You left out a huge equation from this video and that is that 99% of all 4 dim motherboards can not run high-speed ram sticks like 8000 for those type of speeds you will need 2 dim motherboards like the Asus Apex and Gigbyte Tachyon. Also pretty sure all of your results on the higher speeds were incorrect and mostly likely negative due to instability of the pro art board being a 4 Dimm motherboard.
he mentioned it near the end, he uses 4 sticks in his build, and had to lower the speed. i have 192 gigs in one of my z790 stable at 5200. they dont mention that you can use that much ram. except that my z790 maximus extreme does.
For videos like this, at least to get the system setups right, you should really collaborate with guys like BuildZoid from @ActuallyHardcoreOverclocking or others who are related to the XOC scenario. The moment you went with a 4 dimmer for validating 8000. You had me off. Some of the best 4 Dimmer (Ex. 790 Aorus Master X, 790 Aorus Elit X AX, 790 PG Nova, etc.) can only reach 7600-7800 with a really good IMC. Also, some of the major performance benefits come from the tighter secondary and tertiary timings of DDR5 and don't solely rely on the speed and primary timings alone. 7600 with tighter timing can outperform 8000 with untuned or loose timing, and XMP is kind of great at that just to make it work. So this video won't represent a super-accurate apples-to-apples comparison in terms of the scalability of true RAM performance. FYI, for similar reasons, Jay's video is also invalid.
Jay is not as much of an expert as he pretends to be. There are many inaccuracies, dubious approximations and hazardous conclusions in his videos. I'd have more trust in Linus of LTT, he has better knowledge, intuition and rigour in his logic, and if he's presenting, he probably vetted the script. But the one who usually pays the most attention to accuracy is Steve of Gamers Nexus, although I don't think that he covers RAM unfortunately (haven't checked).
@@dismuter_yt check out actually hardcore overlocking. he covers both intel and amd ram overclocking and also cpu and gpu overclocking. hes probably the best source of information in regards to ram. also unlike many overclockers, he actually does tell you the reasons as to why he overclocks something, like lets say cpu/gpu frequency, ram transfer speeds, ram timings, voltages, what a certain timing does, what a certain voltage is for, how certain ram timings affect things, why some timings dont matter, etc. etc. hardware unboxed had him do some ram overclocking once and they did some benchmarks. hes a great source of information. i learned to overclock after watching his videos
It does seem like though, that it's generally been accepted that 6000-6400 is a stable setup with 2 DIMMs installed, so there is that. Anyways, I'm glad this channel exists to benchmark and review PC hardware outside of gaming contexts, this was sorely needed in the PC review space.
Yesterday I mentioned to a creator friend of mine that populating all RAM slots with very fast DDR5 sticks could actually result in his motherboard reducing the speed he will actually get and he said that makes no sense as a higher total RAM memory quantity will always benefit a creator. Thank you Lauri.
You are correct, under a certain situation - overclocking. Outside of that, assuming the ram is in his QVL list or has a high compatibility, he will get the rated speed and timings. Quantity, as your friend suggests, is correct assuming 'all' the ram is used, and the system is now going into swopfile. He would likely benefit from more cache and cores on his CPU. Also, timings may well show a good improvement in reducing the time taken to render or produce his work, obv each app he uses will differ in how it uses and allocates resources from the hardware.
In the old days... Populating them all with smaller sticks to total a larger ammount 'was' the best thing to do... newer motherboards changed the way the Memory lanes are done apparently, so your better off with 'pairs' unless your using a really really expensive MB that can run 4. *quick edit* Found this out myself recently while getting ready for a new PC rebuild/upgrade... had ordered 4 sticks of 16 to get the 64 i wanted while filling up all the slots... found out the next day that would actually overload the Memory control module and cause slowdowns.... soooo gotta return those 4 when they get here tomarrow, and exchange them for a pair of 32gb sticks.
Ended up opting for a 128GB (32GB x 4) kit of Kingston Fury Beast DDR5 5600MT/sec. I had considered 96GB (2 x 48GB) but as there is so little choice available, the pricing seems unreasonably high imo. Hoping that even if I take a small hit on the speed that my RAM runs at because I am populating all four DIMM slots, that the extra total RAM capacity balances that out somewhat when working with big projects.
Were any actual stability test performed with 7800 Mhz? Your board is not capable of such speeds, specs on asus website are BS. You need APEX for sure stable 8000. Please run y-cryncher VST for 4+ hours. Active ram cooling is a must. If ram is unstable then your tests are bad.
Good, helpful video. A kindly gentle correction starting at 3:36 - Asus has a dedicated IMC SP score called MC SP, which is available on their higher-end mobos (I don't know where the cutoff is, but I know a board like the z790 Tuf Gaming doesn't have it available in the bios/uefi). From within the bios, you have to actually run the Get MC SP the first time to get a score (unlike the SP score). If it is available on your mobo, It is found under Extreme Tweaker/AI Features/Get MC SP. So, the SP score does not relate directly to the IMC. I hope this is helpful :)
God, I love you for creating PC content for creative work. Building a PC for gaming was easy because information is everywhere. However, building a PC for my girlfriend, who is a graphic designer, is much harder. There is very little information about what components are better to choose and why, etc.
Not sure what programs your girlfriend uses but if she uses adobe programs like Photoshop they generally benefit from faster single core CPUs and lots of RAM, probably at least 32gb. If she games or does 3d rendering probs gotta look into a decent GPU too. Also look at pugetsystems' articles, they have some good information on these kind of builds.
This is not for a audio creator. I see that all channels now only concentrate on graphics like what about the audio creators 😂😆 how does it work with DAW and vst instruments ?
