Starting to see 5200 and 5600 CL28 28-34-34-89 Kits wondering if the lower CL and decent speed would be good as you no longer will get that latency issue with DDR5
Good video. I always wanted to see differences in a system where only ram speed is the difference. Start at 4800, 5600, 6000,6400,6800,7200,7400 strictly for gaming
Would love to see the 6000 expo kits vs faster xmp kits like 6400 or 6600 with looser timings to see what kind of impact the additional options on expo actually bring.
Well the thing with looser timings is most of the time the are irrelevant. You can get 6000MHz Trident Z5 Neo with CL30, 32 and 36. The only difference is CL=Cas and this are cycles and not a time how long each cycle takes is not defined all timings behind are the same between this rams but thwy have a cost difference of about 80 to 100 dollar from 30 to 36. So just a scame nothing else.
Those are some very inconsistent results you got. Some make no sense, especially in the gaming tests. In some cases the stronger CPU was losing performance with the faster ram, while the weaker one was gaining performance from the faster ram. You should probably check your settings and probably system stability (maybe background processes, OS settings). Also, you didn't list the specs of the ram kits (latency, etc.). Showing them on screen for a couple of seconds would have been helpful and informative. On the Cost Per Frame chart, the bars should be presented primarily to reflect the average framerates with the cost per frame as secondary scale (kinda like Hardware Unboxed do it). That way it paints a more complete/precise picture of the relation between performance and cost. Your presentation would always put the most expensive thing at the top, which is not very useful information.
You really need more data points per CPU to say whether something is a sweet spot or not. You especially need to compare 6000 vs something higher to see if the returns are entirely diminished for that particular CPU. There's nothing at all wrong with testing with 5200 MT/s DDR5, since that's what AMD itself has said is the maximum memory speed officially supported by these CPUs. I think it'd be worth comparing 3600 MT/s as well since that's the officially support max speed for a max capacity configuration.
Given that the 6000MHz kit is CL30 vs CL36 on the 5200MHz with the IF still topping out at 2000MHz on Zen 4, how much of the performance gain for the 6000MHz kit is actually from the lower latency of the kit and not the higher bandiwdth? I would test a 5600MHz CL28 kit vs a 6600MHz CL32-34 one to see if Zen 4 benefits more from lower latency or higher bandwidth.
6000/5200 is about 15% faster, all else being equal. Of course modern processors spend most of their time in ultra-fast L1, L2, and L3 caches, so faster main memory makes *little* *difference* .
probably the cheaper one as, since I would run ecomode and, undervolt and reduce the max clock the GPU , I wont be hitting the max performance, so wont be needed. Why? Electric and Gas prices here in UK gonna rip me a new one.
I'm honestly more interested on how much performance will suffer if you populate all DIMM slots with 4 sticks of RAM (like x4 5200MHz VS x4 6000MHz). I have heard there will be performance reductions but have never heard of how large it will be. That's something that's more important to me, since like many folks, I want to fill in all slots. Will 6000MHz being the higher clocked RAM go down BUT still stay above 5200MHz? Or will 5200MHz go down even lower, keeping the battle have the same results as this video? Maybe both will just match? That's what I want to know.
Nice to see such little difference even with pretty substantial difference in not only speed but timings as well. Ended up getting 5600 28-34-34-89 for 10 bucks less than the 6000 kit you tested but with slightly better latency so there shouldn't be much difference.
just to check, is it a CPU limitation on the speeds of the DDR5 rams and inability to go beyond 3600mhz on 4 sticks of DIMMs? Or is it something that can be resolved with future AMD BIOs driver updates? Because the whole idea and point of upgrading to AM5 platform is for future proofing, and if there is a need to grab a newer AM5 motherboard and newer CPU to support 4 sticks of DIMMs and higher RAM speeds then it becomes pointless to upgrade.
This is the same thing all over again where all of the TH-camrs tested Ryzen 2000, 3000 and 5000 with 3600MHz DDR4 and in the end you might get like 1-2% gain with faster than 3600MHz RAM. So this will be the same just use 6000 MHz DDR5 and stop thinking about it.
I really don't see the point of canned benchmarks and 1080p gaming on such high-end systems. I've looked at several videos testing different memory configurations(e.g. Dual rank, speed etc.) and the common thread is that any performance gain will likely happen in extremely cpu bound scenarios such as 1080p gaming. And these gains, if any, get wiped out at higher resolutions. So, my question is: how representative is 1080p gaming data of other cpu intensitve use cases, such as 4k video editing? The same goes for the synthetic benchmarks. I'm pretty sure when AMD recommended 6000 memory as the sweet spot, gaming and canned benchmarks weren't the only use cases they were considering.
