Seconding this. I'm building a new AM5 PC with 64gb of g.skill 6000 CL30 RAM and hopefully finding a 9800X3D if they'll every stay in stock for more than 5 seconds. Would love any tips specifically on tuning that RAM I can get.
You should consider putting some affiliate links in the description you deserve any compensation if people are gonna go out and buy some products based on your findings
@@SomeoneElse6282bought a 7200 2x16GB teamgroup kit, overclocked to 8000 and 8200 cl34. Got it for $110 on a slight sale. Can probably go higher, haven’t tested yet. 9800x3d, MSI X870-P wifi
I got the GSkill Ripjaws S5. It's lower profile without RGB. Using the D15 G2 cooler. Went with the 64gb 6400 cl 32 kit with my new 9800X3D. Running the Asrock X870E Nova motherboard. This memory ran fine with the EXPO profile at the default 1:2 ratio.. But when I set it back to 1:1, I had to dial it down to 6200 to get it stable. I was then able to tighten the timings from 32-38-38 to 30-37-37. I tried 2200 on the IF clock, but went down to 2133 in the process of trying to get the system stable again. I need to see if I can get it up to 2200 I guess. I haven't gone into the sub timings yet, but I want to. A detailed guide would be very welcome! Any tips for getting 6400 stable would be welcome. I did try raising the memory controller voltage to 1.25v and that got me a couple minutes of game time before crashing, but still unusable. Maybe I just need to suck it up and go higher. I had that 1.4 is safe, but I don't want to stress the chip for gains I won't t really feel in real life. I'm using a 3090FE at real world resolution and graphics settings. The GPU is definitely the bottleneck right now.
Same here. I have adata xpg lancer 64 gb 6400 cl32 hynix a dual rank kit. My mobo Asus rog x870A. I couldnt get Windows 6400 1 to 1 mode. I run at 6200 cl30 tuned sub and secondary timings with buildzoid. Fclk 2066 ( buildzoid says 6200 much stable at 2066 fclk ) . Im happy right now.
some times you need to mess around with the SOC voltage for IF, its a balance between ram speed and IF. sounds like a great build though. it takes some playing, I have been trying out so many different configs on this system I haven't fully figure out what my final settings will be yet.
@@JohnnyRage303 I tried 1.27 vsoc with 6400 1 to 1 mode. It didnt work. Probably if i give it vsoc at 1.30volts it will work. But i dont wanna go higher vsoc above 1.25. Rightnow 6200 1 to 1 vsoc at 1.25 im stable with 2066 fclk. I am lucky to ask buildzod on streaming my timings. And i tuned a lot my timings. A data gives me 67.00 ns latency 66.000mb read 92.000 write.. I am happy actuually.
@@eyluloktay9104 Looks like your are having a very similar experience. Last I checked, Buildzoid didn't have a memory guide newer than 7800X3D? But his info looks very extensive. Maybe over the holiday weekend I'll get into the sub-timings.. I also haven't touched PBO yet. I don't know where AMD is actually standing with warranty stuff. I'm too worried, but there's no reason to take any chances with this chip, when there is so little to gain.
I don't know a thing about RAM timings and tuning them. I simply check qvl on mobo list, plug the sticks in, enable xmp and pray for the best. Luckily, my 64gb 6400mt/s cl32 ram kit works great with xmp enabled with my 9800x3d. Couldn't even tell you if it's running 1:1 or not since i have no idea what that means. Using a b650m aorus elite ax board which didn't even list my RAM on the qvl.
With 3d cache you might not even know when you have some errors. The thing with RAM errors - it often breaks something else along the way. I had frequent GPU driver crashes, tried everything i could imagine until i ran another RAM test and caught a few errors. Tweaked some timings and no more GPU crashes.
My comment didn’t age well. The 64gb 6400mt/s cl32 kit causes intermittent issues with booting up resulting in bsod. I am guessing because this is an xmp kit that could be why. Gskill sells a expo ram kit with the same timings but only up to 48gb and it runs at 1.35v instead of 1.4v
@@e92justin It's not that your comment didn't age well. The way you described how you configure your computer is how 99.99% of people should. It's on the various manufacturers to ensure that the QVL is true. It should just work and if it doesn't they failed. I consider myself an enthusiast, but at the end of the day, I just want things to work without any fuss and expect to accomplish that by, in this context, simply enabling something like XMP or EXPO.
