Seems something is wrong with windows. Linux tests over on phoronix conclude "the Ryzen 9 9950X came out to being 17.8% faster than the Ryzen 9 7950X. The Ryzen 9 9900X meanwhile was 21.5% faster than the Ryzen 9 7900X across this wide mix of workloads. The Ryzen 9 9950X was 33% faster than the Intel Core i9 14900K performance overall and even the Ryzen 9 9900X was 18% faster than the Core i9 14900K"
I think the scheduler is not great. I'd be curious to see win10 compared to 11. Why? AMDs chiplets are more like the classic sockets + northbridge than Intel's. I suspect that the optimisations in 11 for Intel's architecture are being too smart for their own good. If I'm correct, w10 will show better performance than w11.
Some reviewers claim running the benchmarks as Administrator in Windows or running them on Linux gives them about 5-10% boost in performance, could you make a video about that?
@@PineyJustice I think it is: Phoronix did a series with the 9600/9700/9900/9950, and they were way better than their Windows counterparts. I think the Windows scheduler is playing silly bonks again.
The problem is that Windows 10/11 require 64 GB to stop using memory compression for full windows operation. 48GB is not big enough. You get more speed from stopping memory compression than you do from the speed of the ram.
Not on Zen 5. But, for dual CCX/chiplets, Yes. It is too complicated of an architecture. It's like messenger. Once you start adding in more steps (people) to convey rhe information to the next step (person) then it just makes the information less reliable. By the time it reaches the other end, it's garbled and mixed up. Too many involved. Keep it simple. One chip. But they can only make up to 8 cores on one chip. So, if you want more than 8 cores, you lose potential. But those are for production and editing. Not truly a gaming chip.
@@IZZYX-X 6 cores is used for a good amount of time... better saying games are starting to use 8 cores more and more... 6 cores is the bare minimum nowadays.
CAS Latency (and other timings) have always been just as, if not more important, than memory speed. I used DDR4 3600 CL14 memory on my 5000 series procs and it nearly eliminated the stutter issues on higher CL16-18 kits that I tested before, along with a more responsive desktop.
Timing correlates with speed. Example: You can speak very fast on the phone, but you are slow in typing the phone number. In case your phone calls are just one minute each, it is not important how fast you speak to get a certain amount of work done. It's more important how fast you get into ur next call. That's the correlation between memory transfer speed and timings. The nature of random access makes the time required to do the next access very important. It's the main reason why Cache works so well. In first place it reduces dramatically the access time.
It’s never worth buying a cpu from 1 gen to the next. Especially high end ones. People get all gooogly eyes over these charts but forget that in real life you’re saving seconds. Even if you save a full minute, unless it’s enterprise it’s not that serious. Same with gaming. People with build an entire new machine for 15 more frames, from the same game they was already playing, and enjoying just fine…….until they saw a chart
You can manually set ratio back to 1:1 even after 6000. You can then run memory at 6400 and noticable difference in applications that are memory speed dependent. Also very tight timmings will help even more.
most motherborad need 1.35v-1.4v vddq and vddio on asrock board you need 1.4v vdd misc 1.1v-1.2v and mem vdd 1.35v-1.42v or higher if the motherborad doesnt bug out then you can hit 8000mhz stable if you have a good motherboard (300$+) and test the stability with memtest (more progarms are better)
BRO! I say this EVERY TIME you do a standing video.... you NEED TO fix the audio in that room! It sounds like you are in an airplane bathroom lol ... there's gotta be SOMETHING you can do that's off camera to muffle those echos.... When you're sitting and right next to the mic you don't hear it at all.... because you're right next to the mic.... So aside from spending a bunch of money on a new wireless/lapel mic that you can attach to your shirt collar, you would HAVE TO do some sound dampening.... blankets and cloth on large flat, hard surfaces is what will work off camera... - so aside from just NOT doing standing vids you COULD reach out to a company that makes lapel/wireless mics and see if you can do a sponsor spot 🤷♂ that might actually be your best bet :)
7700x with per core negative power curve, always at about 5.5Ghz, 6400MT ram 1:1 CL34 flawlessly stable. No issues especially since last 4/5 agesa updates.
Again this reminds me of the early days of DDR4 and x99. Trying the super fast stuff early on was a mess but BIOS updates, yeilds etc just got better and better and 18 months later faster stuff got stable. I never jump on a new gen and platform on day 1. Leave it a few months and let someone else take the pain.
i am a ram overclocker and i will say for games stay at 1:1 as high as your fabric goes. the penalty is real and for ddr4 you dont see a gain in games till 4800mhz from a 3800mhz and its 1%
Learn this from AM4 when I bought DDR4 4000MT/s. I won’t be migrating to Ryzen 9000, I’m sticking with my 7600X and DDR5 6000MT/s until there is a 50% gain.