1080p is a stress of the CPU, ensuring GPU is not bottlenecking. Usually, 4k is for GPU testing, as resolution scales with GPU usually more so then CPU.
The main benefit of latency would be on expo ram as it tunes the timings for you. Also it’s kinda weird to see so many benchmarks from so many reviewers that don’t use an AMD gpu. Would be more beneficial to at least include a 6900xt comparison in the benchmarks as well, as SAM might have a benefit with faster infinity fabric speeds.
they are all paid by nvidia to use rtx gpus... the excuse of youtubers is always but it's the fastest gpu on the market... forget youtubers and hardware....
Are we? NVIDIA haven’t paid me a penny. They don’t even send us founders edition cards. We use NVIDIA cards for certain tests and AMD cards for other tests to be as fair as possible. No money has been put in our hands though.
What is Zet? Are there any videos out there that do similar testing on the affect of memory for other than gaming performance, such as for photo/video editing etc.
I currently have the ,,I NEED THE TECHNICALLY BETTER COMPONENT'' illness and I love to come to these videos to just calm me down. I have 2 sticks of 5200mhz for a ryzne 9 7950x for 4K gaming and I'm glad to see that it really doesn't matter if my ram was 800mhz faster...
The AMD motherboards all have been getting BIOS updates since launch to improve memory stability and compatibility. I think results would definitely be different if tested today
I'm wondering what the ram speed difference does for the iGPU part of the Ryzen 7000 series. I know the integrated graphics aren't meant for gaming strictly speaking, but in lieu of an actual RDNA 2/3 "G" APU series anytime soon, I might consider this a stopgap.
Hi, I'm not an expert and in my country they don't talk about it. Do you recommend 6000 mhz expo (I ordered them and could cancel the order) or 6400 mhz without expo for a ryzen 7900x?
Just to check, the 6000Mhz kits for AMD are EXPO RAMs right? Because they did mentioned it's capped at 5200Mhz 1DPC and drops to 3600Mhz on 2DPC. And only specific AMD Expo ones can be overclocked above this 5200Mhz limit on 1DPC.
The memory latency matters on Ryzen. Not the speed. You want CL30 6000 is the sweet spot. If your slower RAM had lower latency it's gonna perform better than higher latency higher clocked.
5:32 Why DDR4 6000 and DDR4 5200 have similar READ bandwith for Ryzen 5 7600x (1 CPU die with separate IMC on IO die) .. also why better bandwith for DDR4 6000 using Ryzen 9 7900x ( 2 CPU die with separate IMC on I/O die) ?
Using corsair in expo mode on my asus b650e board with ryzen 7900. After the last Asus incident, should i disable expo mode in bios? I did not update bi
It would be interesting to see what the results would be if both ram kits where the same cas latency. If you want people to value your time and effort this needs to be done.
Can either of these platforms (XMP or XPO) use four DIMMS? I recently built a 12900K system with G-skill DDR 5 5600 memory but stuck with two DIMMS for fear of instability under XMP.
That's a limit of the memory controller. You could possibly get around it with tweaks to the memory settings, but I had to revert to 4800MHz on a system I built with an i9 12900K and 4 sticks of RAM. I wasn't willing to go beyond 1.3V or whatever it was to get XMP speeds. The system was stable, but Blender kept crashing and I'd get errors on memory tests. 13th gen has a better memory controller.
Definitely and it’s something we’ll revisit when kits that are significantly better come out. I’ll update the description with the memory kits used with their timings shortly.
I have ryzen 7 7700 non x and rog strix b650e-f motherboard got a set of 32gb 5600cl28 gskill and it will go to 5800 but not 6000 or gets blue screen so im glad i didnt but 6000 or i have lower it and cause my cl is 28 and not 30 little bit tighter.but idk how to adjust more out of it ?
Ive bought just the 5600 mhz for a r5 7600. It works fine. I'm happy. Im not into overclocking. Pc is fast as hell as it is. Btw. I got all on default but the CPU does jump to 5.4 ghz according to cpu-z and the taskmanager. Thats also the reason I didn't got a 7600x.
I'm not sure about the blender benchmarks really, memory speed is much more than just testing a render. It's how much faster it can scatter objects, remesh or dyntopo, add modifiers etc. Putting together a system with a 7950x with a 6k kit and blender is a huge part of my high intensity use, Ill try to monitor ram usage and throughput to see how much I can saturate teh chips
More like the new 3000 in my opinion. 3600 only became the best value in recent years near the end of DDR4. For many years 3000 or 3200 was the go to option for both Intel and AMD. The DDR5 equivalent of DDR4 3600-3733 will probably only come by the time DDR6 is getting close to debut but a few years from now the very least.
