream wrong, lol.
@@FrameChasers Good on you man, I'm still waiting for Arrow Lake and Z890. Pretty much everything you praised about X670/X870 will be better on Z890 as it supports an additional NVMe storage over those AM5 platforms, has mandatory Thunderbolt 4 and the QuickSync stuff will also be faster as ARL iGPU has been updated to Arc Xe-LPG architecture (no need to add an additional Arc card). CAMM2 might also show up first on Z890. With Zen5 being a failure (in my opinion) the AM5 platform has nothing going for it at this point in time. Maybe X3D will save the day, lol.
The server guy is paying the most, but he wants the lower performing core because he doesn't care to go to direct die and do what we are. AMD could redeem themselves by making an XT version...
Maybe the 9800X3D & 9950X3D will be faster then we think. The 9000 series dies are all binned down from server. When desktop users get the crap leftovers this is the outcome.
Wait, AMD isnt clocking the CCD's to the same peak? No reviews mentioned that (that I can recall, but I admit I havent watched all and have been working crazy work hours)... no wonder performance and L3 latency across the IF is so poor.
DerBauer mentioned it, iirc. If anything it was obvious when max frequencies were shown for various workloads in HWiNFO. Those who wanted to see it, saw it (the reason for core parking requirement on multi-ccd Zen5 cpus).
Looking at the launch delay and the bad reviews, my only theory I could come up with had to do with silicon allocation. Once the product was engineered, there is really no other reason why it would underperform besides binning and segmentation out of the hands of the engineers. The fact that they just put out non-X parts as X parts for the R5 and R7 launch as well as the inventory clawback and launch delay all had to do with either terrible yields that couldn't meet the demand of a July launch, or hoarding good silicon for other segments such as server. Then they spring the core parking on the reviewers right before re-launch (AFTER initial launch delay and clawback) looked like a last minute pivot to lesser quality silicon in all of the consumer SKUs. Let's say theoretically their silicon meets 4 grades: S-Tier (Golden sample), A, B, and C Previous generations like the 5950X would give you 2 A-tier CCDs 5950X: A + A Perhaps the 7950X regressed to either one A-tier and one B-tier, or one A-tier and one silicon lottery 7950X: A + B The 7000 series at least had a spectrum of all silicon where it seems like they did the following 7950X: A + B 7700X: B 7700: C And the 9000 series, once it neared launch, got a massive downgrade to 9950X: A + C 9700X: C During the 5000 and 7000 series they are probably using C grade to stack X3D cache on. By using C-grade in all of their consumer products for 9000, now they can release X3D and claim they did it without having to compromise performance and allowing the same overclocking. Just my theory.
The reason AMD does this is simple: Datacenter quarter to quarter profits doubled this year so milk the datacenter as much as possible even it means screwing the customer
And shouldve launched the 9950x at 459 9950x3d at 559... This would've been a better way to price the top end competitively... Not 659 for a non x3d CPU that has half the CPU turning off.
Just like amd Graphics cards remember the drivers need to be tweaked.. I'm sure that if you give it a few months the performance of the new 9000 series cpus will blow your mind... History is repeating itself as usual for amd launch a graohics card and wait a few months then the drivers have been tweaked and voila you have a respectable gpu.. Intel arc cards remember they werwe dog dirt slow but the drivers were wqorked on and then those intel gpu's were worth considering
AMD trick you, they say UP TO 5.7ghz but then you click on the small print and it says single core burst load... They add further caveats and say it depends on motherboard, cooling, thermal paste etc..
0:58 - And what's with HUB's Windows benchmarks, admin user, 24H2 etc. Nobody has been able to replicate it but AMD users taking this as gospel.. Something isn't right, but then again why are you listening to 'techtubers'.
Poor yields I'd say, they're already putting the weaker ccds in all of the lower tier cpus anyways but I guess they just have a lot of them in their hands.
Holy F___!!! Finally!! Someone who has the same perspective as me. I, too, hate V cache and big.LITTLE BS. I like having just P cores. The E core hybrid technology f___ed up the evolution of processors. Too many memory controllers incorporated in the processor. (DDR4 & 5). Too much compaction in the dies. Too many features to offer, too much going on in one processor. So, you end up with bugs and big issues. It's like an OBD1 car vs an OBD2 car. The OBD2 car has too many f___ing sensors and control modules. Misfires, predetination, pinging, knocking, stalling, etc issues occur easily because there is too much to depend on.