PC Specs: 9800x3d X870E Crosshair Hero RTX 4080 super Zotac Memory Kits: G.SKILL Trident Z5 Neo RGB Series AMD EXPO 32GB DDR5-6000 CL30 Ok so i originally had a MSI Tomahawk x670e MB inside. i was getting Post issues during re-starts ONLY. switched to the Asus x870e crosshair hero, seems like the problem improved BUT its still present. using expo on occasion system wont post from a RESTART, cold starts are usually fine. Having a Few USB issues also. system seems to hang on things like keyboard/mouse. its weird, because otherwise things KILLING benchmark's. any help? reading the forums etc this issue seems to be VERY wide spread, posting issues specifically from reboot's/re-starts! whats the issue here guy's? AM5? AMD? FEELING LIKE 3.7K WAS WASTED! help!
i'm guessing with the 64gb vs 32 gb kit thing. is the 64gb kit is dual sided so it's basically it looks like a 4 stick kit to the cpu. like back in some of the previous gens they were getting better performance with 4 sticks vs 2
I've heard that Star Citizen is supposed to be abnormally impacted by memory. Would have been interesting to see that one included in a test like this.
rule of thumb for the fclk is that its good to sync it in 2 to 1 mode when youre running 8000 Mts+ ram. Anything below just get the fclk as high as it can go, same for the ram.
Noticed XPO/XMP settings kept boot looping but manually OC Ram was the way to go and stable. Interesting having to use the OG way to get new tech working
That's wierd I have had expo and xmp work on all my am5, after the initial Ling start for memory training. Usually I just find you can improve the timings alot from expo xmp. But either way stable is stable.
There's a whole new marketing opportunity opening-up here. We could see memory advertised as "runs best with ...". Potentially COD memory kits could be a thing lol.
Tried 8000 cl40 expo on a x870e godlike from a QVL set and it wouldn't even post. Looks like I won't have any real performance issues just sticking with 6000 cl30 though.
So I came to the conclusion that the 6400 cl32 kit I have is about 1.25% faster overall throughout my testing then my 6000 CL28 I bought to compare it to. I ONLY used EXPO timings and did not do any tuning. I also game at 4k so I was already against the GPU BN in MOST cases.
I would try to force the cpu bottleneck because if you're seeing that difference at 4k i bet it's higher while the cpu has a higher load.. However if your real world application is 4k at the end of the day it is what it is.
@@TheRealDlo Higher mclk with good timings is the way to go even over a lower mclk with excellent timings. Some games and applications are faster with lower latency but overall higher frequency is better
I bought a kit Corsair 64 GB 6600CL32 to pair it with the 7800X3D on an Aorus B650 Elite AX. My goal is to have better on Star Citizen than with my current Lexar 32GB 6000 CL30 I hope that the new feature at Gigabyte of "Turbo BOOST X3D" will allow an improvement in performance. Otherwise, I will try to decrease the timings to have a kit of 6400CL30 or 6200CL28 Cross my fingers
@@JohnnyRage303 Being an Alpha and depending a lot on the performance "server", I think this game is complicated to bench. But with the imminent release of 4.0 it will become much more stable on the server side and I think there will be an interest to do comparative tests. That being said, the game needs 36 GB to not be "pinned", it is a fact already demonstrated. Otherwise, to answer more precisely, walking and doing the same route in the city of Lorville is perfect for making benchs
@@NEIHERUniversStarCitizen This game will quickly become the standard of measurement for "games requiring a lot of resources in CPU and DDR" in the future. Very soon, if it’s not already the case If the performances are good on this title, then all other current games will pass the bar of performance "more than acceptable", like CyberPunk2077
I've made some tests while fiddling with the memory timings and I can absolutely tell that a nicely tuned memory can outrun any of the shown examples. While 6000 cl28 seems good enough, I'd argue that 6400 cl30 FCLK 2133 is a real sweet spot, at least for X3D chips that still have higher CPU-to-memory latency
Can 6400 C30 - 2133 FCLK be doable on a M-die 64GB dual rank? That's what I have but I dunno what I can take it to on a 7000 series while trying to maintain a v1.4 on the ram.