@@Eusebiugh 8400 is the highest "normal" memory you can have. Though not plentiful, if I look in the national pricewatch there are over a dozen kits listed out of 500+ for DDR5. Costly yes but there if you want it.
You can go higher than 5200, but if you do so every time you boot the pc it will be slower because of memory training. For example I have 6000 cl38 DDR5 but I use them at 5200 cl32 to keep the boot time faster. But if you keep your PC always on it's not an issue to go higher than 5200.
Hi, I have upgraded my rams from 5200 cl40 2x16 to 6000mhz cl 32 2x16. But after change my boot time pretty increased. My cpu is 7700x and my mobo is asus rog 670x-a. Do you have any idea about reason? Thanks.
AMD cannot run 7200 at all, it has difficulty running 6400 stable and there is no benefit in speed past 6000MHz as above that the half divider on fclk kicks in and robs performance. Regardless it's still very fast, especially at cl30.
You call the video : Ryzen 7000 series - Is faster memory better. Then why do you insert Intel data ? Like you said in the intro with Intel faster is just better. Then you show they make 7200 memory but you only test the 6000 ? I would love to know what the limits are for memory speed on the AM5 platform. A few tries overclocking that 6000 memory could be interesting and maybe also answer if speed or other settings influence performance most. You are one of the few channels though that I found searching best memory for AM5 that actually tested anything. Most others are rehashes of the AMD presentation or straight up commercials for memory strung together so kudos.
Depends on how much do you need and what's the main use case of your PC. For gaming it's no brainer as 6000 is similar in cost with 5200 and 5600 fo 32GB(and you don't need much for gaming only). So it's like almost free maximizing performance even if by a little. For the rest it's money on the wind when you but 96GB and the difference between 5200 and 5600 is already $70 and it goes a lot more for the 6000 while the difference in the productivity programs that eat all this RAM is negligible at best. Also i've seen on youtube 4k gaming tests with different memory sizes at the same speed and 1% and 0.1% lows are almost twice bigger with 64GB and 96GB comparing with 32GB. But i don't think gamers care that much about 4k gaming or for those little spikes as overall performance won't increase.
For the price, stick with DDR4 for now. 4000 or like 4133 tuned to like CL14 if you can manage it, is faster. Just can't justify DDR5 pricing rn (you'd need a new mobo too). 4000CL16 b-die Vipers for like $160 a kit is a great deal; grabbed myself two of those and off to the races
The advantage of sticking to ONE memory standard is that you avoid scenarios where some consumers are later confused when the DDR5 they bought doesn't function in the board they have because it was DDR4 and they didn't fully understand the choice they were making at the time. Of course there's also the scenario where the consumer knew what they were getting into but still regrets going DDR4 because now DDR5 is finally appealing and cost effective but now they need to buy a whole new board to be able to use it. AMD and Intel weigh this differently because they weigh platform longetivity differently. AMD emphasizes longetivity to benefit consumers most -- and thus considers the scenarios consumers might have to make years down the line -- while Intel emphasizes iteration to generally benefit their partners more. Pros and cons both ways. :T
6000mhz /CL30 ram is basically the same price in mid 2024. No reason to buy cheaper ram. $105 now for Gskill FlareX 6000MT CL30. $110 for the taller Trident Z5 versions. Why save $10-20 on some budget ram. 4 out of 5 AM5 builds have been 6000mt/cl30 with the last being cl32 6400mt that I bought new for $50. I run it at the same spec as the cl30 6000mt though.
I was holding out for AM5 components to drop in price, which has worked out well. Jun/Jul 2024, I am now in the process of sorting out my AM5 build (previously on AM4).
And poof my comment is gone as usual. I said 6000mt cl30 is basically the same price now as the cheaper ram so why bother on higher latency ram at this point. Edit: my comment returned!
Faster ram equals slightly better performance on average. As it should be. Data does not support any "sweet spot" nonsense theory. If you have fewer cores as in 7600x and on a budget just get the 5200 ram and don't overthink anything.
Now try to find ddr4 8000 MHz because that is exactly what my board can go up to finding the ram that's going to have to take a special s*** ton of money to make from Corsair for 2 sticks of ddr4 8000 MHz each with 32GB ram
We need to know the full specs of these ram kits otherwise this data is useless, we do not know the cl timings, we don't know the capacity size if they're dual rank or not and other confounding variables
Zen 4 will be a great upgrade towards the end of next year. AM4 is still a much better value platform. Unless you really need the productivity gains, it's just not worth the price.