I run Memtest first. Instability can be detected in just a few minutes. If it has errors, I won't even try the benchmarks. I haven't had a single crash on AM5 in a year and a half. I have a 6000 CL30 1.35V with lowered timings on my XPG Lancer 6000 CL40 + Asrock B650M PG. 6000 CL28 1.4V also survived the 6 hour test in Memtest. 6200 CL30 is also OK. 6400 CL32 1.4V 1:2 throws errors within 15 minutes. 6000 CL26 1.55V seems to be stable if the temperature is kept below 80°C.
7800x3d running at 6200 1:1. With one major caveat, the memory is at 1.55 volts. I am NOT recommending this for everyone. Temps are what degrade memory, and I have excellent airflow in my case. I monitor heavy gaming sessions with hardware info, and temps never reach 40° celsius on the memory.
There is no chance your memory chips are around 40° with that savage voltage. You are likely lying or must have a problem with the software you are using either. If they are at around 50/55° you arr lucky already.
This RUclips video of yours was historic. The first video that doesn't make our ears ring with plosives, i.e. you don't speak directly into the microphone. Thank you very much! Keep it up!
Memory stability also depends a lot from motherboard itself. Trace length/width, interference signal quality. Remember than current boards (6xx series) are release over a years ago and it's may not even tested with DDR5 as high as 7000. Although AEGSA supports it. CPU and Mainboard are not guarantee to work. See your mainboard specification that what they really test the board up to. It may be just DDR5-6400 or 7200 depending on manufacturer. For DDR5-8000 module that exists, they tested it yes but they don't tell you what CPU / Board combination they using. It must be very high quality though to possible to run that test.
Ryzen processors, which are normally targeted towards consumers, are definitely not with Zen 5. Steve from Hardware Unboxed pointed this out. Zen 5 is not aimed at consumers. Despite AMD showing gaming benchmarks for Zen 5, they were total BS. AMD clearly didn't design Zen 5 for consumers such as gamers. It seems much more focused on AI/enterprise workloads. Something that most consumers don't give a shit about.
You realize that gamers are their minority of consumer right? Like a quarter of their revenue is from gaming. The rest is all servers and ai processing but pop off bud
@@DingleBerryschnapps For general PC use cases such as gaming, 3D rendering, video editing, etc, Zen 5 really is not for that. For those use cases, you are much better off with Zen 4. Same performance at much lower prices. Zen 4 is faster sometimes.
Zen 5 beats zen 4 in data center benchmarks and tasks. Any desktop use case is equal or below Zen 4. Current Zen 5 chips are all built using TSMC 4nm tuned for efficiency. AMD can switch things around and use 4nm performance skew to release this CPU's at higher TDP. But likely they are just waiting on Intel's next "gaming champion" to burry it with Zen 5 X3D that they don't need to release before Intel shows its hand.
Have you tried the Hollyland mic? It’s fairly cheap and has great audio and amazingly easy at concealing on cam. Throwing on cheap acoustic foam and some carpet will help clear echo in a room as well. Audio > visuals
Good work man. It stuns me that reasonably priced RAM is the best for AMD CPUs. I myself tuned cheaper 6000MHz RAM from 36-36-36-96 to 30-34-34-76@ 1.4V with tREFI = 65565! I could not be happier. G-Skill Flair X5 EXPO only RAM with Asrock x670e Steel Legend motherboard. 🤘
I have a kit of CL32 6400MHz RAM that I ran at 6000MHz, 6200,and 6400MHz. I never noticed a difference between them. Timings and channel ranking is more important than frequency.
I don't understand then! DDR5-6400 Confirmed as Sweetspot Speed of Ryzen 9000 "Zen 5" Desktop Processors! As part of the update, Ryzen 9000 "Granite Ridge" should be able to run DDR5-6400 with a 1:1 ratio between the MCLK and FCLK domains. This is a slight increase from the DDR5-6000 sweetspot speed of Ryzen 7000 processors. /TechPowerup/. Infinity Fabric clock (FCLK) has been increased to 2400 MHz /Zen 5/ Now is that a lie? I planned to buy such ram for the 9900X... 6400mhz CL30 and 1:1 ratio! Or will this only be available with 800 series motherboards?
@@reiniermoreno1653 Yes, I want to upgrade to 9900X! So then I have to wait for the 800 motherboards, and then the 6400 can work 1:1? As far as I know, the 800 motherboards will be launched on September 30th. I didn't want 600 or 700 motherboards, only 800, which is my choice! I have a 3900X, I am developing from it, it will be a fairly decent jump, and if it really works with 6400 rams 1:1, I can get quite good performance with CL 30 timing. Otherwise, how can it work with 6400 1:1? So our Infinity fabric should be 3200mhz, or not?
But you can take an 8000mhz kit set it to 6000mhz for faster timings and divide its 8000mhz timings by the difference, so for example my 4000 mhz ddr4 kit had timings of 16-16-16-36CR2 @1.35V and the difference between that and 3200mhz i had to run at for my old cpu was the same as the difference between 6000 and 8000, so for example i divided my timings by 4 since and multiplied times 3 because the new frequency was 3/4 the old one so new timings were 12-12-12-27 and got it to 11-11-11-24CR1 @1.45V of its 1.5V max recommended by samsung for b-die at the time.
ryzen 7000 taught me that timings matter more and that stability requires more than just loading an expo or xmp profile. There's a reddit thread dedicated to stabalizing OC'd ram. I was having a hard time booting 128gb of ddr5 6000 on a 7950x3d. *2 64gb kits. It was only stable when I removed a kit.. but that thread saved me having to return ram or waiting for 9950x3d to be released.