I love how you mention this CPU is good at decompressing fitgirl repacks. 😂😂
Fit girl is a blessing to gamers😁 but not developers...🤣.
Make that a standard benchmark with graphs, like HUB. 😂
When i turned off the e-cores on my 12700K for a test i tought myself: "wow, 8 core Ryzen 7s and older i7s are sluggish AF" .
Fitgirl benchies ftw
AMD users watching this are having their minds blown right now
Yep 😮
I’m an AMD user 🤷♂️
AMD marketing should be fired en masse
@@Natedaskate Do you know what anecdotal means?
🤷
Yes. Anecdotal means not necessarily true or reliable and completely subjective and not based in fact. Thanks.
I’m an AMD user and I love this channel.
The problem is that 95% of gamers are not interested in spending 3 days tuning a cpu and possibly finding out its still unstable after. Reviewers are not wrong testing the out of the box experience as thats 95% of all users. Tuning and channels that revolve around tuning are an extremely niche population. Reviews need to be geared to the setup most people are going to be using. In that case the 7800x3d still offers the best out of the box experience for 95%+ of gamers.
yes, out of the box is how it should be for everyone but xoc dorks
Chaser does really underestimate how much a average Joe wants to configure / optimise their system. Mainstream could do a better job of a dirty overclock section in their reviews .
@@mushbeck Well, yeah, but when I was first trying to OC I found a program called !xSpeed. This OCs Windows and Windows software, but NOT hardware. Don't even ask me how it does; all I know is that it worked on most software I tried. Anyway, that whole saga didn't end well. I eventually got convinced to remove dibeng.dll, lol.
I think most people learn how to OC when they learn the other stuff first, and tbh this not-aggressive gatekeeping is probably for the best. Anyone who wants to find it will.
yeah its not the customer's job to fix your CPU
@@Paradox-om6lm AMD really struggles with their software unfortunately, it's why Nvidia is crushing them
I immediately had to look out my window to see if there were any pigs flying by. Much respect for just calling it how it is, Literally the ONLY channel who does this and it's the main reason why I've been subscribed for a cpl years now. I always know that I can count on Frame Chasers for raw facts W/O any BS
Holy shit he's actually happy...
🤣
I went from excited for Zen 5 to disappointed, to excited again with this video to immensely disappointed.
In my region, 9950X 975$, 13900K 415$, 14900K 430$, 7950X 450$. All Box versions from internet shop with 3Year warranty 🙄
who will buy 9950X at this price.
that was also my criticism of all the reviewers that they never showed whether the AVX-512 values have improved, because one of the main advertisements for the 9000 series from AMD is exactly that point. never saw a PS3 Emulator (RPCS3) speed test, like u always said, most of them not are gamers
Zen 5 improvements compare to Zen 4
The L1 cache per core is increased from 64 KB to 80 KB per core.
A ~16% IPC uplift on average, claimed by AMD.
Memory speeds up to DDR5-5600 and LPDDR5X-7500 are officially supported.
Infinity Fabric clock (FCLK) has been increased to 2400 MHz.
New 2-ahead branch predictor
AVX-512 capabilities have been expanded with Zen 5 with a doubling of the floating point pipe width to 512-bit. Additionally, there is greater bfloat16 throughput which is beneficial for AI workloads.
RPCS3 only 10% improvement. It uses only 128-bit functions of avx512 which were already available in Zen 4.
Listening to Jufes talk about how much he likes an AMD CPU makes me feel like I'm in Bizarro World. He's always said he tries to be an unbiased channel that offers real info, but he ripped AMD so much people just figured he's an Intel shill no matter what he says...this video seems to prove otherwise.
He likes what works and the moments when he feels he is being fairly treated by the company. But yeah, many AMD fans are so religious about it, that I think Jufes just loves to rub in their cognitive dissonances a bit.
The x3d chips were weird topic here. Mostly because most of techtube tests things at stock settings and x3d shines with those. Then most tech beginners get the idea that x3d is the only way to go. Their OC and tuning potential is very low though (for mostly the same reason why they are good out of the box) and that is why they do not compare as favorably to tuned systems. That difference has been very hard for the AMD community to accept and a lot of semireligious discussion has risen from it. Prime grounds for a somewhat heated discussions.
So this is why 9950X requires a scheduler, compared to 7950X that does not.
AMD is covering up the fact that they are putting one bad CCD in that thing. So lame...
AMD has been using one good CCD and one bad CCD since zen 2
Yup.. This was the same for 5900x and 5950x... One good die and one Med or shit CCD.