I watched throught his just to see a chart comparison of video rendering performance only to be disappointed. It should of been the first statistic reviewed, because it's the one thing you are literally waiting 5 minutes on your computer to do. 1 or 2fps improvements in games are irrelevant. Shaving off a minute from your render time is a big deal. You would think a TH-cam contnt creator would be similiarly aware of how much the video rendering statistic is, and not.. gaming performance. So my tip for future videos, video rendering performance, video rendering performance, video rendering performance... first in your charts. Take any one of your video projects as a real world test, render it out on each machine, and post thetimes.
Ive ordered a 13700k (z690) with ddr5 5200 venegance. The extra More money for the ddr5 6000 ive spend in a atx 3.0 power supply for upgrading later to a rtx 4xxx 😂
Memory kits used:
Corsair Dominator Platinum DDR5 RGB 5200MHz (2x16GB) 36-38-38-74 1.25v
G.Skill Trident Z5 Neo 6000MHz (2x16GB) 30-38-38-96 1.35v
Please do 6000MHz vs faster RAM next? I'm curious to see how much faster you need to go to offset the "sweet spot".
I cannot get the 6000mhz to boot with the z5 neo kit. Currently running 5200
@@Dad_Kwhat motherboard do you have?
I went with 5600mhz with my 7600x..i think i saved money instead of going for 6000mhz..not worth it at all in my opinion.
Starting to see 5200 and 5600 CL28 28-34-34-89 Kits wondering if the lower CL and decent speed would be good as you no longer will get that latency issue with DDR5
Good video.
I always wanted to see differences in a system where only ram speed is the difference.
Start at 4800, 5600, 6000,6400,6800,7200,7400 strictly for gaming
do you still have the dead RTX 3090? I'd like to take a look at what killed it.
So we think we know what it was and happy to discuss further with you on Discord (discord.gg/eTeknix) to confirm.
👀👀
7800X3D with cheap RAM for the win.
12700k @ 5.2 ghz with cheap vipers ddr4 4000 cl15*
@@weyo14 that would still be slower
Yes, a cpu that doesn't exist ftw. wut?
@@insertnamehere4419 this CPU will be released in 2023 first half.
@@Arx07 Yes, I'm aware of the roadmap thank you.
Would love to see the 6000 expo kits vs faster xmp kits like 6400 or 6600 with looser timings to see what kind of impact the additional options on expo actually bring.
Well the thing with looser timings is most of the time the are irrelevant.
You can get 6000MHz Trident Z5 Neo with CL30, 32 and 36. The only difference is CL=Cas and this are cycles and not a time how long each cycle takes is not defined all timings behind are the same between this rams but thwy have a cost difference of about 80 to 100 dollar from 30 to 36. So just a scame nothing else.
Those are some very inconsistent results you got. Some make no sense, especially in the gaming tests. In some cases the stronger CPU was losing performance with the faster ram, while the weaker one was gaining performance from the faster ram. You should probably check your settings and probably system stability (maybe background processes, OS settings).
Also, you didn't list the specs of the ram kits (latency, etc.). Showing them on screen for a couple of seconds would have been helpful and informative.
On the Cost Per Frame chart, the bars should be presented primarily to reflect the average framerates with the cost per frame as secondary scale (kinda like Hardware Unboxed do it). That way it paints a more complete/precise picture of the relation between performance and cost. Your presentation would always put the most expensive thing at the top, which is not very useful information.
here we go, why can't you guys accept ram speed really doesn't matter that much?
Very detailed test. Impressive. Thank you & commenting for the algorithm.
Great work. Extensive testing. Really appreciate it! Got you a new subscriber
Thanks for the sub!
You really need more data points per CPU to say whether something is a sweet spot or not. You especially need to compare 6000 vs something higher to see if the returns are entirely diminished for that particular CPU. There's nothing at all wrong with testing with 5200 MT/s DDR5, since that's what AMD itself has said is the maximum memory speed officially supported by these CPUs. I think it'd be worth comparing 3600 MT/s as well since that's the officially support max speed for a max capacity configuration.
Given that the 6000MHz kit is CL30 vs CL36 on the 5200MHz with the IF still topping out at 2000MHz on Zen 4, how much of the performance gain for the 6000MHz kit is actually from the lower latency of the kit and not the higher bandiwdth? I would test a 5600MHz CL28 kit vs a 6600MHz CL32-34 one to see if Zen 4 benefits more from lower latency or higher bandwidth.
We’ll be testing using the 6000MHz kit at 6000mhz all the way down to 5200mhz to see how the scaling is with the same timings. It’s on our to do list.
6000/5200 is about 15% faster, all else being equal.
Of course modern processors spend most of their time in ultra-fast L1, L2, and L3 caches, so faster main memory makes *little* *difference* .
The juice depends on the squeeze applied. (2:33)
Amazing content. I was really looking for a video like this.
I was surprised by the small gains for 800MT/s. Thankfully the RAM I bought (6000) was only £7 more than the 5200 speed memory :)
What speed memory would you pair with a Ryzen 7000 CPU?