My team group X2 16GB (32GB) 8000MT/s kit in AIDA64 reads 63,805 MB/s with a latency of 59.9 ns, using tweaked timings. In games, I believe latency is more important than bandwidth. FCLK 2000 and UCLK 2000. Ram 38-46-38-48
I rum 9800X3D on 5.5Ghz@1.250V ,Kinston (Hynix) 32GB 6400Mt/s on CL28 1:1 on 1.480V (with fan top of stiks because if the ram reach 50C crash) and FCLK 2133 @ 1.260V. My latency is 66ns (all stock was around 90ns,stock CPU with only EXPO on 6400 was 80ns) I play alot to run FCLK on 2200 because the latency will drop to 62-63ns but unfortunately the system is not stable
The higher your soc voltage, the harder it is to stabilize fclk. But soc stabilizes memory frequency. So it's a balancing act. But your setup sounds awesome. Nice settings.
is there something wrong with corsair cl30 6000 kits? i have a 32gb vengeance rgb kit and looking at 64gb kits for Tarkov, and i like those dominator platinums, but i rarely see them in this kind of videos. is there a reason, except for price?
There's nothing wrong with corsair many times there xmp profiles are a little more aggressive which is good. You probably don't see them as much because they are more expensive and if you are tuning the kit anyways the brand doesn't matter as much. They are all the same hynix a or m die underneath.
Thank you for sharing 😊 are you using nitro setting for 6400? Man I am having a hard time with 9800 2x32 on a x870e strix. I’ve only gotten loose 6000 even close to remotely stable. Anything else either errors right away or hangs on a 2f or 46 I think. Pretty frustrating
I have 64 gb 6400 mhz cl32 kit. I couldnt run that 6400 1 to 1 mode. So i watched buildzoid videos and asked him my secondary timings. I tweaked a lot. With my 9800x3d im running at 6200 cl30 fclk 2066 with tweaked secondary timings. Pbo +75 total 5.3ghz I have cinebench r23 multicore 24960 score. Time spy cpu score 17440 I have 4080 super and with the shadow of the tomb raider utlra performance dlls im getting 406 avarage fps.
Great idea, what is the rank of modules tested? I think the 64GB modules may have an edge thanks to being dual rank. The fabric is a big bottleneck of am5. Same read speed for ddr5 6000 and 8000 are crazy.
I remember with AM4, dual rank helps gaming performance even though other benchmarks like AIDA64 show no performance difference. Seems to still be true.
6200 CL28 1.435V with very tight timings and DDR Nitro settings with 2200 FCLK completely stable (OCCT, Y-Cruncher, 24 hours in Stalker 2 no issues) AIDA Read: 70190 Latency 64.8 Can run 6400 but with nowhere as good timings, 6200 tight gives me best results. FCLK needs proper testing to see if stable or you will run into issues. Benchmate Y-Cruncher 0.8.5 PI-2.5b will quickly pick up issues, then longer tests can be run. 2066 for 6200 is better than 2100 and maybe 2133 but 2166 and 2200 overtake that advantage. Could NOT get FCLK stable higher than 2066 until raising 'VDD MISC' Voltage to 1.25V ASUS ROG STRIX X670E-F GAMING WIFI 32gb 2x16 Corsair Vengeance 32GB CL30 1.4V sticks
Thanks for all of that info, this video was really just me trying to find what direction I wanted to go. I'm gonna get into tuning more in depth soon. This is some good additional info to steer my setup.
I've only played around with 2 stick kits. I know with 4 you can't clock as high. I thought last gen with some of the bios updates people had gotten 6000 stable but I could be wrong.
Just got my 9800X3D today and it memory overclocks better than my 7950X3D on my X670E Carbon. I have a 7800 CL38 XMP Adie Gskill kit that is on most of the AM5 mobos QVL but XMP or tuned didn't work on my 7950X3D at the rated speed. XMP and tuned work at 7800 and 8000 (haven't gone higher because I figure it won't work) on my 9800X3D. My 7950X3D was able to do tight timing 6200/2067, but I feel like this 9800X3D might be able to do 6400 which I will try soon. Pretty stoked to have 8000/2000 working with tuned timings just because I think its fun and low SOC voltage is dope.