Motherboard plays as big of a role as the IMC in most cases, as maintaining a stable connection with such speeds from the ram slots to the cpu socket is a big factor too. Most motherboards will have a verified minimum rating to accompany what the cpu itself can handle. Also hot tip for people who build in small form factor PCs for creator workloads and high ram demands; G.Skill and Corsair make some very competitive high-speed 96GB dual ram kits (2x48gb), and they're perfect for ITX, or just for anyone who don't wanna take the speed penalty risk of fully populating a 4-slot board. Works wonders and is such a big step up from the 64gig standard, especially in 3d rendering and other ram capacity intensive loads. And the stability offering great speed for those evening gaming sessions with zero compromise :D
That the DDR5-6800 and DDR5-7800 are sometimes slower shows that the RAM XMP Profiles were unstable. If fully stable the scores for the faster RAM would be noticeably higher. You and Jay2Cents did similar mistakes. His XMP profiles were also unstable. And this will happen to many people who don't know it better and buy this higher end Ram Kits... Thats why PC Building is not as easy as it seems... People gonna run into problems or will be missinformed after watching videos like this or the one from Jays2Cents.
@@tek_soup no he doesn't try. If he would there would be more research into this unexpected behaviors. What he is doing is feeding the Algorithm and trying to generate clicks and affilate sales like most other TH-camrs. And because I'm a SI I constantly have to deal with misinformed people/customers.
@@tek_soup Nothing to do with not liking him. He's a Businessma like I am. It's an objective and honest critique of his content. And I do this effort because my customers link his videos sometimes when I consult them. I do sell creator PCs. His Videos often have misinformation, and when I spot it I try to educate in hope he reads it and does it better the next time. I told you I'm a SI (System integrator) and have to deal with misinformed people all the time.
oh ok, they ALL Leave Something out. i usually leave a comment bitching about it, after i try to go do it. i just setup 25gbe between two computers, id say 99% vids on networking 10gbe, left out all the important stuff, they just all talk about the hardware and wire, but Nothing about, setting it up. Constantly, there is half ass videos, and unfinished guides, i see. yeh pisses me off. He left out, that you need to stick the ram in 1B and 2B on the sticks of ram, i wasted hours, trying to figure out. Cause ive only bought the full amount. on his previous Ram Video. @@BarisYener
Very useful, unfortunately V-Ray, Blender and Octane were not in the tests. It is also necessary to repeat the tests in AMD due the impact of the RAM speed on Infinity Fabric. And then, one question: if I have two 64GB kits, I must put the two sticks of a same kit on channel A and the other two on channel B or I must put a one stick of each kit on channel A and so on.
He also had unstable XMP, thats why some DDR5 6800 and 7800 results are slower. If it was fully stable, the scores for DDR-6800+ should be even higher.
I'd say, if you multitask, 4 dimms(128/192gb) at 4800 is a lot better. For very specific purposes or if you want to save money, 2 dimms (96gb) 6400 will do.
These tests don't represent the actual RAM kits on the market though, since only the frequency was lowered and not the CAS latency. Most "good" RAM kits on the market aim for at least 10ms latency. To calculate RAM latency, multiply the CL by 2000 then divide by the frequency (or use a RAM latency calculator). In your next video, why not test different kits that have similar RAM latency? Here are some examples of 10ms kits: 5600MHz CL28 6000MHz CL30 6400MHz CL32 6800MHz CL34 7200Mhz CL36 7600MHz CL38 8000MHz CL40
So it's still the same issue as previous generations. More than two sticks starts to become a stability issue, unfortunate, but at DDR5, really shouldn't be much of an issue. Making 4800mhz the new 3200mhz stable.
The total system cost increase to go above 6000Mhz is exponential. It's only worth it for a very few people to do that. I think right now, 6000Mhz is the sweet spot.
So nearly doubling the frequency and yet just a 3-8% increase 😅 was ddr5 a gimmick just to sell more ram and newer kits ? I personally prefer a stable system than an unstable one with random crashes that will just negate that 3% increase. If i find myself rebooting every hour 😂
Let me just clarify, it is not Mhz bu MT/s (which devided by 2, as DDR has in it's name) = Mhz. Interesting analysis, but you should have mentioned their CL next to speed to get better picture.
dude, didnt watch the video, but .. :D teamgroup t-create makes great ddr5 ram kits, t-create sounds like for creators and its hynix-m die low tier for low latency or hynix-a die high tier for high rates (good for intel) its low profile 33 mm compatible with all air coolers and has good thermal pad which allows them to be overclocked to 8000 mhz without need for extra cooling, t-create expert ddr5 7200 hynix-a die is top tier ram my recommendation to all whose motherboards can handle it, for amd or lower tier intel systems there are 6000 cl30, 6400 cl32 versions.
I'm planning to jump soon on AM5, already ordered 7800X3D, what would you highly recommend exactly kit for best budget RAM, as well for the motherboard,,,gaming only my setup, no overclocking,,,I would appreciate any feedback
@@gezimlimoni2319 7800x3d hardly benefits from ram speeds or oc'd ram its main advantage is cache, 6000 cl30 g skill flare x5, t-create or any other brand should be sufficient without oc, for the motherboard there is asrock b650m hdv for 120 and 200 ish msi b650m mortar/b650 tomahawk boards that are best value boards with good vrm and audio (black).
well going to build 5k Euro PC in Q1 2025, my 5800x3d/6950xt will be too slow when I get my 480Hz 1440p oled. if I am going 5K for PC I will be sure I am getting 8000Mhz 48gb or even 8400Mhz 48gb ram memmory for ryzen 9800x3d just to squize 10fps more and better 1% lows, I know I will use ram mem for like 6 years so why not pay more for it, my 3200Mhz is 7 years old I use for 5800x3d.
You left out some factors as others have said. On a z790 Apex encore with 14900ks and 4090 strix at 4k gaming a 48 gb 8200mhz liquid cooled kit will be within 1-2 fps in things as a 64gb 6400 mhz kit so no speed isn’t everything. Due to games like DCS hitting over 40 gb of ram usage in large mp servers I took my 8200 kit out and put the 64 back in after testing.
maybe high speed RAMs would show it self in low capacity and mostly in copy or transferring large data. these benchmarks are for CPU speed(most CPU clock limited to ~4.8ghz) either the RAM at 4800mt/s, and are not compatible with RAM speed only and above 5.6ghz which is more than cpu clock. unziping 1tb of data might be the better test.