Whatever gives the best performance in the size I want/can afford.
probably the cheaper one as, since I would run ecomode and, undervolt and reduce the max clock the GPU , I wont be hitting the max performance, so wont be needed.
Why? Electric and Gas prices here in UK gonna rip me a new one.
I'm honestly more interested on how much performance will suffer if you populate all DIMM slots with 4 sticks of RAM (like x4 5200MHz VS x4 6000MHz). I have heard there will be performance reductions but have never heard of how large it will be. That's something that's more important to me, since like many folks, I want to fill in all slots. Will 6000MHz being the higher clocked RAM go down BUT still stay above 5200MHz? Or will 5200MHz go down even lower, keeping the battle have the same results as this video? Maybe both will just match? That's what I want to know.
Nice to see such little difference even with pretty substantial difference in not only speed but timings as well. Ended up getting 5600 28-34-34-89 for 10 bucks less than the 6000 kit you tested but with slightly better latency so there shouldn't be much difference.
One thing to remember is that higher speed memory also runs on higher voltage which will increase power use, especially on idle.
bro its just ram, how much power could it draw
@@Sartheris Enough to make a difference on idle.
just to check, is it a CPU limitation on the speeds of the DDR5 rams and inability to go beyond 3600mhz on 4 sticks of DIMMs?
Or is it something that can be resolved with future AMD BIOs driver updates?
Because the whole idea and point of upgrading to AM5 platform is for future proofing, and if there is a need to grab a newer AM5 motherboard and newer CPU to support 4 sticks of DIMMs and higher RAM speeds then it becomes pointless to upgrade.
This is the same thing all over again where all of the TH-camrs tested Ryzen 2000, 3000 and 5000 with 3600MHz DDR4 and in the end you might get like 1-2% gain with faster than 3600MHz RAM. So this will be the same just use 6000 MHz DDR5 and stop thinking about it.
Wish I could have given you two thumbs up, one for the video itself, and one for the hilarious monitor ad!
0:50 Sponsor killed me softly
- Auntie, what are you watching?
- It's uh, it's not what you think.
- wow, it's so big 😂
Amazin explanation. Subbed!
Seems like 5200mhz would be perfectly fine for ryzen 7000's
Why are AMD bandwidth number so low on the 7900x compared to intel? 🤔I understand on the 7600x because of 1CCD, but...
beacause of the chiplet design
Thank you for your time.
Can you test 6400 on Ryzen???
I really don't see the point of canned benchmarks and 1080p gaming on such high-end systems. I've looked at several videos testing different memory configurations(e.g. Dual rank, speed etc.) and the common thread is that any performance gain will likely happen in extremely cpu bound scenarios such as 1080p gaming. And these gains, if any, get wiped out at higher resolutions.
So, my question is: how representative is 1080p gaming data of other cpu intensitve use cases, such as 4k video editing? The same goes for the synthetic benchmarks. I'm pretty sure when AMD recommended 6000 memory as the sweet spot, gaming and canned benchmarks weren't the only use cases they were considering.
1080p is a stress of the CPU, ensuring GPU is not bottlenecking. Usually, 4k is for GPU testing, as resolution scales with GPU usually more so then CPU.
The main benefit of latency would be on expo ram as it tunes the timings for you. Also it’s kinda weird to see so many benchmarks from so many reviewers that don’t use an AMD gpu. Would be more beneficial to at least include a 6900xt comparison in the benchmarks as well, as SAM might have a benefit with faster infinity fabric speeds.
they are all paid by nvidia to use rtx gpus... the excuse of youtubers is always but it's the fastest gpu on the market... forget youtubers and hardware....
Are we? NVIDIA haven’t paid me a penny. They don’t even send us founders edition cards. We use NVIDIA cards for certain tests and AMD cards for other tests to be as fair as possible. No money has been put in our hands though.
@@suadmemo2412 Then report them to the FTC if that's the case, since they're breaking the law by not declaring that financial interest.
What is Zet?
Are there any videos out there that do similar testing on the affect of memory for other than gaming performance, such as for photo/video editing etc.
I currently have the ,,I NEED THE TECHNICALLY BETTER COMPONENT'' illness and I love to come to these videos to just calm me down. I have 2 sticks of 5200mhz for a ryzne 9 7950x for 4K gaming and I'm glad to see that it really doesn't matter if my ram was 800mhz faster...
The AMD motherboards all have been getting BIOS updates since launch to improve memory stability and compatibility. I think results would definitely be different if tested today
i did update bios and from 6400/32 expo stable i got nice downgrade to 6000/30 ... stupid
I'm wondering what the ram speed difference does for the iGPU part of the Ryzen 7000 series. I know the integrated graphics aren't meant for gaming strictly speaking, but in lieu of an actual RDNA 2/3 "G" APU series anytime soon, I might consider this a stopgap.