I would love to see you tweak the ddr5 8000 kit for better timings. Memory bandwidth and latency might not matter that much for gaming but it does for AI inference from system memory!
My main goal was to look at the xmp kits and I did a little bit of primary settings to see what direction I wanted to move in the future. I am going to dive more into tuning these kits further in the future.
the best kits to buy is 2x16 8000 c38 because they are the best bins and always hynix a die and then punch in the timings yourself, if you don't want to tune it yourself and only enable xmp you should at least put trefi to 65000, be careful tho you should put a fan over the ram of it gets too hot because of the high trefi
So MCLK/UCLK synchronisation with FCLK doesn't seem to be optimal with the 9800x3D in gaming... that is unfortunate for those with OCD, I guess, but fortunate for the rest of us. Pushing the IF above 1:3 in 1:1 and getting better results isn't as surprising to me than breaking 2:1 synchronisation... I guess for applications beyond gaming and on non-3D processes sticking with 8000:2000 and tightening the timings might work out better though.
What do you think about the cl28 kit form gskill that's at 6000mhz? the neo royal one. Just wondering since you didn't review it in this video. It seems to me in this video lower latency is better but, just curious.
I just saw those were released, amd does love low latency. I had the royal kit on am4 and was happy with the product. I haven't tried them in ddr5 yet. Usually it's their top line so I would bet they are binned pretty well.
@johnnyrage just curious how come you didn't set the fclk to 2200mhz on the 32GB kit? Stability issues or no? I have the same 32GB kit you showed in the video and am running it @2200mhz getting 69k MB/s read, 92k MB/s write and 74ns latency. The only thing I had to do was set my soc voltage to 1.15 to get it stable after stress testing. I'm running a 9800x3d on a ASUS x670e MB.
Honestly I ran the 32gb kit at 2200 for the Aida and cinebench tests but was leaving the house for Thanksgiving. And most the feedback I've had people seemed to think I was crazy running 2200 not 2:3, so I limited the full gaming runs I completed mostly just due to time. Seems like on this video alot of people are more curious about the 2200 so when I get home I might do some more tests.
I wonder if the larger memory chips help reduce the number of rows involved in the majority of read commands? Overall it might reduce the number of clock cycles per command required. An uneducated guess anyway haha
My first main goal was to just compare the xmp /expo performance for people that just want to plug and play. The second i was just playing around with settings as I went. the 64gb 6400mhz kit was the most flexible, my 32gb 6000 kit is the oldest and would regularly not post as i tried to add performance (thats the silicon lottery) I was unable to run as many profiles on it. The 8000mhz kit I could have explored more and I will in the future, I'm also finding these high end kits like high voltage you can push them as far as you are comfortable. But simply the 6400 got the most attention because that kit was very flexible in the time I was filming. I'm testing these kits again on my 9950x as well. I'm trying to keep it a little more standard since I've got that feedback.
Expo 1 is usually the faster profile. You want to at least enable that. You can find a little more performance with tuning. But it's up to you if you want to research that.
So the G.skills 6000 64gb CL28 you are using is 4 dimms of 16gb correct??? Because g skills own site as well as Newegg and Amazon all say that G.skills doesn’t make 6000, 64gb, 2x32gb dimms. Please confirm?
if you can get there :/ mine does 2133 max stable. I can get 2167/2200, but the amount of tweaking voltages to get that stable there is 0 room to o.c. ram. Undervolting soc = flck stability, but lowers ram stability. 2133 with ram @6400mt/s is about where I'm fine setting the scales.
@@Dante_S550_Turbo i haven’t been able to run 6400 below vsoc at 1.25. Most of the people can easily run at 6400 at 1.20 vsoc. So for stable 6400 i need to run at 1.28 vsoc at least. At this high vsoc voltage fclk become instability. So i ended up with 6200 tighter sub and secondary timings. With lover vsoc at 1.16. And fclk 2200. It is rock solid stable. No vt3 errors no karhu errors.
I found a few cases where my framerate increased when I went from 32 to 64 on my current system. star citizen and MSFS both gave me a significant framerate increase weirdly. I got the 48gb kit on my dresser my 9800x3d is in the mail. so new system coming up.