It is bcos of the subtimings. Using xmp etc, usually slower ram gets better timings but also subtimings. So just by intyping it manualy at high speeds u will get better results. Worth to say that some programs relay on certain subT so thats why u might think it is faster. But it is not. 4800 for sure is good for 192 or 256GB rig.
this seems like an intel platform judging by the mobo but i didnt see it stated explicitly yes intel cares about ram "speed" amd ofc is the opposite, you will do better with lower timings architecture differences, thought it wouldve been mentioned since a blanket statement is incorrect. also 4 dims will never reach the same "speeds" as a 2 dimm board. this has to do with the memory controller. and a touch about the mobo layouts as well
Thanks for this video.... Very helpful! I've been a freelance video editor for almost 4 years now. While I'm running 128GB of RAM, clocked at 4800MHz, I always thought higher RAM frequency mattered more than memory capacity.
The best Sweetspot is a Ram with the highest Clockspeed at the lowest Latency u can get. If u can get a DDR5 CL30 8000mhz Ram and it is working with your CPU, BUY IT!
Low Latency is in my case the most important thing in a Ram, because the clockspeed will always fast enough for 99% of all tasks. I would say everything higher than 6000mhz I'd waste of money, a 8000mhz can give more Performance, but with what Latency? C48? It's a big Latency bump compared to Cl30 and Latency is really important, more than Clockspeed...
Easy: if your CPU ONLY eating for example only up to 6000mhz and everything above 6000mhz its crashes, that means, buy 6000mhz with the lowest CAS Latency you can find in the market, check every Latency time, compare and buy the one with the lowest number... And if you want an 8000mhz Kit, your motherboard is a second factor and Windows a third factor that is coming in if it's works or not
I have this same motherboard with a x2 16 GB 8000MT RAM inside of it running stable at 7400MT. All you need to do is tweak the voltages a bit. Set DRAM volts to 1.44 or 1.45. Can even go to 1.5 if you have enough circulation on your case. Then downvolt with an offset of about -.060 with Intel's XTU and lower the Watts/volts a bit and you're good. Would completely change your results.
I run 6000mhz 64gb with Ryzen 7900x And it's very stable after motherboard firmware update. Not a single crash/bluescreen for month's But I choose RAM very carefully and check the memory support lists like crazy. I will be able to later upgrade to 128gb and keep the speed according to Asus memory support list
I'd like to see a follow up video to this for XMP profiles and for x2 48gb =96GB of ram. I've using my CORSAIR VENGEANCE DDR5 RAM 96GB (2x48GB) 6400MHz CL32 6400Mhz. I'm running mine at 6400Mhz.
Hopefully g skills will have 10 ghz ram with the tightest timings and lowest latency as possible. I'm waiting for the day when ram is 10ghz, all cpu cores including performance and efficiency cores running at 10 ghz and gpu 10ghz, even 100gb read speeds and 100 write speeds ssd for transferring and downloading uncompressed 4k video files. The option for no decoding and no encoding while working with apps with no compression at all. Basically I want the speeds to be done with a single click of a button with no waiting time at all including games with no loading screens at all during the era of 90s game consoles like the super nintendo, sega genesis and n64 since the games would go instantly to the main menu since there is no loading screen. As soon as I click the button on a mouse the 100gb or 1tb file is already done transferring the files or done downloading instantly since I don't like waiting and prefer to be done without waiting. For resolution always 4k even when I play games. I wish games support dual gpus, evga rtx titan and evga rtx kingpin in the same pc build with native 4k resolution, no upscaling at all, ray tracing always enabled, dlss always off, motion blur always off, v sync always off, unlimited fps always enabled for the supported games and can it run crysis mode in every graphics settings for crysis remastered and ray tracing experimental boost always enabled so the game can demand every bit of hardware with unlimited settings since that's how I always game in 4k. I wish there was an option for can it run crysis mode in every games graphics in the graphics settings with ray tracing experimental boost.
@@theTechNotice in this case you just need a high rez volume simulation like fire or smoke, or water simulation. They can saturate the RAM very fast :)
6:07 Keeping the timings the same while decreasing frequency means that you increase latency, because the timings are expressed in clock cycles. Fewer clock cycles per second = more time per clock cycle. So you're testing both lower throughput and higher latency at the same time. The problem with that, is that lower-clocked kits also usually have timings that keep latency pretty much the same as the higher clocked ones. So if the idea was to show what kit to buy, you should have adjusted timings to keep latency constant. If some benchmarks are very latency-dependent, your results will not reflect reality. Also as others have pointed out, those are not MHz but MT/s. DDR has 2 transfers per clock cycle, hence the name "Double Data Rate". You keep switching between both as if they're the same thing, you need to sort out that confusion for your next videos 🙂
Hello for DDR I am having ASUS pro art X570 MB with 64 Gb Gskill Ripjaw 3600 RAM (ryzen 9 5900x CPU) but this ram is not mentioned in the QVL list. so should I change while upgrading to 128gb or Non QVL ram can be used and enabled XMP? Kindly suggest
Happy with my two sticks of ddr5, 64gb at 6000, but wonder if I could keep those speeds with 4 sticks and 128gb? I will look into that. Thanks for another great video👍
Love your explanations and videos. Do you have any video on RAM and CPU and SSD for software development...virtualization? I use Visual Studio, hyper-v, VS Code and sometimes I like to work on Cyberlink PowerDirector to learn how to edit videos and audio. Would love to see more recommendations for creators, developers/programmers, and IT specialists.
And the performance for DAW? As you should know VSTi’s are very RAM demanding in capacity so 4 Dimms are pretty much a must, so what’s the optimum there? I think you may find there are immensely more audio creators as video and find it a shame that audio community are a bit forgotten, maybe if you started a audio creator channel you could, with quality content that you try to produce for video, a large following develop quite easily, had hopes for more audio focus from you as at the start there was some, I believe increased for a while, but feel like it’s been dropped again……
Put the ram in dimm slots a2 and b2. Do not put it in slots a1 and b1. Asus mothernoards have a bios issue where xmp profiles will not run well on ram on slots a1 and b1. I ran into this same issue. Once I put it in a2 and b2 it worked fine. I have the same exact motherboard.
yeh, i learned that the hard way yesterday, i left a comment like yours, on his ram video. cause i am 99% always filling it up, went for speed this time.