Some people also get really screwed over when there was the covid chip shortage and both CCD's on a 5950x was only slightly faster then 5900x.. :(
The big reason for that is that the cross-CCD latency is 2.5x as much as previous gens. If games use cores on both chips, it definitely hurts performance.
no really amd has had issues windows scheduler for a while now. but all the cries that windows users are having are non-existent on linux. but I am not saying peps need to switch to linux but windows has always had issues with cpu scheduling for a long time now. but then again intels mixed cores were a tire fire when they released it and they worked w/ MS to make the Ecore&Pcore thing work poorly at minimum.
bro most people are not going to adjust 50 settings if its 3% faster in default settings vs old gen thats what reviewers will say and their not wrong its not on the user
if you're leaving that much to be acquired over 23% in tuning comparing to tech jesus himself, is insane your opinion is wrong but let me tell you WHY it's wrong. You get what you pay for and if you paid for the best you should get the best and if you buy the best tool for the job it better do its fucking job.
@@NeverLucky-RubberDucky yeah im clearly the insane one here.
Well, the idea is that configured right, this thing is genuinely great for workflow and gaming, without X3D dual CCD problems from a 7950X3D. That more performance is genuinely achievable for those who seek it. Again, this is a channel directed at people who are going to tune when they get their mits on something.
Why are you watching my channel if you’re not interested in tuning? Lolwut
I think the buyers of this CPU are people who are akin to tuning and getting max performance. If you are buying products that cost this much, you are the kind of customer who love max performance irregardless of parts cost in time and money. These are for enthusiasts and PC builders. I love watching this channel because FC only caters to the top 0.00001% of pc builders. Almost all of my friends who subscribe to FC and his Patreon are buying all the products he tests. I have bought every version of 5x00x3d 7800x3d 13900k/kf/ks and 14900k. Dude everything that Dufus has reviewed. If you listen to his content, you would realize that tuning is the real deal when working with these high end products. It’s so extreme that even tuning ram is important. buy the FC ram and you’ll see how tuning really changes your system and 1%lows. I hope this helps and also makes you excited for tuning as well.
I'm going to guess the fafo is disabling smt and the second ccd, then overclocking the first ccd.
More interested in the Core Ultra 285K.
Actually the 2 CCDs are not an issue for me cuz I like to VM it on Proxmox.
If you have a partner, you can cut the 9950X in HALF making it a 9700X for both of you. Now, all you need is 2 monitors, 2 GPUs and have fun playing on a single machine. And another way is, you can still run the 1CCD for gaming, and the other slower CCD, just run a local LLM like Llama3 or idk... maybe a video encoder or something... You got 16 cores with fully enabled AVX512, so push the multitasking to the limit! That CPU should be excellent for RPCS3...
RPCS3 only primarily uses 128bit AVX512 instructions so, as demonstrated by TechPowerUp which benchmarked RPCS3, it's within margin of error relative to the 7950X (though interestingly the 9700X was a bit faster than the 7700X to the point of not being within margin of error?)
For reference you can read more info on this from the recent blog post from the developer of y-cruncher about Zen5's AVX512 (a simple ctrl+f for "rpcs3" on the post will take you to the corresponding text).
it's 10% faster than Zen 4. Ask the devs(RPCS3).
Pbo +150. -33 Curve x5 very cool, low watts and great performance on my 9900x I also have approached the 6ghz for fun although dialled it back for daily use and longevity.
I notice the "mainstream" often tell people to avoid the CPUs I buy....
In-fact you can achieve benchmark scores (Cinebench etc) only attainable using liquid nitrogen or water chillers with 7950/7900x with just a well put together custom WC loop with this 9000 series.
A very fun CPU for those that like to tweak :)
Always the first manufactured Dies with different process like zen 5 its on 4nm with dual ccds have much bigger gap between 2 ccds as nod improve with the time this gap in Mhz will be closed a lot, that's happen in all AMD cpus with 2 ccds as release it, also the first ccd will improve with all core boost close to max speed in a future, but is much better than intel as 14900K have only 3 cores of 24c to boost at max speed (6000 mhz)
15:30, even as an AMD fan I have to agree, they should use the lower binned CCDs in the 8 core CPUs.
They're already doing it. That's why the 9700X is a 65W chip instead of 105W like the previous gen 7700X.
@@Patrick73787 That explains why my i7-12700K is still so competitive vs both?