You should show cl34 or cl36 8000mhz ram with tight timings. With my own cl34 M-die configuration, i can reach cl-34-44-44-80. I got the read speed of average 93K. FLCK 2200
I really hate to be that guy but tuning the sub timings isn’t that hard and without doing that this becomes a pretty misleading video. Title should be “if you are unwilling to work at your memory these are your best options” or something like that. 8000cl36 (fclk=2000) or 7600cl34 (fclk=1900) AIDA latency ends up around 5ns faster than 6000cl28 or 6400cl30 with about the same bandwidth on a single ccd CPU. Seems like all the 9000 series CPUs will do ~8200 with no problem and some will push 8400+
A big warning for x870e Asrock mobo users G.Skill even on the qvl list are giving Asrock mobo users major issues including not even booting up at all. Then the expo isn't stable if you are lucky enough to have a g skill kit that boots up. Some have reported weird glitches with voltage management. It'ssuppose to run at 1.1v/1.4v depending on default and xmp. Actually running on your system @ .9v/1.2v which is unstable and possibly the reason why they aren't working well.
Has anyone have any experience with the CORSAIR VENGEANCE RAM I have a CL30 32GB 6000 on my recent 9800x3d build and I've noticed some lock ups when restarting the PC randomly. Should I return in and go for the G.SKILL Trident Z5 Neo RGB Series 64GB 6000 with expo?
Corsair vs g.skill corsair usually run slightly tighter timings at higher voltage. My example I have a corsair 6000cl30 that is 1.4v xmp with tighter timings and a g.skill 6000cl30 1.35 with slightly looser timings. If you're going to tune either brand will be fine. Just running xmp corsair might have a slight advantage, probably unpreceptable. I haven't had lockups with either.
having more ram does not gain you any performance what it allows you to do is have more ram. some games like star citizen wont run on 32GB for example and it should be a good example of whats to come in the future.
do u game with google open ,,,a few tabs watch youtube or stream at the same time? then you need more ram honestly i linked 2 videos the other day in 2 other toutube channels about ram a game like star citizen uses 27-34 gigs ram on 1440p and uses 44-56 gigs on 4k this all depends on areas however star citizen is not even complete yet when all the players load in and all the AI npcs i think its gonna be even worse as off now you may see 1-4 people running around with about less the 10 AI npcs in city sections they still want to add live AI that walks around ..does shopping and flying there own spacships their aim is to make it look like real life many people and npcs in areas to look lively like new york streets no idea if they will ever get it right just dont go less then 32gigs at the moment games are growing hungry and game worlds and data keeps increasing stalker 2 uses 22+ gigs at 1440p and u need space to run windows also windows 10 and 11 can compress..to about a small 4 gigs in size on ssd or nvme ..but still needs 4-9 gigs to run itself so dont go below 32 gigs 48 is better as it is A-die memory on ddr5 and 64 is also good..a little better future prooffing vram for gpus on stalker 2 ....8 gigs makes the game not run and so look at 12 gig vram on gpus and up..to safe yourself
@@themarketgardener ahhh true, im doing same building new pc. Not sure if ill end up with a 5000 series card though, i feel like better to wait for super/ti versions and thats too long for me
i have same board and the same ram 6000 cl30, do you not have issues when putting expo? my pc refuses to boot with expo, i have to disable memory context restore for it to work and it takes ages to boot
No I've used the 32gb kit in a msi x670e, asrock b650e, and now the asus x870e. It's ran in expo on all of them. First boot is slow then it's trained and boots fast.
@ pretty odd, i have the same asus, 9800x3d and the same exact ram and without memory context restore disabled it wont boot at all.. do you have a bios beta version perhaps?
68ns thats high… go for low 40s, even your main timings are bad 🙈 you will see huge improvement with mem tweaking. fclk tuning is not so great on this chips. its good on the G chips. next year Ryzen Max+ will come out, and than the mem tweaking will be realy cool
I just purchased this cpu with x870e aorus pro ice and 6000 mhz 2 sticks rams and CL 30 but dam everytime i restarting my pc led for RAM goes on ,after restarting will boot but the main issue is stuttering on desktop i cant even play games , the gpu is up to 99% utilisation at max freq , wtf is happening, i had intel before but I haven't reinstalled windows just uninstalling the intel chipset and all others, no EXPO enabled, the rams are on the QLV list of the mobo !!! It passed 2 day , i am going crazy......