You just made a great case for "Diminishing Return" which is way too common in the IT industry in recent years... Also a great example of false advertisement, these RAM sticks do not say that the advertised speed is OC (and sadly theoretical more than practical) plus they don`t mention that adding more sticks will lower frequency even more! To get the conclusion and final point 7800MTs vs 4800MTs is 10% faster (in synthetic benchmarks, the real-world application can be even lower) so on 1 hour-long video render you get a whopping 6 min shorter wait!!! And the price difference between those modules isn`t exactly 6min / 10% compatible :P Sadly entire IT industry has turned to Quick Money Grab...thankfully they usually put RGB lights on top of RAM sticks, I guess it`s to make you feel less sad while your butt is still hurting (for paying premium) !
Please read more comments... His results here are inaccurate. The faster RAM Kits should have higher scores but they don't because it was unstable. It's the same BS like jays2cents. This is overclocking and he is uneducated in this regards. It's the typical normie buying high-end hardware and not understanding the results here. And the manufacturers don't highlight this possible problems but they do communicate it with extra ** notes etc. As usual, rtfm
@@BarisYener "This is overclocking and he is uneducated in this regards. It's the typical normie buying high-end hardware and not understanding the results here" I love it when people consider entering bios and turning on the XMP profile "overclocking", btw I thought that the entire point of buying high-end hardware was to see the difference in performance (aka get better performance) but if you say that understanding the results of synthetic benchmarks is the priority...then I will throw away my 24 year experience of building PCs and buy the book "Synthetic Benchmark Results for Dummies - how to understand something you don`t see" !
@@alexdimitri9669 I'm not sure if you intentionally misunderstand me or if I was not clear enough. His results are flawed since the scores he get in shown puget benchmarks (which are NOT synthetic btw.) are lower then expected. This too low difference and sometimes even worse performance scores with the higher end XMP indicate that the RAM was runing unstable and that the on-die ECC correction brought him through those tests (which again, are NOT synthetic) but with the cost of performance. And it's also a phenomenon that people who build a couple of PCs in their live think they are experts. And from what you write I can clearly assume that you have no experience in overclocking and what terms you need to watch out for. Yes manufacturers are not clear enough in communicating some of this stuff, but they do and people who are PROs know about this stuff. You clearly don't. Typical Dunning-Kruger-Effect.
@@alexdimitri9669 "To get the conclusion and final point 7800MTs vs 4800MTs is 10% faster (in synthetic benchmarks, the real-world application can be even lower) so on 1 hour-long video render you get a whopping 6 min shorter wait!!! " It's not useless to get free performance. The Module prices are very close for example fast DDR5-5600 is not much different to DDR-6400. You can buy Hynix A-DIe and tune it to even higher MT/s and get free performance. This rubbish you write here is just the problem on the internet. If you tune your DDR5 right, there can be up to 15% - 20% performance gain in real world szenarios. Maybe even more. DDR5 is a complicated technology since you need 3 parts that can run the higher MTs + the knowledge to tune it. Puget Benchmarks are real world tests. PS: I'm not wasting more of my time here in the comments. Go do you! Bye.
![DON'T BUY THIS CPU! 👉 4 Mistakes to AVOID as Creators! [More Expensive & WORSE]](http://i.ytimg.com/vi/8M_zoDKw7XA/mqdefault.jpg)
Silicon Pottery, did anyone catch that? 🤔 😅
Yes! Haha
@@bsatyam- It made me laugh. Silicone Lottery would be a better and more appropriate name for it.
😂😂plenty of seasoned creators know exactly why you call it "pottery". The L word gets your content blocked, regardless of context!
lol - yes, we caught it... nice play on words... letters ;) Shared it with a YT creator friend of mine.
It should actually mean silicon prediction.
You left out a huge equation from this video and that is that 99% of all 4 dim motherboards can not run high-speed ram sticks like 8000 for those type of speeds you will need 2 dim motherboards like the Asus Apex and Gigbyte Tachyon. Also pretty sure all of your results on the higher speeds were incorrect and mostly likely negative due to instability of the pro art board being a 4 Dimm motherboard.
Yes, his scores are a result of unstable XMP. I also already mentioned it in my comments.
he mentioned it near the end, he uses 4 sticks in his build, and had to lower the speed. i have 192 gigs in one of my z790 stable at 5200. they dont mention that you can use that much ram. except that my z790 maximus extreme does.
Im doing 16gbs x 4 rams at 6000mhz no issue. 14900k and mag b70m wifi II.
yes because that extremely slow ram for intel @@azntactical4884
frame chasers in youtube the best to learn for ram
For videos like this, at least to get the system setups right, you should really collaborate with guys like BuildZoid from @ActuallyHardcoreOverclocking or others who are related to the XOC scenario. The moment you went with a 4 dimmer for validating 8000. You had me off. Some of the best 4 Dimmer (Ex. 790 Aorus Master X, 790 Aorus Elit X AX, 790 PG Nova, etc.) can only reach 7600-7800 with a really good IMC. Also, some of the major performance benefits come from the tighter secondary and tertiary timings of DDR5 and don't solely rely on the speed and primary timings alone. 7600 with tighter timing can outperform 8000 with untuned or loose timing, and XMP is kind of great at that just to make it work. So this video won't represent a super-accurate apples-to-apples comparison in terms of the scalability of true RAM performance. FYI, for similar reasons, Jay's video is also invalid.
100% true and great comment
Jay is not as much of an expert as he pretends to be. There are many inaccuracies, dubious approximations and hazardous conclusions in his videos. I'd have more trust in Linus of LTT, he has better knowledge, intuition and rigour in his logic, and if he's presenting, he probably vetted the script. But the one who usually pays the most attention to accuracy is Steve of Gamers Nexus, although I don't think that he covers RAM unfortunately (haven't checked).
@@dismuter_yt check out actually hardcore overlocking. he covers both intel and amd ram overclocking and also cpu and gpu overclocking. hes probably the best source of information in regards to ram. also unlike many overclockers, he actually does tell you the reasons as to why he overclocks something, like lets say cpu/gpu frequency, ram transfer speeds, ram timings, voltages, what a certain timing does, what a certain voltage is for, how certain ram timings affect things, why some timings dont matter, etc. etc. hardware unboxed had him do some ram overclocking once and they did some benchmarks. hes a great source of information. i learned to overclock after watching his videos
I wasn't even able to finish watching this LOL!
would love to see a comparison of the corsair dominator vs t create expert
It does seem like though, that it's generally been accepted that 6000-6400 is a stable setup with 2 DIMMs installed, so there is that.