They are doing it 100%. bought 3 9600x and all max out between 5450 and 5550 all core. Then again, zen 5 at 4.9ghz matches zen 4 at 5.625ghz so there's more to the story than "AMD using crappy ccds"
The very principal behind the CCDs works out in how workloads end up being handled. Your light loads will shuffle around your highest clockimg cores. As the load increases to more threads those clocks collapse anyway. There have always been inter-CCD issues in games, because the other is basically empty in those moments when the thread jumps.
I would love to see you review the 9800X without the CCD Latency issues.
The only thing consistent with Zen5's launch, is how inconsistent it is.
Remember in the late 90s hitting 300 MHz was a significant technological achievement for cpus which costs only a few hundred bucks. So just think about it that way
Why not buy another 9950x and see if it is better binned?
Doesn't AMD have an iGPU! how come these programs don't utilize AMD's iGPU like they Utilize INTEL's quick sync?
I'm certain all the best chips go to the server market. Reminds me when the consumer market had a larger slice of the pie and Xeon and Opteron chips fit on consumer boards which also had the best OC.
Indubitably
Jufus visibly gloating without the need for a humble brag because he put in the time and has earned it.
Zen 5 is only legit if up coming arrow lake is crap. I just don’t want to regret after 2 months when arrow lake releases. At this moment, I would say just wait. BTW, I can live without 6GHz @1.55v.
You are not the real cod bro
14900k 5.8 at 1.34 lol. 6.0 won't change anything for gaming lol.
Test it.
Arrow Lake Leaks are already out.
You only need a 10900k.
My 4090 is my bottleneck on my 7800x3d and 14900k maxed systems.
I'm nowhere near cpu bottlenecked, and would be fine on a 10900 lol....
@@Hardware_Neutral That's irrelevant. CPUs not only need to not be a bottleneck, they need a lot of compute in reserve to take care of potential stutter in intense scenarios that are barely tested by benchmarks.
@JAnx01 have you tested it lol?
10900K is the only last legit CPU with 10 P-Cores. If you don't care about power, it's the golden CPU
Power Efficiency is real nice in the summer. My 13900k required a fucking window-unit AC to keep my room bearably cool
I get it that you don't want to share the secrets, but you should've mentioned more of the reasons why the mainstream techtubers are wrong.
Basically everything you said boils down to scheduling = bad. Which is a total non-issue for 9700X and 9600X, which also had subpar results in their reviews. So that can't be the only reason why Zen 5 is .... weird.
9950x also costs ~ $200 dollars more than the 7800x3d and MOST people can't be bothered to spend hours upon hours of fine tuning a CPU that SHOULD give the performance experience you were able to message out of it, out of the box (in a perfect world right). Anyways the 9950x is a really interesting CPU but after a 6-year hiatus from pc building I just built a new pc and went with the 7800x3d for all of the obvious reasons (wanted easy ;) and couldn't be happier with it. I do like to tinker though and with minimal effort in the oh so scary bios 😨 it's incredible how efficient even the 7800x3d is/can be. Max load, yes water cooled (not direct die f that), I NEVER exceed 83C. Just ran a single clip CR23 while this vid was playing with 10 tabs open in Edge and COD minimized, 5050 all core, 1.07v average for cores, 83c Max TDie, 78w package power max; 18357 was my score, not too shabby for a little 8 core (considering everything else that's running). But yeah, the 9950x is a great CPU and if video editing and workstation tasks are your use case and the price doesn't bother you absolutely get it; but the 7800x3d is equally great for its use case and it DOES NOT SUCK at everything accept gaming. 😒"Terrified and scared shitless of a bios screen." 🤣🤣 "What's the goal?" To not support Intel in its current state, AMD is MORE SO on the right track. Intel will no doubt get there shit together but until then
It is very easy to tune 9950x.1.SMT OFF 2.Curve optimizer -20 all cores.3.Tune memory .READY.OR:1 SMT off,2 OC at max the good ccd-Oc at max the bad CCd.3.Tune memory.READY.The same you can do with 7950x.And the best for everyone is the 12 core.1SMT off,Curve optimizer,2.Memory tuning READY
25:30
~"core to core and CCD to CCD latency is not an issue"
Let me guess…AMD made it so the infinity fabric runs at half speed by default or something ?
News flash, for me!
I own the R9 9950X, but I had no idea that there are speed differences between the 2 CCDs!