1. Could you go to the Windows Device Manager and find out if the Display Adapter works properly? 2. Go to windows defender and check if memory integrity in on, if it isn't then look what driver is causing it to be off. 3. Try to boot into windows safe mode, then check the GPU utilization, if it isn't 99% anymore in desktop then it means you might have accidentally installed crypto miner 4. In the command prompt window, type SFC /scannow. Let it run and then reboot computer 5. It's highly recommended to take the bullet and nuclear option and just clean install the windows.
@ I found the issue, fir the stuttering was the rizer cable which is 3.0 and i have tested the ram modules and they are ok but still i have 0D and ram error code and when i have this error after i am restarting the pc the mobo its reading only one ram module i have to turn off to get her to read both sticks and i think could be because of the ssd which is 3.0 and i see on the motherboard diagram that there is no pcie 3.0 for the slot, they communicate directly with cpu through pcie 5.0/4.0 , both my ssd are 3.0 and one it's connected to chipset through 4.0 connection and the other one share pci x16 with gpu but the pcie x16 i see its only 5.0/4.0 , and the first ssd slot which is directly connected to cpu its empty, i have tried with only one ssd connected to chipset (last ssd connector ) and it seems tha was no 0D error or ram , in conclusion i am not sure if the ssd its the cause but definitely the rizer cable was one , i will buy a new 5.0/4.0 one to see if that its the issue
I have had problems in the past with this. In the bios PCIE channels are set to AUTO by default. you can manually set them to 3.0 / 4.0 / 5.0 per channel when you are running a bunch of different devices like this sometimes it can help to manually set them by device rather than leave in auto. But ultimately for me that was a temp fix until I just upgraded everything to PCIE 4.0 or higher.
@ The stuttering was because of the rizer cable which was 3.0 , i am going to order a 4.0 but the 0D and ram error still there , if i am not restarting the pc everything works fine
@@JohnnyRage303it's the most played game on Steam. Most competitive eSport ever. Test at 1080p. Do 3 runs at high to get that average, 3 runs at all low settings and get that average. Competitive players really really want to have the best hardware possible.
Hi, i bought gskill 2 x32gb 6000 cl30 36 36 memory kit, i have a 9800x3d and strix x870e-e, I'm a casual gamer, should i just set everything on expo and im good? I don't wanna tinker so many things for small benefits.. But If it gains so much i could do... Can someone help?
It's a personal choice the expo profiles are huge over the base frequencies. You can always find a little more performance tuning. You just gotta decide if it's worth it to learn.
Many people recommend using 6000mhz cl30 or cl32. but where I live the price of 6000mhz and 6400mhz have similar prices. Which do you think is better for me to take, and is 6400mhz plug and play?
I bought a 2x32GB G.Skill Trident Z DDR5-6400/CL32 kit, running on a Gigabyte AORUS Pro X with a 7950X3D. I tightened the timings a bit, set FCLK to 2133, and left MCLK:UCLK at a 1:1 ratio. Runs great. Read is around 80,000, and Latency is 78.3. My system wouldn't boot with the default XMP profile and FLCK=2200. Be aware that only about 25% of AM5 CPU's can handle that FLCK=2200, while about 80% can handle 2133.
Worrying about 4K res on Time Spy CPU test is wrong. The CPU test is isolated not to care about resolution and rendering. If you look at just the CPU score and not overall, Extreme would be a better idea on any 8 core or higher core count CPU. For 8 core normal Time Spy still works (it stops scaling well past 8 cores). In general unless you are running a grossly wrong test (too light), all 3DMark graphics tests are purely GPU bound, and all CPU tests are purely CPU bound (with the caveat that older CPU tests do not scale past 8 cores very well. TS Extreme CPU test does scale, as does the CPU Profile test)
im on 6400 cl30 2133fclk and have latency 66, when i go 8000 cl 36 2000fclk i get 65, sometimes 64 if i tune it tighter so idk about these results. BUT i am on A die 32gb Also, tightening timings for gaming increases lows not the max or average that much, less dips=better
Hi i think yo not have eneabled suport for high bandwitch becasue my 8000mt 40 48 48 127 1.35V ram have over 63 000+ in aida and latency 66ns (in normal windows boot) Aorus x870e 9800x3d fabric on 2000. I try 6400 cs30 6000 cl 28 but best performance i have on 8000 40 anyway thx for test and you work
when I can get my hands on the CUDIMM 9600 I am gonna shoot for the moon with that. but initially I tried to OC my 8000 to 8800 and it was gonna take higher voltage than I wanted. I have RAM FANs in the mail maybe once I get them I will see with cooling if I can get it up there.