Anyways, I'm glad this channel exists to benchmark and review PC hardware outside of gaming contexts, this was sorely needed in the PC review space.
Yesterday I mentioned to a creator friend of mine that populating all RAM slots with very fast DDR5 sticks could actually result in his motherboard reducing the speed he will actually get and he said that makes no sense as a higher total RAM memory quantity will always benefit a creator.
Thank you Lauri.
You are correct, under a certain situation - overclocking.
Outside of that, assuming the ram is in his QVL list or has a high compatibility, he will get the rated speed and timings.
Quantity, as your friend suggests, is correct assuming 'all' the ram is used, and the system is now going into swopfile.
He would likely benefit from more cache and cores on his CPU.
Also, timings may well show a good improvement in reducing the time taken to render or produce his work, obv each app he uses will differ in how it uses and allocates resources from the hardware.
It's actually pretty massive. Im talking losing like 1200mhz just going from 2 to 4 sticks
In the old days... Populating them all with smaller sticks to total a larger ammount 'was' the best thing to do... newer motherboards changed the way the Memory lanes are done apparently, so your better off with 'pairs' unless your using a really really expensive MB that can run 4.
*quick edit*
Found this out myself recently while getting ready for a new PC rebuild/upgrade... had ordered 4 sticks of 16 to get the 64 i wanted while filling up all the slots... found out the next day that would actually overload the Memory control module and cause slowdowns.... soooo gotta return those 4 when they get here tomarrow, and exchange them for a pair of 32gb sticks.
Ended up opting for a 128GB (32GB x 4) kit of Kingston Fury Beast DDR5 5600MT/sec. I had considered 96GB (2 x 48GB) but as there is so little choice available, the pricing seems unreasonably high imo.
Hoping that even if I take a small hit on the speed that my RAM runs at because I am populating all four DIMM slots, that the extra total RAM capacity balances that out somewhat when working with big projects.
Were any actual stability test performed with 7800 Mhz? Your board is not capable of such speeds, specs on asus website are BS. You need APEX for sure stable 8000. Please run y-cryncher VST for 4+ hours. Active ram cooling is a must. If ram is unstable then your tests are bad.
Makes no fckn sense
Good, helpful video. A kindly gentle correction starting at 3:36 - Asus has a dedicated IMC SP score called MC SP, which is available on their higher-end mobos (I don't know where the cutoff is, but I know a board like the z790 Tuf Gaming doesn't have it available in the bios/uefi). From within the bios, you have to actually run the Get MC SP the first time to get a score (unlike the SP score). If it is available on your mobo, It is found under Extreme Tweaker/AI Features/Get MC SP. So, the SP score does not relate directly to the IMC. I hope this is helpful :)
download more ram
Do you know a pirate webpage to download it for free?
God, I love you for creating PC content for creative work. Building a PC for gaming was easy because information is everywhere. However, building a PC for my girlfriend, who is a graphic designer, is much harder. There is very little information about what components are better to choose and why, etc.
Not sure what programs your girlfriend uses but if she uses adobe programs like Photoshop they generally benefit from faster single core CPUs and lots of RAM, probably at least 32gb. If she games or does 3d rendering probs gotta look into a decent GPU too. Also look at pugetsystems' articles, they have some good information on these kind of builds.
10% improvement is negligible compared to the loss of stability you might encounter
ECC RAM is a better investment rather than 8000MTs RAM
This is not for a audio creator. I see that all channels now only concentrate on graphics like what about the audio creators 😂😆 how does it work with DAW and vst instruments ?
Motherboard plays as big of a role as the IMC in most cases, as maintaining a stable connection with such speeds from the ram slots to the cpu socket is a big factor too. Most motherboards will have a verified minimum rating to accompany what the cpu itself can handle.
Also hot tip for people who build in small form factor PCs for creator workloads and high ram demands; G.Skill and Corsair make some very competitive high-speed 96GB dual ram kits (2x48gb), and they're perfect for ITX, or just for anyone who don't wanna take the speed penalty risk of fully populating a 4-slot board. Works wonders and is such a big step up from the 64gig standard, especially in 3d rendering and other ram capacity intensive loads. And the stability offering great speed for those evening gaming sessions with zero compromise :D
That the DDR5-6800 and DDR5-7800 are sometimes slower shows that the RAM XMP Profiles were unstable. If fully stable the scores for the faster RAM would be noticeably higher. You and Jay2Cents did similar mistakes. His XMP profiles were also unstable. And this will happen to many people who don't know it better and buy this higher end Ram Kits... Thats why PC Building is not as easy as it seems... People gonna run into problems or will be missinformed after watching videos like this or the one from Jays2Cents.
There is a Millions of Misleading videos. on you tube, take your pick. at least he is trying to do the right thing.
@@tek_soup no he doesn't try. If he would there would be more research into this unexpected behaviors. What he is doing is feeding the Algorithm and trying to generate clicks and affilate sales like most other TH-camrs. And because I'm a SI I constantly have to deal with misinformed people/customers.
well then, why you watching? you don't like him or his channel. @@BarisYener
@@tek_soup Nothing to do with not liking him. He's a Businessma like I am. It's an objective and honest critique of his content. And I do this effort because my customers link his videos sometimes when I consult them. I do sell creator PCs. His Videos often have misinformation, and when I spot it I try to educate in hope he reads it and does it better the next time. I told you I'm a SI (System integrator) and have to deal with misinformed people all the time.
oh ok, they ALL Leave Something out. i usually leave a comment bitching about it, after i try to go do it. i just setup 25gbe between two computers, id say 99% vids on networking 10gbe, left out all the important stuff, they just all talk about the hardware and wire, but Nothing about, setting it up. Constantly, there is half ass videos, and unfinished guides, i see. yeh pisses me off. He left out, that you need to stick the ram in 1B and 2B on the sticks of ram, i wasted hours, trying to figure out. Cause ive only bought the full amount. on his previous Ram Video. @@BarisYener
Great video! Would love to see a latency comparison between a bunch of sticks.