Do your old tuning guides eventually get posted elsewhere "out in the open" so to say, or do they remain exclusive forever to "god-tier" supporters on the discord? This came to mind when I was tuning an i7-4770K today (which probably predates your channel but you get the idea).
Did they fix the idle consumption on the 9000 series? The 7800x3d idles at like 25w.
Boo hoo
I love how you don't have the CPU and spread misinformation about it
@@fefsefefesfsefsef2733 When I was using that chip it was idling around 20-25 watts.
He asked if it's fixed and you act like a child instead of answering the damn question or providing actual idle wattages.
No surprise your brain has melted when you put anime footfetish porn in your benchmark thumbnails.
@@fefsefefesfsefsef2733 It does idle at 20+ watts, I can confirm since I own the damn thing and a multitude of posts all over the web confirming 20-30 watt idle draw.
In your benchmark afterburner says you run it at 4800(default), most people will run 6000 mt/s or more on a 7800x3d.
That will lower your idle draw.
It's sad you call misinfo on something that is just a fact.
Next time close the animefootfetish porn and enable expo, or maybe copy buildzoids "ez ryzen timings" for some free performance.
I don't know about desktop as of now but in laptops Meteor Lake beats Zen5 Strix Point in idle power draw and battery life.
About the 2 CCD binning, it is normal and has been in place for at least 2 gens. First CCD always clocks higher, at least since Zen 3, by at least 100MHz. Also, jumping from CCD0 to CCD1 would have even worse perf than just the top frequency difference, due to the extremely high core to core latencies between CCDs. The reason it shouldnt make a difference in multi-core workloads, just like the previous series, is that threads are fully applied to CCD0 first, and due to power constraints, by the time you hit CCD 1, you'll be way below 5.4 anyway for that many active cores.
Makes me feel like yield is not good, if they had enough 5.7 bins they could have made both chips on the top chip 5.7 bins without a crazy price tag.
I finally realized who you remind me of: Bob Odenkirk!
Btw, I'm wondering how close is the gap between 7950X and 9950X when both CPUs get the FC special treatment? That's a video I'd like to see.
I agree the ccds should have the same peak frequency.
I just got the 7950X3D - would you bother to upgrade to the 9950X?
no! just wait for 2 to 3 more gens
So when intel launches their new SKU's with 10-12 P-Cores thats basically the only thing any of us will want
Is that even in the horizon? Seems like Intel is stuck on 8 P-cores for now.
One chiplet and one shitlet is pretty standard for AMD. I ran into this issues on my 3900x. I would typically tune mine by adjusting a negative curve in PBO (once the curve optimizer was added), figuring out what my max VF Curve point was. Then manually locking in that voltage, slightly lowering until I found a good stable point, and then locking in the multi. Then turn PBO totally off. Looks to me, at least from all these reviewers, that about 1.35 at 5.7 seems to be the high end "cap" of the good chiplet curve and 5.4 at 1.35 for the bad... but the bad one hits that 1.35 VF Curve at like 5.1 ghz.. which is aids. Full discloser... I have not tested this chip yet so this is all speculation based on my previous experience. They say they are going to unlock the new X3D chips and let us change voltages on those now... That might be interesting with direct die cooling.
CCD inequality is the reason that 9950X3D has insignificant chance to be true
- expensive to manufacture
Both ccd being binned enough to maintain 3d requirements would be rare, they better off making 2 9950x instead
The good thing is at the end of the manuf node they could release a sweep the floor low clock 3d still at profit.
They are charging extra for these good CCDs since being used in non-consumer skus. Nothing is left behind in AMD factory, every piece of silicone that doesn't meet the requirements goes to the next lower level product. It's cost effective and great for them
They are making so much profits in data center right now with Q1 to Q2 earnings doubled this year so two high quality ccds on 9900x and up is too expensive when they could sell this one high ccd for epyc instead. we really need intel to be competitive again especially in server as amd is trying to have their cake and eat it too
The other great AMD product was Zen 3 mobile stuff. I have it as 5600H it's just great at everything for it's purpose. Monolithic cores with decent iGPU for 1080p out there and hours of battery efficiency. It's not throttling on battery, compared to my last laptop with best 10gen intel and 2070.
In your testing, did you disable the good CCD in bios and test the clocking ability of the bad CCD?
Ryzen 7950x vs. Ryzen 9950x (both delided), using the integrated graphics, CPU cooled with Kraken 360, running Handbrake (2 jobs simultaneously), x265, 1080p, 2pass. The FPS numbers are for 2nd pass and the numbers depends on the source that is being worked with.