I have a Ryzen 3700x paired with DDR4 3600mhz RAM. Now I think random shutdown and rest happening because of RAM speed. In my GSkill Ram there is only 1 profile i.e., 3600mhz So how do I set 3200mhz for stability? Or is 3600mhz really an issue? Please help me out 🙏🏽
Having fast ram doesnt mean its actually running at that speed. It doesn't default to the higher speeds in bios unless you set it so. Easy to check using CPUZ. Run a memtest at the current settings. Check Hard drive health ie the usual troubleshooting.
Makes sense, I have an old ryzen 3600, and I was running 3600mhz, i put in 4 sticks of 3600mhz, and it buckled completely. The ryzen mem controllers are not amazing, and they seem to have a hard time running high speeds it feels like. I hadnto reduce speeds to around 3333 mhz to get any kind of stabvility
It is also related to cl30 vs cl40+ and the rest of the sluggish timings over 6000mhz. Further, 6000 to 8000 bandwidth when it is useful is running a heck load of work or serving a lot of clients. Then the 8000 will be beneficial. The ideal is the Trident.Z 6400mhz @ cl32 2x32gb which is still very cheap. Im running mine at 6000 cl30 and at 1.38v instead of 1.4
instead I got an epyc 9124. I have 12 memory channels at 4800. So in addition to having more total memory bandwidth, I can hold more ram. The motherboards were about the same price but the cpu was +$500.
The new chipsets on the the (upcoming) motherboards isn't going to help. Memory accesses don't go through the chipset -- it's strictly CPU-to-RAM. If there are AEGISA updates to help with this, those can be incorporated into the existing mobo's as easily as the new ones.
Any news on whether 128gb is stable? I never upgraded my 64gb on 7950x due to issues, I need 128gb absolutely minimum to work before considering any changes
im running a 7800x3d bundle from microcenter it came with 6000mhz ram and it runs at that speed great. ill try to get it to run at 6400mhz and see what happens.
I've been running 6400 cl 30 on my 2 7800x3d systems and 7900X since the launch of them. I am using the Asus ROG x670E gene board that is a 2 dimm board.
Best performance with 1-to-1 DDR5 6000CL28, possibly G.Skill Trident Z5. If your CPU memory controller is good enough, you can try overclocking the up to 6400 (still 1-to-1 ratio)
hey meld could it be the crashes are from MS GAME BAR ISSUE IN WINDOWS... BECAUSE THAT IS EXACTLY WHAT HAPPENED WITH THE 7800x3d and 7950x3d you need GAME BAR ENABLED to make them work the same is with the 9900 and 9950x and am sure the X3d chips will have the same issue ... THIS IS MS WINDOWS ISSUE NOT AN AMD ISSUE OK GET THAT CORRECT...
You hyped Zen 5 so much, why didn't you review it against all the current CPU's for us all to see. Much more interesting than seeing it with different ram speeds?
Why didnt you spend more time elaborating on the importance of tighter timings on the AM5 platform? It's fairly well established that the lower the CL, the better the performance. The problem in the market is 48gb DDR5 with low timings is almost non existant and not worth the $ at all if value is concidered. Your tests definately show this is still the case for the 9000 series.
Hate to say this bud, but it’s the UCLK and FCLK that need to sync up, even if that means dropping the FCLK to 1900 when running 7600 MHz memory speeds. Having the memory speed at, for example 6400, it doesn’t make sense to run 1:2 that since the fabric speeds will just crawl along at 1600MHz which just makes everything run slowly, so you would be better off letting the fclk do its own thing. But at 8000MHz everything is synced up again, same as at 6000. R23 uses no memory to run the benchmark so you won’t see gains there, but R24 does hit the memory hard which is why you had the crashes. Just a stability problem causing that. Doesn’t need to be a crazy motherboard like a Gene or Apex, there are a couple good itx boards that OC memory well too. ASUS X670e-i or even the itx B650 version is a good choice 👍🏻
still using my old 3900x with 32gb ram CL12 witha 2070Super Im still debating whether to upgrade CPU or GPU.. or just get a whole new system.. although i think i should wait for the nvidia 5000 gpus for that. what yall think.
I stabalized my 7950x3d to 7200MHz but I also tightened my timings and adjusted voltages, My memory Beats all benchmarks like this compared to 6000MHz and even 6200MHz theres is only a few instances where it loses performance, i got the latenct lower than 6000MHz modules its actually faster than my old kit but the bandwidth is 15% higher so any benchamrks that uses bandwidth is dramatically faster. Do it if you need the bandwidth but if not then i would say stick with 6000. but I'm leaving it at what I have because most tests are actually faster than slower.