Very useful, unfortunately V-Ray, Blender and Octane were not in the tests. It is also necessary to repeat the tests in AMD due the impact of the RAM speed on Infinity Fabric. And then, one question: if I have two 64GB kits, I must put the two sticks of a same kit on channel A and the other two on channel B or I must put a one stick of each kit on channel A and so on.
we would love to see latency tests as well! amazing video
Noted!
it's not megahetz it's megatransfers, as in MT/s
DDR speed is measured in MT/s, the Kingston box is accurate
---
ASUS has also a IMC score 🤫
I actually hope to never see a jay 2 sent video again ever in my life.
100%
He also had unstable XMP, thats why some DDR5 6800 and 7800 results are slower. If it was fully stable, the scores for DDR-6800+ should be even higher.
Hey man I was just wondering the same, thanks
Glad I could help
I'd say, if you multitask, 4 dimms(128/192gb) at 4800 is a lot better. For very specific purposes or if you want to save money, 2 dimms (96gb) 6400 will do.
more ram is great for primo cache and speed up a hard drive. insane fast.
i love my 192GB vengeance quad kit.. im not that much into gaming, its more for Adobe, davinci etc. and its perfect
These tests don't represent the actual RAM kits on the market though, since only the frequency was lowered and not the CAS latency.
Most "good" RAM kits on the market aim for at least 10ms latency. To calculate RAM latency, multiply the CL by 2000 then divide by the frequency (or use a RAM latency calculator).
In your next video, why not test different kits that have similar RAM latency?
Here are some examples of 10ms kits:
5600MHz CL28
6000MHz CL30
6400MHz CL32
6800MHz CL34
7200Mhz CL36
7600MHz CL38
8000MHz CL40
would love a video on how latency affects RAM speed
So it's still the same issue as previous generations. More than two sticks starts to become a stability issue, unfortunate, but at DDR5, really shouldn't be much of an issue. Making 4800mhz the new 3200mhz stable.
I would really have liked to see a 3D rendering software tested. Something like corona benchmark
would you render that on GPU?
@@theTechNotice no 50% of renders are cpu based
Now with X870E series, you can run faster ram more efficiently and consistently.
The total system cost increase to go above 6000Mhz is exponential. It's only worth it for a very few people to do that. I think right now, 6000Mhz is the sweet spot.
So nearly doubling the frequency and yet just a 3-8% increase 😅 was ddr5 a gimmick just to sell more ram and newer kits ? I personally prefer a stable system than an unstable one with random crashes that will just negate that 3% increase. If i find myself rebooting every hour 😂
99% Jays sayd is bull 💩
Soon as I got 8000 stable I've never tried to go higher
Hello TN!
Can you make comparison on the just release new Blender 4.1 version on an Intel Arc 770?
Let me just clarify, it is not Mhz bu MT/s (which devided by 2, as DDR has in it's name) = Mhz. Interesting analysis, but you should have mentioned their CL next to speed to get better picture.
dude, didnt watch the video, but .. :D teamgroup t-create makes great ddr5 ram kits, t-create sounds like for creators and its hynix-m die low tier for low latency or hynix-a die high tier for high rates (good for intel) its low profile 33 mm compatible with all air coolers and has good thermal pad which allows them to be overclocked to 8000 mhz without need for extra cooling, t-create expert ddr5 7200 hynix-a die is top tier ram my recommendation to all whose motherboards can handle it, for amd or lower tier intel systems there are 6000 cl30, 6400 cl32 versions.
I'm planning to jump soon on AM5, already ordered 7800X3D, what would you highly recommend exactly kit for best budget RAM, as well for the motherboard,,,gaming only my setup, no overclocking,,,I would appreciate any feedback
@@gezimlimoni2319 7800x3d hardly benefits from ram speeds or oc'd ram its main advantage is cache, 6000 cl30 g skill flare x5, t-create or any other brand should be sufficient without oc, for the motherboard there is asrock b650m hdv for 120 and 200 ish msi b650m mortar/b650 tomahawk boards that are best value boards with good vrm and audio (black).
It's 8000 MT/s, it's running at 4000 MHz.
And the memory controller would be at only 2000 mhz on Intel since you're in gear 2
I'm going to make a comment saying mhz and not mt/s simply for the engagement.
Appreciate that, I'll answer Mt/s for the engagement!
well going to build 5k Euro PC in Q1 2025, my 5800x3d/6950xt will be too slow when I get my 480Hz 1440p oled.
if I am going 5K for PC I will be sure I am getting 8000Mhz 48gb or even 8400Mhz 48gb ram memmory for ryzen 9800x3d just to squize 10fps more and better 1% lows, I know I will use ram mem for like 6 years so why not pay more for it, my 3200Mhz is 7 years old I use for 5800x3d.
wish you add benchmarks for v-ray, blender, audio / video encode/decode
They really said "mini-itx who?" with these RAM sticks LMAO
Get AMD Ryzen 7000 or 8000 series CPU with Asrock B650E or X670E motherboard..
I am so afraid of stability issues when it comes to ram speed I only buy stock speed and save the extra money.
My kinda guy!
Jay was proven to be wrong in that video… again. He really needs to stop mass producing garbage videos.
You left out some factors as others have said. On a z790 Apex encore with 14900ks and 4090 strix at 4k gaming a 48 gb 8200mhz liquid cooled kit will be within 1-2 fps in things as a 64gb 6400 mhz kit so no speed isn’t everything. Due to games like DCS hitting over 40 gb of ram usage in large mp servers I took my 8200 kit out and put the 64 back in after testing.
maybe high speed RAMs would show it self in low capacity and mostly in copy or transferring large data. these benchmarks are for CPU speed(most CPU clock limited to ~4.8ghz) either the RAM at 4800mt/s, and are not compatible with RAM speed only and above 5.6ghz which is more than cpu clock. unziping 1tb of data might be the better test.
I have this question before and you replied now you created a new video with this lol nice
I think you need to take timings into consideration...