Why not use a 6950XT?
Check the DDR5 vs. DDR4 Benchmarks from gamers nexus, he's gets into why using AMD is not the best methodology for DDR5.
@@vherostar The only D4 vs D5 video that GN has is nearly a year old, so not relevant now with Ryzen 7000.
@@vherostar Also the 6950XT is a GPU...
what resolution are these games?
Great analysis Andy!
18:49 Q: what is holding up those G.Skill DDR5 memory modules …… enquiring minds want to know 😅👍🤣.
Amazing video
Thanks a lot
Hi, I'm not an expert and in my country they don't talk about it. Do you recommend 6000 mhz expo (I ordered them and could cancel the order) or 6400 mhz without expo for a ryzen 7900x?
Just to check, the 6000Mhz kits for AMD are EXPO RAMs right?
Because they did mentioned it's capped at 5200Mhz 1DPC and drops to 3600Mhz on 2DPC.
And only specific AMD Expo ones can be overclocked above this 5200Mhz limit on 1DPC.
Whaaat
@@Hasmatix you can check official website for Ryzen 700. It's stated in spec sheets
Great vid! Many people are thinking about saving some money, when buying AM5, and some might think about getting slower&cheaper RAM.
The memory latency matters on Ryzen. Not the speed. You want CL30 6000 is the sweet spot. If your slower RAM had lower latency it's gonna perform better than higher latency higher clocked.
5:32 Why DDR4 6000 and DDR4 5200 have similar READ bandwith for Ryzen 5 7600x (1 CPU die with separate IMC on IO die) .. also why better bandwith for DDR4 6000 using Ryzen 9 7900x ( 2 CPU die with separate IMC on I/O die) ?
Using corsair in expo mode on my asus b650e board with ryzen 7900. After the last Asus incident, should i disable expo mode in bios? I did not update bi
It would be interesting to see what the results would be if both ram kits where the same cas latency. If you want people to value your time and effort this needs to be done.
to know if it's truly the sweetspot you'd need to test 5200/6000/6400/7000?
Can either of these platforms (XMP or XPO) use four DIMMS? I recently built a 12900K system with G-skill DDR 5 5600 memory but stuck with two DIMMS for fear of instability under XMP.
That's a limit of the memory controller. You could possibly get around it with tweaks to the memory settings, but I had to revert to 4800MHz on a system I built with an i9 12900K and 4 sticks of RAM. I wasn't willing to go beyond 1.3V or whatever it was to get XMP speeds. The system was stable, but Blender kept crashing and I'd get errors on memory tests. 13th gen has a better memory controller.
@@lordofdrones Thanks. Very helpful. 🙂
What about 4 dimms vs 2 dimms for 7000?
Um this is half the story! CAS latency cannot be ignored
Definitely and it’s something we’ll revisit when kits that are significantly better come out. I’ll update the description with the memory kits used with their timings shortly.
Description updated.
This is exactly what I was looking for
I have ryzen 7 7700 non x and rog strix b650e-f motherboard got a set of 32gb 5600cl28 gskill and it will go to 5800 but not 6000 or gets blue screen so im glad i didnt but 6000 or i have lower it and cause my cl is 28 and not 30 little bit tighter.but idk how to adjust more out of it ?
Ive bought just the 5600 mhz for a r5 7600. It works fine. I'm happy. Im not into overclocking. Pc is fast as hell as it is.
Btw. I got all on default but the CPU does jump to 5.4 ghz according to cpu-z and the taskmanager. Thats also the reason I didn't got a 7600x.
I'm not sure about the blender benchmarks really, memory speed is much more than just testing a render. It's how much faster it can scatter objects, remesh or dyntopo, add modifiers etc.
Putting together a system with a 7950x with a 6k kit and blender is a huge part of my high intensity use, Ill try to monitor ram usage and throughput to see how much I can saturate teh chips
apparently on amd web says ryzen 7000 run max 5200mhz, would i have any problem if i buy ddr5 6000mhz on my 7700x
Seems like DDR5 5600 will be the new 3600 DDR4
Then what's the 4400 MHz? I wonder I wonder.
DDR4 4400.
Timings will play a big factor over time I believe instead of raw bandwidth from speed.
More like the new 3000 in my opinion. 3600 only became the best value in recent years near the end of DDR4. For many years 3000 or 3200 was the go to option for both Intel and AMD. The DDR5 equivalent of DDR4 3600-3733 will probably only come by the time DDR6 is getting close to debut but a few years from now the very least.