I just got my first PC for gaming. I have an Intel i9 14900 KS and Corsair DDR-6000 CL-30. I can’t get any games to load without crashing. I’m new so any ideas if this is the same problem?
Different timings doesn't tell anything about the frequency performance cause very tight timings even the best DDR4 kits can completely annihilate the lowest end DDR5 kits because of their way lower timings, like I have a 4400 CL17 and there's even way better... Also what's the absolute max frequency you can get on your particular system with manual 1:1 ratio fully stable tested???
No it’s still that AMD’s got issues with infinity fabric and the ram controler and just can’t sort them out even after 7 years of Ryzen AMD’s still a slave to ram specs and it’s still a dice roll on how well it works.
Might do more trying to push the latency as low as possible. However, Windows is such a mess right now with these CPUs I don't think you are going to get reliable data.
Wake me up when the 9800X3D is out
shoutet like a lion
Need the 3D in my life as well.
I need 3d cache on apu that are focused on graphics
if im saving money from not buying a gpu , i dont want to spend it on ram lol
Naw I'm betting 9950x3d well be faster since they will learn from the mistakes of the 7950x3d
You made hearing "Wake me up, when September ends"
Seems something is wrong with windows. Linux tests over on phoronix conclude
"the Ryzen 9 9950X came out to being 17.8% faster than the Ryzen 9 7950X. The Ryzen 9 9900X meanwhile was 21.5% faster than the Ryzen 9 7900X across this wide mix of workloads. The Ryzen 9 9950X was 33% faster than the Intel Core i9 14900K performance overall and even the Ryzen 9 9900X was 18% faster than the Core i9 14900K"
I think the scheduler is not great.
I'd be curious to see win10 compared to 11.
Why? AMDs chiplets are more like the classic sockets + northbridge than Intel's. I suspect that the optimisations in 11 for Intel's architecture are being too smart for their own good.
If I'm correct, w10 will show better performance than w11.
@@myne00I think you can be Rigth!
yeah i heard its a windows bug slowing down the 9000 series
Windows is partnered with Intel. They purposefully make amd run slower. Theirs been several lawsuits due to this
it was a completely different benchmark suite?
Some reviewers claim running the benchmarks as Administrator in Windows or running them on Linux gives them about 5-10% boost in performance, could you make a video about that?
Yeah it looks like something is funky with windows drivers/scheduler/something
@@PineyJustice Windows is the best advertisement for Linux
@@PineyJustice I think it is: Phoronix did a series with the 9600/9700/9900/9950, and they were way better than their Windows counterparts. I think the Windows scheduler is playing silly bonks again.
@@PineyJustice always has been
Somebody should let the c**** know at Microsoft about this problem then.
The problem is that Windows 10/11 require 64 GB to stop using memory compression for full windows operation. 48GB is not big enough. You get more speed from stopping memory compression than you do from the speed of the ram.
Does moving up to 96GB help even more?
I have 128 for model programs but 10 and 11 are both horrible OSs 11 keeps getting worse though
6000 is the sweet spot
Not on Zen 5. But, for dual CCX/chiplets, Yes. It is too complicated of an architecture. It's like messenger. Once you start adding in more steps (people) to convey rhe information to the next step (person) then it just makes the information less reliable. By the time it reaches the other end, it's garbled and mixed up. Too many involved. Keep it simple. One chip. But they can only make up to 8 cores on one chip. So, if you want more than 8 cores, you lose potential. But those are for production and editing. Not truly a gaming chip.
@@asmongoldsmouth9839 games are more and more start to using 6 cores, and 8 cores cpus would do a lot good.
@@IZZYX-X 6 cores is used for a good amount of time... better saying games are starting to use 8 cores more and more... 6 cores is the bare minimum nowadays.
Why not try 6200 and 6400? Additionally, two akingtimings, especially sub-timings can give quite massive gains.
@@mickybaus6848 anying higher in some system it could be not stable
All the Ryzen 9000 series CPUs I've tested ran fine at 6400 CL30 on the MSI MEG X670E ACE Motherboard w/ Kingston Fury 8000 48 GB kit
CAS Latency (and other timings) have always been just as, if not more important, than memory speed. I used DDR4 3600 CL14 memory on my 5000 series procs and it nearly eliminated the stutter issues on higher CL16-18 kits that I tested before, along with a more responsive desktop.
Timing correlates with speed. Example: You can speak very fast on the phone, but you are slow in typing the phone number. In case your phone calls are just one minute each, it is not important how fast you speak to get a certain amount of work done. It's more important how fast you get into ur next call.
That's the correlation between memory transfer speed and timings. The nature of random access makes the time required to do the next access very important. It's the main reason why Cache works so well. In first place it reduces dramatically the access time.