Yes! Working on that idea!
It is bcos of the subtimings. Using xmp etc, usually slower ram gets better timings but also subtimings. So just by intyping it manualy at high speeds u will get better results. Worth to say that some programs relay on certain subT so thats why u might think it is faster. But it is not. 4800 for sure is good for 192 or 256GB rig.
brilliant vid, thx!! 👏👏👏
AMD responds better with lower timings , same freq, vs Intel. Basically for most games 2800mhz cl14 delivers same fps as 3800mhz standard cl.
this seems like an intel platform judging by the mobo
but i didnt see it stated explicitly
yes intel cares about ram "speed"
amd ofc is the opposite, you will do better with lower timings
architecture differences, thought it wouldve been mentioned since a blanket statement is incorrect.
also 4 dims will never reach the same "speeds" as a 2 dimm board.
this has to do with the memory controller. and a touch about the mobo layouts as well
This is my motherboard but i cant see it working..even though it is aproved...plus f corsair..i stick with 96 2 dims st 6400 is better😊
Thanks for this video.... Very helpful! I've been a freelance video editor for almost 4 years now. While I'm running 128GB of RAM, clocked at 4800MHz, I always thought higher RAM frequency mattered more than memory capacity.
Thanks for sharing!
Great video keep up the good work I have the Corsair platinum ddr5 7200 32gb using i7 13700k
Can you also test this ram with 3d software? Zbrush, maya, substance painter for example. Great vid!
Why are you using the Orange Blue theme. that's my theme
now put a fan on your ram and increase tREFI to max and you will see the real difference
Ok but what about CL latency?
Very good question!
So I cannot benefit from fast RAM unless I overclock RAM?!!
But what is the point they sell such ram and such speed if they do not seem to work?
The best Sweetspot is a Ram with the highest Clockspeed at the lowest Latency u can get. If u can get a DDR5 CL30 8000mhz Ram and it is working with your CPU, BUY IT!
Low Latency is in my case the most important thing in a Ram, because the clockspeed will always fast enough for 99% of all tasks. I would say everything higher than 6000mhz I'd waste of money, a 8000mhz can give more Performance, but with what Latency? C48? It's a big Latency bump compared to Cl30 and Latency is really important, more than Clockspeed...
Easy: if your CPU ONLY eating for example only up to 6000mhz and everything above 6000mhz its crashes, that means, buy 6000mhz with the lowest CAS Latency you can find in the market, check every Latency time, compare and buy the one with the lowest number...
And if you want an 8000mhz Kit, your motherboard is a second factor and Windows a third factor that is coming in if it's works or not
7800 was 100% not stable on that 4 dimm motherboard.
Try 5600mhz vs 7200mhz?
Test games next pls
I have this same motherboard with a x2 16 GB 8000MT RAM inside of it running stable at 7400MT. All you need to do is tweak the voltages a bit. Set DRAM volts to 1.44 or 1.45. Can even go to 1.5 if you have enough circulation on your case. Then downvolt with an offset of about -.060 with Intel's XTU and lower the Watts/volts a bit and you're good. Would completely change your results.
I run 13700kf cant run 100% with 7200 34-44-44-44-96 :( and need lower it to 6800 but timing is very low 32-39-39-39-28 with xmp tweak.
Its 2x 16gb
Plz tell me which one is better 6400mhz cl32 or 6800mhz cl34 for ryzen 5 7600
I run 6000mhz 64gb with Ryzen 7900x
And it's very stable after motherboard firmware update.
Not a single crash/bluescreen for month's
But I choose RAM very carefully and check the memory support lists like crazy.
I will be able to later upgrade to 128gb and keep the speed according to Asus memory support list
I'd like to see a follow up video to this for XMP profiles and for x2 48gb =96GB of ram. I've using my CORSAIR VENGEANCE DDR5 RAM 96GB (2x48GB) 6400MHz CL32 6400Mhz. I'm running mine at 6400Mhz.
Hopefully g skills will have 10 ghz ram with the tightest timings and lowest latency as possible. I'm waiting for the day when ram is 10ghz, all cpu cores including performance and efficiency cores running at 10 ghz and gpu 10ghz, even 100gb read speeds and 100 write speeds ssd for transferring and downloading uncompressed 4k video files. The option for no decoding and no encoding while working with apps with no compression at all. Basically I want the speeds to be done with a single click of a button with no waiting time at all including games with no loading screens at all during the era of 90s game consoles like the super nintendo, sega genesis and n64 since the games would go instantly to the main menu since there is no loading screen. As soon as I click the button on a mouse the 100gb or 1tb file is already done transferring the files or done downloading instantly since I don't like waiting and prefer to be done without waiting. For resolution always 4k even when I play games. I wish games support dual gpus, evga rtx titan and evga rtx kingpin in the same pc build with native 4k resolution, no upscaling at all, ray tracing always enabled, dlss always off, motion blur always off, v sync always off, unlimited fps always enabled for the supported games and can it run crysis mode in every graphics settings for crysis remastered and ray tracing experimental boost always enabled so the game can demand every bit of hardware with unlimited settings since that's how I always game in 4k. I wish there was an option for can it run crysis mode in every games graphics in the graphics settings with ray tracing experimental boost.
I wish you had tests in 3D applications too.
I'd need some big projects that would saturate the RAM instead of GPU rendering. ON GPU rendering there is no difference :)
@@theTechNotice in this case you just need a high rez volume simulation like fire or smoke, or water simulation. They can saturate the RAM very fast :)
6:07 Keeping the timings the same while decreasing frequency means that you increase latency, because the timings are expressed in clock cycles. Fewer clock cycles per second = more time per clock cycle. So you're testing both lower throughput and higher latency at the same time.
The problem with that, is that lower-clocked kits also usually have timings that keep latency pretty much the same as the higher clocked ones. So if the idea was to show what kit to buy, you should have adjusted timings to keep latency constant. If some benchmarks are very latency-dependent, your results will not reflect reality.
Also as others have pointed out, those are not MHz but MT/s. DDR has 2 transfers per clock cycle, hence the name "Double Data Rate". You keep switching between both as if they're the same thing, you need to sort out that confusion for your next videos 🙂
15:41 Extra 12 GB... your math is correct, 48 - 36 = 12. However the Kingston kit is 32 GB, not 36 GB 😅
So it's 16 GB more for the Corsair.