Wow on the game testing you dropped the fabric clock speed to make the 32 gig look really bad on purpose. Why didn't you keep the 22000 FCLS? Worse testing I've ever seen I'll wait for gamer nexus to make a proper and honest video.
I originally ran all the games at their 2:3 ratio and expo / Xmp settings, then I started branching out a bit to get a direction that I want to dedicate my time to, I've seen alot of back and forth on fclk freq and latency claims and I'm driving myself towards the setup I want and sharing my findings. By all means I enjoy those channels too. I'm just a dude playing around with hardware and sharing what I find. Sorry if you thought I was trying to skew the numbers.
@@JohnnyRage303 I can honesty respect that answer and overall it was a good video until you didn't post all the data halfway thru. It seams you are trying to get people to buy more ram than needed. Your running an expensive motherboard and the 80000 RAM is also really expensive. But the first half was very informative. I'll try to adjust my timing. Keep going I know you will improve with every video you make good job man.
Dude sorry but u sont eeally know what ur talking about..u havent dine ebough testing with enough ram..and u never explaij how u got the cl with tuning..really unproductive video
Hey thanks for watching. My main goal was to really compare the xmp performance because I see alot of people just parroting what worked on the 7800x3d and i think a majority of users just enable xmp and call it a day. I do think the new generation has a bit more capability. I started to play a little bit with the primary timings just to get an idea of where I should focus in the future. Probably was a little disjointed in hindsight. What would you like to see different?
different games =different needs data is used different ly across games all dev programming so one does better here and another does better over there its all meh
hey, I'm also currently using the gskill royal neo cl28 32gb Is it normal for a RAM to glow a different color of blue at the top with RGB. I changed the slots of my Asus x870e Crosshair but the RGB lights are buggy. If I unplug the whole thing and restart it, the RGB is clean. It bugs again when playing. I updated everything in the bios and windows
Can you please do a video on how you tuned the 64GB 6000 CL28?
Seconding this. I'm building a new AM5 PC with 64gb of g.skill 6000 CL30 RAM and hopefully finding a 9800X3D if they'll every stay in stock for more than 5 seconds. Would love any tips specifically on tuning that RAM I can get.
You can check if your bios has memory try it in the OC- DRAM section, and try what boots, then what is stable through benchmarks
seconding this, would love to see a video on this
You should consider putting some affiliate links in the description you deserve any compensation if people are gonna go out and buy some products based on your findings
it's simple
aim for hynix A-die or m-die.
Adjust timinigs
always stay on 1:1
primary timings not so important, but secondary are!
well said friend
8000 1:2 with fclk 2000 for gaming is better because you are synced but need a good mobo and also not every cpu will do 8000
@@xpatrikpvp too much work and you gotta spend serious cash on it and get lucky too
@@SomeoneElse6282bought a 7200 2x16GB teamgroup kit, overclocked to 8000 and 8200 cl34. Got it for $110 on a slight sale.
Can probably go higher, haven’t tested yet.
9800x3d, MSI X870-P wifi
@@kurtsaidwhatIs it stable in OCCT?
I got the GSkill Ripjaws S5. It's lower profile without RGB. Using the D15 G2 cooler. Went with the 64gb 6400 cl 32 kit with my new 9800X3D. Running the Asrock X870E Nova motherboard.
This memory ran fine with the EXPO profile at the default 1:2 ratio.. But when I set it back to 1:1, I had to dial it down to 6200 to get it stable. I was then able to tighten the timings from 32-38-38 to 30-37-37. I tried 2200 on the IF clock, but went down to 2133 in the process of trying to get the system stable again. I need to see if I can get it up to 2200 I guess.