If you already have the 7950xt, you're good it's not worth upgrading this time.
what is 7950 xt
@@SigmaProEdition ryzen 4070 anniversary edition
Its worth getting a 7950xtx it will be 30+ faster trust me bro 👍
It’s never worth buying a cpu from 1 gen to the next. Especially high end ones. People get all gooogly eyes over these charts but forget that in real life you’re saving seconds. Even if you save a full minute, unless it’s enterprise it’s not that serious. Same with gaming. People with build an entire new machine for 15 more frames, from the same game they was already playing, and enjoying just fine…….until they saw a chart
@@SigmaProEdition i think he meant 7950x
You can manually set ratio back to 1:1 even after 6000. You can then run memory at 6400 and noticable difference in applications that are memory speed dependent. Also very tight timmings will help even more.
I got 6400 memory, but I'm running it at 6000 for stability.
What cpu? I wanna do the same for my 9950x
@@newbiegain117 7800x3d
DONT BUY 9950x its only 3% faster than 7950x buy that one instead its a lot cheaper@@newbiegain117
Pre-ordered?
@@newbiegain117
None asked
most motherborad need 1.35v-1.4v vddq and vddio on asrock board you need 1.4v
vdd misc 1.1v-1.2v and
mem vdd 1.35v-1.42v or higher if the motherborad doesnt bug out
then you can hit 8000mhz stable if you have a good motherboard (300$+)
and test the stability with memtest (more progarms are better)
That's why I got expo 64GB 6000 CL 30. Still waiting for other PC parts to complete the build.
CL48 6000 ?
I could bought 6000MHz 32GB DDR5 CL30 to 36 for 100 to 120€. Did you pay like 70€?
BRO! I say this EVERY TIME you do a standing video.... you NEED TO fix the audio in that room! It sounds like you are in an airplane bathroom lol ... there's gotta be SOMETHING you can do that's off camera to muffle those echos.... When you're sitting and right next to the mic you don't hear it at all.... because you're right next to the mic.... So aside from spending a bunch of money on a new wireless/lapel mic that you can attach to your shirt collar, you would HAVE TO do some sound dampening.... blankets and cloth on large flat, hard surfaces is what will work off camera...
- so aside from just NOT doing standing vids you COULD reach out to a company that makes lapel/wireless mics and see if you can do a sponsor spot 🤷♂ that might actually be your best bet :)
Sounds ok to me, U need to chill Stop nickpicking. It's not like ur ears are bleeding or something.
It sounds fine
saying "you are in an airplane bathroom" is a crazy flex, we get it uve been in one
Or YOU CAN NOT WATCH THE VIDEO, NO ONE WILL MISS YOU
dude what do you mean there is no problem
7700x with per core negative power curve, always at about 5.5Ghz, 6400MT ram 1:1 CL34 flawlessly stable. No issues especially since last 4/5 agesa updates.
1:57 sounds like you said “I bought a kid” 😂
Again this reminds me of the early days of DDR4 and x99. Trying the super fast stuff early on was a mess but BIOS updates, yeilds etc just got better and better and 18 months later faster stuff got stable. I never jump on a new gen and platform on day 1. Leave it a few months and let someone else take the pain.
video starts at 3:23
thanks now i know what memory i should buy
Not important because it's an overclock and getting it to be stable could prove near impossible.
Yeah even expo 6000 was unstable at release for me but bios updates helped
i am a ram overclocker and i will say for games stay at 1:1 as high as your fabric goes. the penalty is real and for ddr4 you dont see a gain in games till 4800mhz from a 3800mhz and its 1%
Learn this from AM4 when I bought DDR4 4000MT/s. I won’t be migrating to Ryzen 9000, I’m sticking with my 7600X and DDR5 6000MT/s until there is a 50% gain.
I run Memtest first. Instability can be detected in just a few minutes. If it has errors, I won't even try the benchmarks. I haven't had a single crash on AM5 in a year and a half. I have a 6000 CL30 1.35V with lowered timings on my XPG Lancer 6000 CL40 + Asrock B650M PG. 6000 CL28 1.4V also survived the 6 hour test in Memtest. 6200 CL30 is also OK. 6400 CL32 1.4V 1:2 throws errors within 15 minutes. 6000 CL26 1.55V seems to be stable if the temperature is kept below 80°C.
7800x3d running at 6200 1:1. With one major caveat, the memory is at 1.55 volts.
I am NOT recommending this for everyone.
Temps are what degrade memory, and I have excellent airflow in my case. I monitor heavy gaming sessions with hardware info, and temps never reach 40° celsius on the memory.
Why not just run it at 6000 and not have to monitor anything? You will lose approximately 0.12-0.25% performance.
There is no chance your memory chips are around 40° with that savage voltage. You are likely lying or must have a problem with the software you are using either.
If they are at around 50/55° you arr lucky already.
@@badass6300I assume those percentages are based on based on gaming. I see good gains with respect to productivity.
@@jolly7188ok, buddy.
@@dixielanddigital7952 what 3-5%?
i have 3700X and 32gb 3800mhz ddr4 stable! cl16, i think i found a good combo for this cpu, thou im swapping it to 5800x3d soon.
Sounds like Samsung B-die chips.