6000 memory at CL30, versus 5200 memory at CL36.
And still, the performance boost is 5% or less.
6000 is definitely NOT worth the money.
Retest with new AGESA for DDR5-8400. Wonder if the 1:2IF hampers or helps.
Didnt see anyone with 8400. Maybe 8000 and on a single ccd 7800x3d i think 6600 flck 2200 1:1 is best.
@@Eusebiugh 8400 is the highest "normal" memory you can have. Though not plentiful, if I look in the national pricewatch there are over a dozen kits listed out of 500+ for DDR5. Costly yes but there if you want it.
@@jamegumb7298 didnt see anyone with at least boot ar 8400 but working stable.
if 6000 mhz is faster, then why does amd say max speed is 5200?
You can go higher than 5200, but if you do so every time you boot the pc it will be slower because of memory training. For example I have 6000 cl38 DDR5 but I use them at 5200 cl32 to keep the boot time faster. But if you keep your PC always on it's not an issue to go higher than 5200.
@@agostinog do you feel it's a waste of money for you since u got 6000, but running at 5200? But I guess u still benefit from the lower CL?
@@Mewzyc yes this is the deal, and between 6000 cl38 and 5200 cl2 the difference is very minimal anyway
Hi, I have upgraded my rams from 5200 cl40 2x16 to 6000mhz cl 32 2x16. But after change my boot time pretty increased. My cpu is 7700x and my mobo is asus rog 670x-a. Do you have any idea about reason? Thanks.
My 7950x crashes when I set the ram 5400 ram to expo. Maybe I have a bad chip.
What ram? Seems these processors struggle above 6000mhz at the moment.
@@eTeknix I fixed it. I needed to update my chipset drivers from AMD website opposed to the MB website.
I would like to see how 7200 memory performs on both AMD and Intel
yeah 7000+ pls
@@iikatinggangsengii2471 I am using Corsair 5600 CAS 36 EXPO, so yes I would be interested in seeing 7200.
AMD cannot run 7200 at all, it has difficulty running 6400 stable and there is no benefit in speed past 6000MHz as above that the half divider on fclk kicks in and robs performance. Regardless it's still very fast, especially at cl30.
You call the video : Ryzen 7000 series - Is faster memory better. Then why do you insert Intel data ? Like you said in the intro with Intel faster is just better. Then you show they make 7200 memory but you only test the 6000 ? I would love to know what the limits are for memory speed on the AM5 platform. A few tries overclocking that 6000 memory could be interesting and maybe also answer if speed or other settings influence performance most. You are one of the few channels though that I found searching best memory for AM5 that actually tested anything. Most others are rehashes of the AMD presentation or straight up commercials for memory strung together so kudos.
I felt let down when you didn't throw air quotes and do the slight zoom the last time you said sweet spot. 😂
What is also funny is that all of these performance numbers are just for your setup when people might get better performance with the same CPUs. 😆
😁😁Haha what a joker with the Intro. Thanks for the informative Video.
So max ram speed is 5200 for 7600x, how does it benefit from 6000 speed?
It runs 6000 ram at 6000 speed just fine, 5200 is not the max.
Please make a comparison between 6000MHz vs 7200+mhz on am5
Since the difference of the price is low now in 2024, I would pick the 6000mhz kit
Depends on how much do you need and what's the main use case of your PC. For gaming it's no brainer as 6000 is similar in cost with 5200 and 5600 fo 32GB(and you don't need much for gaming only). So it's like almost free maximizing performance even if by a little.
For the rest it's money on the wind when you but 96GB and the difference between 5200 and 5600 is already $70 and it goes a lot more for the 6000 while the difference in the productivity programs that eat all this RAM is negligible at best.
Also i've seen on youtube 4k gaming tests with different memory sizes at the same speed and 1% and 0.1% lows are almost twice bigger with 64GB and 96GB comparing with 32GB. But i don't think gamers care that much about 4k gaming or for those little spikes as overall performance won't increase.
Personally, i would only go for the higher speed memory if cheaper or if i already squeezed out all i could out of other hardware.
Doesn't Cas Latency matter?
Can I buy 5600 C36 MHz RAM and using it at 5200 C40?
I thought the new terminology was mega transfers?
What about 5600 MHz? This is the middle between 5200 and 6000 MHz and fast as cheap as 5200 MHz. And 5600 MHz RAM can also run with 5200.
For the price, stick with DDR4 for now. 4000 or like 4133 tuned to like CL14 if you can manage it, is faster. Just can't justify DDR5 pricing rn (you'd need a new mobo too). 4000CL16 b-die Vipers for like $160 a kit is a great deal; grabbed myself two of those and off to the races
You can't run DDR4 on Ryzen 7000
@@notlNSIGHT tell NewEgg that.