- 7950x with 142000 PPT, 110000 TDC, 170000 EDC, Curve optimizer -20 => 20-22FPS (No asm=avx512, same FPS with asm=avx512) => Temperature: 65C-75C (fans audible)
- 9950x with 142000 PPT, 110000 TDC, 170000 EDC, Curve optimizer -30 => 20-22FPS (No asm=avx512) => 28-31FPS (With asm=avx512) => Temperature: 50C-63C (fans silent)
- 9950x with 230000 PPT, 160000 TDC, 225000 EDC, Curve optimizer -30 => 34-38FPS (With asm=avx512) => Temperature: 70C-85C (fans audible)
The 9950x with 142000 PPT etc. runs a couple of hundred MHz slower than 7950x with 142000 PPT etc. but manages to get the same work done and is way cooler. I guess the extra 2.1 billion (or how many there are) transistors per CCD come pretty handy. So this is pretty cool, but when asm=avx512 is enabled... sweeet lord. This performance is exactly what I have expected from this CPU.
I didn't play much around with BIOS (curve shaper, etc), but I guess I could get another 5-10% out of the CPU since I don't see 100% utilization when running Handbrake. Probably -20 would maintain a more stable Voltage flow to the CPU and it would run more smooth. Noticed, if I do simple tasks, once CCD is parked, don't see any sings of life in it :D.
Didn't reinstall Windows, uninstalled the drivers (chipset and graphics), cleared registry, removed all tracks, reinstalled drivers and that's it. No game bar or anything else, Power profile was set to performance since the dawn of time.
guys, should i wait for 9800x3d or just get 9950x ?
i want the best cpu for highest fps possible in games (ultra with 4K res) paired with rtx 5090 when it comes out.
Your "4k ultra" fps will not be cpu limited 😉
I bought AMD and found out im gonna cry myself to death now
So, does this mean that the 9700x will all be good dies, bad dies, or a random mix? EDIT: Just saw on AMD's page that it says "up to 5.5 ghz" compared to the 9950x's "5.7 ghz."
This is where the rubber meets the road boys. There’s no one better
You would still need FAFO to keep games on 1 CCD to avoid penalty of crossing the fabric from 1 ccd to another.
That's something I also noticed going from my 13600k to the 7800x3d, when not in game the i5 is better for nearly everything else.
should have kept the 13600k, if u tuned it, it would have been within 5-10% in gaming and like 40% faster in everything else.
If Intel wasn't giving me a full $600 refund for my Nov 2022 13900k, I would definitely use it over my 7800x3d. 7800x3d is seriously overrated. Thankfully, Intel is paying for my ArrowLake upgrade :)
@@jjlw2378True the 7800x3d is amazing in games that benefit extra cache. In games that care more for IPC, clock speed, more cores, faster memory it’s mediocre. In workstation tasks it’s not a great CPU.
@@ZackSNetwork The Raptor Lake has also nice cache, it was improved significantly over the Alder Lake. IMO, the best mid-range CPU is the 7700 or the 7700X with fast memory.
As for the 13600K, the E-cores definitely make rendering and the whole Windows experience faster, especially the boot up time, but I had to disable because the VRM on the motherboard couldn't handle it. Intel went from making 90W i5s to making 180W i5s in just a couple of years. The montherboard manufacturers changed nothing, so the cheaper boards have exactly the same VRM as 5-10 years ago and can't handle 100+W CPUs. The problem is intel doesn't care.
And also, HP are scamming their enterprise customers by shipping workstation PCs with fixed 135W/65W PL1/PL2 limits on i7s and i9s, no VRM heatsinks on their pathetic boards and completely inadequate cooling, somehow worse than the intel's stock cooler. Hence, 135W can be held for like 10 seconds under load, then you have a 65W CPU (they're 220W - 250W CPUs btw).
I cannot think of a scenario where the 7800X3D is good enough to justify itself other than for max fps in esports games at 1080p.
Are both ccds clocked the same on the 7950x or are they at different frequencies like the 9950x? If they’re both at the same frequency on the 7950x then that would make the 7950x a better choice for me.
Shouldn't be. Amd are doing that since zen2
@@valentin3186 is there a way to go into the bios and get the cores on both ccds locked to a specific frequency?
@@eliteleaf5305yes. He is doing per-CCD overclock. I should try that again also on my 5950x. I briefly tested it ages ago and only saw 100MHz difference then. It would be worth it if I could get at least 200MHz.