They must be single rank for sure.
This RUclips video of yours was historic. The first video that doesn't make our ears ring with plosives, i.e. you don't speak directly into the microphone. Thank you very much! Keep it up!
spend the extra money on a better CPU, but then that better CPU needs better memory... it's a vicious cycle
Why do you sound like you're 9 feet away from the mic?
Memory stability also depends a lot from motherboard itself. Trace length/width, interference signal quality. Remember than current boards (6xx series) are release over a years ago and it's may not even tested with DDR5 as high as 7000.
Although AEGSA supports it. CPU and Mainboard are not guarantee to work. See your mainboard specification that what they really test the board up to. It may be just DDR5-6400 or 7200 depending on manufacturer.
For DDR5-8000 module that exists, they tested it yes but they don't tell you what CPU / Board combination they using. It must be very high quality though to possible to run that test.
Ryzen processors, which are normally targeted towards consumers, are definitely not with Zen 5. Steve from Hardware Unboxed pointed this out. Zen 5 is not aimed at consumers. Despite AMD showing gaming benchmarks for Zen 5, they were total BS. AMD clearly didn't design Zen 5 for consumers such as gamers. It seems much more focused on AI/enterprise workloads. Something that most consumers don't give a shit about.
Most computers are used for things other than gaming. 👍
@@DingleBerryschnappsSometimes i do gaming, but mostly i used my computer to make electronic simulation, design schematic and pcb. Sometime i coding.
You realize that gamers are their minority of consumer right? Like a quarter of their revenue is from gaming. The rest is all servers and ai processing but pop off bud
@@DingleBerryschnapps For general PC use cases such as gaming, 3D rendering, video editing, etc, Zen 5 really is not for that. For those use cases, you are much better off with Zen 4. Same performance at much lower prices. Zen 4 is faster sometimes.
Zen 5 beats zen 4 in data center benchmarks and tasks. Any desktop use case is equal or below Zen 4. Current Zen 5 chips are all built using TSMC 4nm tuned for efficiency. AMD can switch things around and use 4nm performance skew to release this CPU's at higher TDP. But likely they are just waiting on Intel's next "gaming champion" to burry it with Zen 5 X3D that they don't need to release before Intel shows its hand.
Have you tried the Hollyland mic? It’s fairly cheap and has great audio and amazingly easy at concealing on cam. Throwing on cheap acoustic foam and some carpet will help clear echo in a room as well.
Audio > visuals
RYZEN 9 7900x with. 6000 being the sweet spot.
Anything above 6400 was a total disaster. I have become an expert in clearing CMOS...
With a 7800X3D.
this was great to see you testing hardware too :)
there should be a benchmark showing the 8000 ram running at 6000 but with tighter timing.
Prob better to wait for the new chipset to come out and go based off that.
You're testing memory frequency and using cinebench which almost shows no scaling with memory.
Good work man. It stuns me that reasonably priced RAM is the best for AMD CPUs.
I myself tuned cheaper 6000MHz RAM from 36-36-36-96 to 30-34-34-76@ 1.4V with tREFI = 65565! I could not be happier.
G-Skill Flair X5 EXPO only RAM with Asrock x670e Steel Legend motherboard. 🤘
I have a kit of CL32 6400MHz RAM that I ran at 6000MHz, 6200,and 6400MHz. I never noticed a difference between them. Timings and channel ranking is more important than frequency.
Performance stops increasing after 6000, that's why it's called a sweet spot
@@the_golden_s5682 I run my kit at 6200MHz. It's my own sweet spot.
Me personally I'd rather the slight latency penalty.
I need the memory bandwidth for the workload I'm doing.
The optimal FCLK for DDR5 6000 is 3000MHz, but in reality, the limit is 2000MHz...
When people say 16gb of ram is enough they usually can't afford more and are trying to convince themselves and not you :P
I don't understand then! DDR5-6400 Confirmed as Sweetspot Speed of Ryzen 9000 "Zen 5" Desktop Processors! As part of the update, Ryzen 9000 "Granite Ridge" should be able to run DDR5-6400 with a 1:1 ratio between the MCLK and FCLK domains. This is a slight increase from the DDR5-6000 sweetspot speed of Ryzen 7000 processors. /TechPowerup/.
Infinity Fabric clock (FCLK) has been increased to 2400 MHz /Zen 5/
Now is that a lie?
I planned to buy such ram for the 9900X... 6400mhz CL30 and 1:1 ratio! Or will this only be available with 800 series motherboards?
As always AMD works after a year of updates, if you want to upgrade you should wait for the new MBs and after BIOS and drivers updates
@@reiniermoreno1653 Yes, I want to upgrade to 9900X! So then I have to wait for the 800 motherboards, and then the 6400 can work 1:1? As far as I know, the 800 motherboards will be launched on September 30th. I didn't want 600 or 700 motherboards, only 800, which is my choice! I have a 3900X, I am developing from it, it will be a fairly decent jump, and if it really works with 6400 rams 1:1, I can get quite good performance with CL 30 timing. Otherwise, how can it work with 6400 1:1? So our Infinity fabric should be 3200mhz, or not?
i thinks it's only for new chipset
Doubt that chipset will rule that one. Most of the time the issue lies up to the CPU RAM controllers imo.