How come there's no 7000+ DDR5 here.
Shame amd didn't leave the option of ddr4 because ddr4 can get in the 40's in latency
The advantage of sticking to ONE memory standard is that you avoid scenarios where some consumers are later confused when the DDR5 they bought doesn't function in the board they have because it was DDR4 and they didn't fully understand the choice they were making at the time. Of course there's also the scenario where the consumer knew what they were getting into but still regrets going DDR4 because now DDR5 is finally appealing and cost effective but now they need to buy a whole new board to be able to use it.
AMD and Intel weigh this differently because they weigh platform longetivity differently. AMD emphasizes longetivity to benefit consumers most -- and thus considers the scenarios consumers might have to make years down the line -- while Intel emphasizes iteration to generally benefit their partners more. Pros and cons both ways. :T
by this 7700x with ddr5 6000mhz should be the sweet spot
6000mhz /CL30 ram is basically the same price in mid 2024. No reason to buy cheaper ram. $105 now for Gskill FlareX 6000MT CL30. $110 for the taller Trident Z5 versions. Why save $10-20 on some budget ram.
4 out of 5 AM5 builds have been 6000mt/cl30 with the last being cl32 6400mt that I bought new for $50. I run it at the same spec as the cl30 6000mt though.
I was holding out for AM5 components to drop in price, which has worked out well.
Jun/Jul 2024, I am now in the process of sorting out my AM5 build (previously on AM4).
If we want big improvement over ddr4 3600 we needs 6600+ ddr5
I paused on every frame and you look handsome in all of them. NO HOMO♥️♥️
no 7200mhz??
And poof my comment is gone as usual. I said 6000mt cl30 is basically the same price now as the cheaper ram so why bother on higher latency ram at this point.
Edit: my comment returned!
Faster ram equals slightly better performance on average. As it should be. Data does not support any "sweet spot" nonsense theory. If you have fewer cores as in 7600x and on a budget just get the 5200 ram and don't overthink anything.
Thank you for making me feel worse about my decision to go with the 6000 MHz ram. lol
Why feel bad? Anything less is leaving performance on the table.
AIDA64 memory latency is NOT milliseconds, it's nanoseconds.
Otherwise, it would be 30x times slower than HDD...
My 7900x with ddr5 6000 showed higher results than almost every test done here regardless of CPU. In some cases 30% higher than the highest.
How recently did you test it cause this video is from a while ago and new hardware tends to always need updates
How the h did the 3090 die?!
Now try to find ddr4 8000 MHz because that is exactly what my board can go up to finding the ram that's going to have to take a special s*** ton of money to make from Corsair for 2 sticks of ddr4 8000 MHz each with 32GB ram
5600mhz seems to be the sweet spot
The solution is to OC 5600 Kingston to 6000, it's actually the only sensible way for now.
Can I use xmp ddr5 on am5
i put my 3200 cl14 ripjaws
2x 16)to 3600 cl14 and called it a day :) and saved alot of cash.
We need to know the full specs of these ram kits otherwise this data is useless, we do not know the cl timings, we don't know the capacity size if they're dual rank or not and other confounding variables
Added into description and pinned comment too.
Memory kits used:
Corsair Dominator Platinum DDR5 RGB 5200MHz (2x16GB) 36-38-38-74 1.25v
G.Skill Trident Z5 Neo 6000MHz (2x16GB) 30-38-38-96 1.35v
IT WOULD BE A GOOD IDEA TRY 5600MHZ
Zen 4 will be a great upgrade towards the end of next year. AM4 is still a much better value platform. Unless you really need the productivity gains, it's just not worth the price.
I watched throught his just to see a chart comparison of video rendering performance only to be disappointed.
It should of been the first statistic reviewed, because it's the one thing you are literally waiting 5 minutes on your computer to do. 1 or 2fps improvements in games are irrelevant. Shaving off a minute from your render time is a big deal. You would think a TH-cam contnt creator would be similiarly aware of how much the video rendering statistic is, and not.. gaming performance. So my tip for future videos, video rendering performance, video rendering performance, video rendering performance... first in your charts.
Take any one of your video projects as a real world test, render it out on each machine, and post thetimes.
Thanks
I didn’t know Alder lake could get that hight In latency 😂 mine is at 46 ns. Everyone should go watch frame chasers video on latency scaling
Ive ordered a 13700k (z690) with ddr5 5200 venegance. The extra More money for the ddr5 6000 ive spend in a atx 3.0 power supply for upgrading later to a rtx 4xxx 😂
5200 will gimp your 13700k, his benches are wrong you want 6600MHz +
I wish you included some esports titltes
Can't believe your wife helped you with that ad lol
Ha. She loved it :D