Creators yes it is better but for general creating aftwr 6400 its dinishing returns period
Hello for DDR I am having ASUS pro art X570 MB with 64 Gb Gskill Ripjaw 3600 RAM (ryzen 9 5900x CPU) but this ram is not mentioned in the QVL list. so should I change while upgrading to 128gb or Non QVL ram can be used and enabled XMP? Kindly suggest
What about the newer JEDEC DDR 5 standard? 5600 MT/s without any XMP nor EXPO. As an industrial standard should also work with 4 modules.
Happy with my two sticks of ddr5, 64gb at 6000, but wonder if I could keep those speeds with 4 sticks and 128gb? I will look into that.
Thanks for another great video👍
jay is almost never right
Love your explanations and videos. Do you have any video on RAM and CPU and SSD for software development...virtualization? I use Visual Studio, hyper-v, VS Code and sometimes I like to work on Cyberlink PowerDirector to learn how to edit videos and audio. Would love to see more recommendations for creators, developers/programmers, and IT specialists.
Hello Sir can I install 16gb single stick Corsair 5200mhz
I am big fan of you guys!
Have some 6000 and 6400 kits, tried hard to QVL pair between all 3 machines. 2 sticks for balancing.
intel have found their homie !!! very informative video anyway :)
Kingston is A-Die
Corsair is M-Die
Jay's video was wrong. He was obviously unstable.
What rams are best for AMD?
Great video!
Great channel!
I have to agree with Jayz 2 Cents.
And no games?
Can too high memory speed cause black screens while gaming?
yes of course
these are not megahertz
BEST
Gigabyte also has silicone prediction but they call it biscuits lol
lol
And the performance for DAW? As you should know VSTi’s are very RAM demanding in capacity so 4 Dimms are pretty much a must, so what’s the optimum there? I think you may find there are immensely more audio creators as video and find it a shame that audio community are a bit forgotten, maybe if you started a audio creator channel you could, with quality content that you try to produce for video, a large following develop quite easily, had hopes for more audio focus from you as at the start there was some, I believe increased for a while, but feel like it’s been dropped again……
help me spread the latest poll on the channel :)
@@theTechNotice what poll?
Do a latency test!
48-32=16 gb
3:57
East Asian students
👍🏻👍🏻👍🏻👍🏻👍🏻
48-36=12 but it is 48-32 LOL
Put the ram in dimm slots a2 and b2. Do not put it in slots a1 and b1. Asus mothernoards have a bios issue where xmp profiles will not run well on ram on slots a1 and b1. I ran into this same issue. Once I put it in a2 and b2 it worked fine. I have the same exact motherboard.
yeh, i learned that the hard way yesterday, i left a comment like yours, on his ram video. cause i am 99% always filling it up, went for speed this time.
You just made a great case for "Diminishing Return" which is way too common in the IT industry in recent years...
Also a great example of false advertisement, these RAM sticks do not say that the advertised speed is OC (and sadly theoretical more than practical) plus they don`t mention that adding more sticks will lower frequency even more!
To get the conclusion and final point 7800MTs vs 4800MTs is 10% faster (in synthetic benchmarks, the real-world application can be even lower) so on 1 hour-long video render you get a whopping 6 min shorter wait!!!
And the price difference between those modules isn`t exactly 6min / 10% compatible :P
Sadly entire IT industry has turned to Quick Money Grab...thankfully they usually put RGB lights on top of RAM sticks, I guess it`s to make you feel less sad while your butt is still hurting (for paying premium) !
Please read more comments... His results here are inaccurate. The faster RAM Kits should have higher scores but they don't because it was unstable. It's the same BS like jays2cents. This is overclocking and he is uneducated in this regards. It's the typical normie buying high-end hardware and not understanding the results here. And the manufacturers don't highlight this possible problems but they do communicate it with extra ** notes etc.
As usual, rtfm
@@BarisYener "This is overclocking and he is uneducated in this regards. It's the typical normie buying high-end hardware and not understanding the results here"
I love it when people consider entering bios and turning on the XMP profile "overclocking", btw I thought that the entire point of buying high-end hardware was to see the difference in performance (aka get better performance) but if you say that understanding the results of synthetic benchmarks is the priority...then I will throw away my 24 year experience of building PCs and buy the book "Synthetic Benchmark Results for Dummies - how to understand something you don`t see" !
@@alexdimitri9669 I'm not sure if you intentionally misunderstand me or if I was not clear enough. His results are flawed since the scores he get in shown puget benchmarks (which are NOT synthetic btw.) are lower then expected. This too low difference and sometimes even worse performance scores with the higher end XMP indicate that the RAM was runing unstable and that the on-die ECC correction brought him through those tests (which again, are NOT synthetic) but with the cost of performance.
And it's also a phenomenon that people who build a couple of PCs in their live think they are experts. And from what you write I can clearly assume that you have no experience in overclocking and what terms you need to watch out for. Yes manufacturers are not clear enough in communicating some of this stuff, but they do and people who are PROs know about this stuff. You clearly don't. Typical Dunning-Kruger-Effect.
@@BarisYener Sorry, I am not a PRO in useless discipline, call me a caveman I just build computers that work (as intended).
@@alexdimitri9669 "To get the conclusion and final point 7800MTs vs 4800MTs is 10% faster (in synthetic benchmarks, the real-world application can be even lower) so on 1 hour-long video render you get a whopping 6 min shorter wait!!! "
It's not useless to get free performance. The Module prices are very close for example fast DDR5-5600 is not much different to DDR-6400. You can buy Hynix A-DIe and tune it to even higher MT/s and get free performance.
This rubbish you write here is just the problem on the internet. If you tune your DDR5 right, there can be up to 15% - 20% performance gain in real world szenarios. Maybe even more. DDR5 is a complicated technology since you need 3 parts that can run the higher MTs + the knowledge to tune it.
Puget Benchmarks are real world tests.
PS: I'm not wasting more of my time here in the comments. Go do you! Bye.
u are to late