@@rayiusflyrant268 chipset doesn't manage memory, that was until 12 years ago CPU and drivers does
But you can take an 8000mhz kit set it to 6000mhz for faster timings and divide its 8000mhz timings by the difference, so for example my 4000 mhz ddr4 kit had timings of 16-16-16-36CR2 @1.35V and the difference between that and 3200mhz i had to run at for my old cpu was the same as the difference between 6000 and 8000, so for example i divided my timings by 4 since and multiplied times 3 because the new frequency was 3/4 the old one so new timings were 12-12-12-27 and got it to 11-11-11-24CR1 @1.45V of its 1.5V max recommended by samsung for b-die at the time.
ryzen 7000 taught me that timings matter more and that stability requires more than just loading an expo or xmp profile. There's a reddit thread dedicated to stabalizing OC'd ram. I was having a hard time booting 128gb of ddr5 6000 on a 7950x3d. *2 64gb kits. It was only stable when I removed a kit.. but that thread saved me having to return ram or waiting for 9950x3d to be released.
I have a Ryzen 3700x paired with DDR4 3600mhz RAM.
Now I think random shutdown and rest happening because of RAM speed.
In my GSkill Ram there is only 1 profile i.e., 3600mhz
So how do I set 3200mhz for stability?
Or is 3600mhz really an issue?
Please help me out 🙏🏽
Having fast ram doesnt mean its actually running at that speed. It doesn't default to the higher speeds in bios unless you set it so. Easy to check using CPUZ. Run a memtest at the current settings. Check Hard drive health ie the usual troubleshooting.
Makes sense, I have an old ryzen 3600, and I was running 3600mhz, i put in 4 sticks of 3600mhz, and it buckled completely. The ryzen mem controllers are not amazing, and they seem to have a hard time running high speeds it feels like. I hadnto reduce speeds to around 3333 mhz to get any kind of stabvility
It is also related to cl30 vs cl40+ and the rest of the sluggish timings over 6000mhz. Further, 6000 to 8000 bandwidth when it is useful is running a heck load of work or serving a lot of clients. Then the 8000 will be beneficial. The ideal is the Trident.Z 6400mhz @ cl32 2x32gb which is still very cheap. Im running mine at 6000 cl30 and at 1.38v instead of 1.4
instead I got an epyc 9124. I have 12 memory channels at 4800. So in addition to having more total memory bandwidth, I can hold more ram. The motherboards were about the same price but the cpu was +$500.
why his voice is pitch high and pitch low at the same time?
6400 can be run 1:1 too! All you need to do is to enforce it in UEFI/BIOS.
DDR5 CAS timings are out of control, we need CAMM2 vs Stocket.
The new chipsets on the the (upcoming) motherboards isn't going to help. Memory accesses don't go through the chipset -- it's strictly CPU-to-RAM.
If there are AEGISA updates to help with this, those can be incorporated into the existing mobo's as easily as the new ones.
Watching this with my 5800x3d and 3600mhz CL14 memory just fine.
Any news on whether 128gb is stable? I never upgraded my 64gb on 7950x due to issues, I need 128gb absolutely minimum to work before considering any changes
My i7 13700k breezes threw 6600, no crashing jittering or freezing!
your CL timings are way too high u can get 7400mhz with cl34 to get lower latency which is whats actually gives performance
Aren't the 16gb sticks A die and 24gb sticks M die? Wouldn't that also affect performance
im running a 7800x3d bundle from microcenter it came with 6000mhz ram and it runs at that speed great. ill try to get it to run at 6400mhz and see what happens.
I've been running 6400 cl 30 on my 2 7800x3d systems and 7900X since the launch of them. I am using the Asus ROG x670E gene board that is a 2 dimm board.
I wish you would have ran all the memory speeds and benchmarks again with pbo enabled maybe in the next video
so with the new motherboards can I get 6000 on all 4 memory slots ?
Best performance with 1-to-1 DDR5 6000CL28, possibly G.Skill Trident Z5. If your CPU memory controller is good enough, you can try overclocking the up to 6400 (still 1-to-1 ratio)
i can get it to the beach cause i'm not wasting that sand
hey meld could it be the crashes are from MS GAME BAR ISSUE IN WINDOWS... BECAUSE THAT IS EXACTLY WHAT HAPPENED WITH THE 7800x3d and 7950x3d you need GAME BAR ENABLED to make them work the same is with the 9900 and 9950x and am sure the X3d chips will have the same issue ... THIS IS MS WINDOWS ISSUE NOT AN AMD ISSUE OK GET THAT CORRECT...
i have the r7 7700 non x (180usd) running t-create 32gb(16x2) at 6000mhz and works fine
That the tip of the iceberg for Ram explanation
You hyped Zen 5 so much, why didn't you review it against all the current CPU's for us all to see. Much more interesting than seeing it with different ram speeds?