A little misleading.. mainstream is serving the everyday PC user who isn't going to spend a ton of time tinkering with the hardware they purchased. Gamers Nexus and you have entirely different consumers or end users in mind. I am in between both types of end user currently which is why I watch both. Also not doubting your skills at fine tuning hardware, Steve and yourself are just creating content for different markets.
what if you just disable the second CCD. it would still suck but interesting how it would perform then.
We got it, it's great in the only game you care about (COD) with tuned memory I suppose, but what about other games? Also, not knowing how you tested this stuff won't help the average users nor prove the mainstream wrong, lol.
All games, and don’t care about anyone outside of my discord
@@FrameChasers Good on you man, I'm still waiting for Arrow Lake and Z890. Pretty much everything you praised about X670/X870 will be better on Z890 as it supports an additional NVMe storage over those AM5 platforms, has mandatory Thunderbolt 4 and the QuickSync stuff will also be faster as ARL iGPU has been updated to Arc Xe-LPG architecture (no need to add an additional Arc card). CAMM2 might also show up first on Z890. With Zen5 being a failure (in my opinion) the AM5 platform has nothing going for it at this point in time. Maybe X3D will save the day, lol.
The server guy is paying the most, but he wants the lower performing core because he doesn't care to go to direct die and do what we are. AMD could redeem themselves by making an XT version...
Does the 9950x suffer from the AMD dip?
So 14900k is still ur preferred CPU currently?
Amd silicon variation on launch is always big there can be even 10% diffrence on stock settings.
Maybe the 9800X3D & 9950X3D will be faster then we think. The 9000 series dies are all binned down from server. When desktop users get the crap leftovers this is the outcome.
in my country the 7800x3d its 400E and the 9950x its 660E so i'll save that 260E and wait in windows to decompress fitgirl repack 5-10 min slower :D
One question Jufus Win10 or 11 for the 9950x?
Trust me bro.
please use black background on notepad
7950X3D got a better X3D CCD than 7800X3D and same CCD as 7950X that hits 5.7GHz. Maybe we wait for 9950X3D?
Wait, AMD isnt clocking the CCD's to the same peak? No reviews mentioned that (that I can recall, but I admit I havent watched all and have been working crazy work hours)... no wonder performance and L3 latency across the IF is so poor.
DerBauer mentioned it, iirc. If anything it was obvious when max frequencies were shown for various workloads in HWiNFO. Those who wanted to see it, saw it (the reason for core parking requirement on multi-ccd Zen5 cpus).
Looking at the launch delay and the bad reviews, my only theory I could come up with had to do with silicon allocation. Once the product was engineered, there is really no other reason why it would underperform besides binning and segmentation out of the hands of the engineers.
The fact that they just put out non-X parts as X parts for the R5 and R7 launch as well as the inventory clawback and launch delay all had to do with either terrible yields that couldn't meet the demand of a July launch, or hoarding good silicon for other segments such as server. Then they spring the core parking on the reviewers right before re-launch (AFTER initial launch delay and clawback) looked like a last minute pivot to lesser quality silicon in all of the consumer SKUs.
Let's say theoretically their silicon meets 4 grades: S-Tier (Golden sample), A, B, and C
Previous generations like the 5950X would give you 2 A-tier CCDs
5950X: A + A
Perhaps the 7950X regressed to either one A-tier and one B-tier, or one A-tier and one silicon lottery
7950X: A + B
The 7000 series at least had a spectrum of all silicon where it seems like they did the following
7950X: A + B
7700X: B
7700: C
And the 9000 series, once it neared launch, got a massive downgrade to
9950X: A + C
9700X: C
During the 5000 and 7000 series they are probably using C grade to stack X3D cache on. By using C-grade in all of their consumer products for 9000, now they can release X3D and claim they did it without having to compromise performance and allowing the same overclocking.
Just my theory.
Sounds plausible.
The reason AMD does this is simple: Datacenter quarter to quarter profits doubled this year so milk the datacenter as much as possible even it means screwing the customer
All media: AMD processors are great!
FCh: AMD is crap!!
All media: AMD processors are bad!
FCh: AMD is great!
Let's get deeper! farm amd fanboys turn the hate into love, you will be their hero. "See i knew Ryzen5 wasn't that bad, thanks for proving it Jufes"
For kick is there a subscription fee?