The results might not look good for Zen 5, that is why.
Maybe because the video is about memory compatibility, OC support and system stability?
Why didnt you spend more time elaborating on the importance of tighter timings on the AM5 platform? It's fairly well established that the lower the CL, the better the performance. The problem in the market is 48gb DDR5 with low timings is almost non existant and not worth the $ at all if value is concidered. Your tests definately show this is still the case for the 9000 series.
Hate to say this bud, but it’s the UCLK and FCLK that need to sync up, even if that means dropping the FCLK to 1900 when running 7600 MHz memory speeds. Having the memory speed at, for example 6400, it doesn’t make sense to run 1:2 that since the fabric speeds will just crawl along at 1600MHz which just makes everything run slowly, so you would be better off letting the fclk do its own thing. But at 8000MHz everything is synced up again, same as at 6000.
R23 uses no memory to run the benchmark so you won’t see gains there, but R24 does hit the memory hard which is why you had the crashes. Just a stability problem causing that.
Doesn’t need to be a crazy motherboard like a Gene or Apex, there are a couple good itx boards that OC memory well too. ASUS X670e-i or even the itx B650 version is a good choice 👍🏻
still using my old 3900x with 32gb ram CL12 witha 2070Super Im still debating whether to upgrade CPU or GPU.. or just get a whole new system.. although i think i should wait for the nvidia 5000 gpus for that. what yall think.
I stabalized my 7950x3d to 7200MHz but I also tightened my timings and adjusted voltages, My memory Beats all benchmarks like this compared to 6000MHz and even 6200MHz theres is only a few instances where it loses performance, i got the latenct lower than 6000MHz modules its actually faster than my old kit but the bandwidth is 15% higher so any benchamrks that uses bandwidth is dramatically faster. Do it if you need the bandwidth but if not then i would say stick with 6000. but I'm leaving it at what I have because most tests are actually faster than slower.
6400 will be sweetspot on 800 series motherboards
Memory has always been down to clock cycles and timings not Mhz. It's been known for years, but they still try to overcharge for Mhz it's a scam. 😒
Uh... "down to clock cycles", "beer they charge for mhz"
You do know what a clock cycle is and how it is innately tied to mhz, right?
love how this beaver knows soo much about tech. subscribed
I just got my first PC for gaming. I have an Intel i9 14900 KS and Corsair DDR-6000 CL-30. I can’t get any games to load without crashing. I’m new so any ideas if this is the same problem?
check out if it's not because of the 13/14 generation voltage problems.. it's a widespread issue
Still waiting for the news that the Zen5 launch on desktop is a massive flop.
AMD has made a really bad cpu with the 9000 model and left it open for Intel
what a blunder AMD has made
Different timings doesn't tell anything about the frequency performance cause very tight timings even the best DDR4 kits can completely annihilate the lowest end DDR5 kits because of their way lower timings, like I have a 4400 CL17 and there's even way better...
Also what's the absolute max frequency you can get on your particular system with manual 1:1 ratio fully stable tested???
Give it a few years and we will have 8000 kits with good timings.
what is up with the audio?
Not sure how you don't fly away with how much you move your hands when you talk.. LOL!
mainstream techtuber learn how to tune memory challenge:
The echo from your mic placement is distracting to watching your channel!
Try it an intel CPU it will work perfectly
Please dude dont throw pc parts around
It's ridiculous how amd can't make a 8000 memory perform on their chips
Intel is kinda in the same boat too 🤔
Your mic died?
Undervolting and nice timings are the key
On my ryzen 7900x i got my ram stable at 7200mhz cl32 more bios updates are making higher speeds better on the asus x670e hero
Finnaly a video that isn't just about news.
Probably, they rebrand the 7000 series ,
i manage to get 8000mhz working on my g skill 7600mhz 40 46 46 120 i get 23k in cinebench r23 on my ryzen 7 9700x
All these overclockers complaining that mid level CPUs weren't using OC RAM based on outdated info from the 1st and 2nd gen Ryzens.
No it’s still that AMD’s got issues with infinity fabric and the ram controler and just can’t sort them out even after 7 years of Ryzen AMD’s still a slave to ram specs and it’s still a dice roll on how well it works.
No more threadrippers???
Might do more trying to push the latency as low as possible. However, Windows is such a mess right now with these CPUs I don't think you are going to get reliable data.
Unless you have A dye on your ram then don’t even bother.
however, if you happen to have a fan laying around then perhaps. just be wise lol
Unless you have A dye on your ram then don’t even bother.
however, if you happen to have a fan laying around then perhaps. just be wise lol
Very useful information. Thank you! I am a subscriber.
5500x3d and 4070 gddr6 is coming
Nany nan fity eks... 😆
I mean, it is the same imc die used both Zen 4 and Zen 5, thus i highly doubt that there can be any real gains with ram speed
WE DONT GOT ANY ISSUES RUNNING ANY MHZ ON iNTEL LOL