To watch nah.
you expect to get the best bin on the flagship cpu but the reality is Ryzen cpus bins are second behind Epyc cpus
Ofc, since first Zen because its where the money is
Have you thought about trying the AM5 Epyc yet?
Second ccd is slower for power efficiency and temps. It's best that way.
alright you have convinced me to join your discord :) supporting your work make sense
AMD Sales: Sorry customers, don’t worry we will release the 9950xtx edition where both CCD will hit 6.0Ghz! 😅
9950X Core-to-Core latency is garbage, measured by anadtech review, 7950X has way lower latency AMD doing something silly again...
It's not core to core it's ccd to ccd, and it dost matter for gaming or productivity you don't know what you are talking about
@@fefsefefesfsefsef2733 mistyped, thnx for the correction. btw core to core latency also got worse.
@@davidszep3488 Who cares when games run only on 1 ccd and for rest it does not matter
@@fefsefefesfsefsef2733 Sure if you process lasso everything.
The 9950X has worse core-to-core latency than the 7950X even in a single CCD.
And shouldve launched the 9950x at 459
9950x3d at 559... This would've been a better way to price the top end competitively... Not 659 for a non x3d CPU that has half the CPU turning off.
Just like amd Graphics cards remember the drivers need to be tweaked.. I'm sure that if you give it a few months the performance of the new 9000 series cpus will blow your mind... History is repeating itself as usual for amd launch a graohics card and wait a few months then the drivers have been tweaked and voila you have a respectable gpu.. Intel arc cards remember they werwe dog dirt slow but the drivers were wqorked on and then those intel gpu's were worth considering
Should have matching bin on both CCD's on top end chip
what about windows 24h2 update boost?
Should have had a 9940X (5.4ghz) and 9970x (5.7ghz) or something
They didn’t pull this crap with the 5950x. Both CCD’s would hit full boost. If they gave 2 good CCD’s this would be great!
They did if you have b2 stepping or later CPU, the silicon quality was so much improved that they all hit 5.1ghz single core
AMD trick you, they say UP TO 5.7ghz but then you click on the small print and it says single core burst load... They add further caveats and say it depends on motherboard, cooling, thermal paste etc..
the same is true for intel though, it's just how it is nowadays I guess
@@jabronilifestyle Thats not true, the 14900k is at 5.7 all core, 6ghz 2 core boost.
5.5/5.6 ghz under heavy load
7800x3D is for people who want a console like experience
nice to see jufes this giddy again, can see that new tech energy is back 😅
0:58 - And what's with HUB's Windows benchmarks, admin user, 24H2 etc. Nobody has been able to replicate it but AMD users taking this as gospel..
Something isn't right, but then again why are you listening to 'techtubers'.
Zen 5 is going to be great for gaming laptops
A little of topic, HU claims fsr 3.1 is blurry mess, rdn3 is a disaster, and zen 5 is a fllop, the pattern shows whats going on there....
14900k for the win
The thing is Zen 5 despite being a meme, is still actually... good
Meanwhile Intel's 13th and 14th gen is just a meme
The x670e creator have no second clock generator....so i think you cant reach your goals with the ne creator board
You've paid for the thing, why wouldn't you want to see how much performance can be had and what it is capable of?
why would amd put a ccd meant for the 9900x on the 9950x??????
Poor yields I'd say, they're already putting the weaker ccds in all of the lower tier cpus anyways but I guess they just have a lot of them in their hands.
Was not expecting that at all
Does this mean that the 9700X can be as strong as the 7800X3D?
With unlocked power limits it's basically my 12700K with higher single core speed.
Doesn’t GN test at stock settings? How are you going to hate on that? I could understand if they claimed they OC’d it and you compare.
Is it just me or did anyone to their bone felt like this is exactly what's gonna happen when Jufus benchmark the 9950x???
💌💌💌
They should make a 9950xt3d with 2 good x3d ccds
Dude I can't wait for your next video!
Now this is a shocker, but not really.
Right
Holy F___!!! Finally!! Someone who has the same perspective as me. I, too, hate V cache and big.LITTLE BS. I like having just P cores. The E core hybrid technology f___ed up the evolution of processors.
Too many memory controllers incorporated in the processor. (DDR4 & 5). Too much compaction in the dies. Too many features to offer, too much going on in one processor. So, you end up with bugs and big issues.
It's like an OBD1 car vs an OBD2 car. The OBD2 car has too many f___ing sensors and control modules. Misfires, predetination, pinging, knocking, stalling, etc issues occur easily because there is too much to depend on.