I haven't gone into the sub timings yet, but I want to. A detailed guide would be very welcome! Any tips for getting 6400 stable would be welcome. I did try raising the memory controller voltage to 1.25v and that got me a couple minutes of game time before crashing, but still unusable. Maybe I just need to suck it up and go higher. I had that 1.4 is safe, but I don't want to stress the chip for gains I won't t really feel in real life. I'm using a 3090FE at real world resolution and graphics settings. The GPU is definitely the bottleneck right now.
Same here.
I have adata xpg lancer 64 gb 6400 cl32 hynix a dual rank kit. My mobo Asus rog x870A. I couldnt get Windows 6400 1 to 1 mode. I run at 6200 cl30 tuned sub and secondary timings with buildzoid. Fclk 2066 ( buildzoid says 6200 much stable at 2066 fclk ) . Im happy right now.
some times you need to mess around with the SOC voltage for IF, its a balance between ram speed and IF. sounds like a great build though. it takes some playing, I have been trying out so many different configs on this system I haven't fully figure out what my final settings will be yet.
if its stable its good, seems like whatever you can get with low latency wins. There is no magic number, alot of different ways to win.
@@JohnnyRage303 I tried 1.27 vsoc with 6400 1 to 1 mode. It didnt work. Probably if i give it vsoc at 1.30volts it will work. But i dont wanna go higher vsoc above 1.25. Rightnow 6200 1 to 1 vsoc at 1.25 im stable with 2066 fclk. I am lucky to ask buildzod on streaming my timings. And i tuned a lot my timings. A data gives me 67.00 ns latency 66.000mb read 92.000 write.. I am happy actuually.
@@eyluloktay9104 Looks like your are having a very similar experience. Last I checked, Buildzoid didn't have a memory guide newer than 7800X3D? But his info looks very extensive. Maybe over the holiday weekend I'll get into the sub-timings..
I also haven't touched PBO yet. I don't know where AMD is actually standing with warranty stuff. I'm too worried, but there's no reason to take any chances with this chip, when there is so little to gain.
I don't know a thing about RAM timings and tuning them. I simply check qvl on mobo list, plug the sticks in, enable xmp and pray for the best. Luckily, my 64gb 6400mt/s cl32 ram kit works great with xmp enabled with my 9800x3d. Couldn't even tell you if it's running 1:1 or not since i have no idea what that means. Using a b650m aorus elite ax board which didn't even list my RAM on the qvl.
With 3d cache you might not even know when you have some errors.
The thing with RAM errors - it often breaks something else along the way. I had frequent GPU driver crashes, tried everything i could imagine until i ran another RAM test and caught a few errors. Tweaked some timings and no more GPU crashes.
My comment didn’t age well. The 64gb 6400mt/s cl32 kit causes intermittent issues with booting up resulting in bsod.
I am guessing because this is an xmp kit that could be why. Gskill sells a expo ram kit with the same timings but only up to 48gb and it runs at 1.35v instead of 1.4v
@@e92justin It's not that your comment didn't age well. The way you described how you configure your computer is how 99.99% of people should. It's on the various manufacturers to ensure that the QVL is true. It should just work and if it doesn't they failed. I consider myself an enthusiast, but at the end of the day, I just want things to work without any fuss and expect to accomplish that by, in this context, simply enabling something like XMP or EXPO.
PC Specs:
9800x3d
X870E Crosshair Hero
RTX 4080 super Zotac
Memory Kits:
G.SKILL Trident Z5 Neo RGB Series AMD EXPO 32GB
DDR5-6000 CL30
Ok so i originally had a MSI Tomahawk x670e MB inside. i was getting Post issues during re-starts ONLY. switched to the Asus x870e crosshair hero, seems like the problem improved BUT its still present. using expo on occasion system wont post from a RESTART, cold starts are usually fine. Having a Few USB issues also. system seems to hang on things like keyboard/mouse. its weird, because otherwise things KILLING benchmark's. any help?
reading the forums etc this issue seems to be VERY wide spread, posting issues specifically from reboot's/re-starts! whats the issue here guy's? AM5? AMD? FEELING LIKE 3.7K WAS WASTED! help!
Really need some help with this kids xmas gonna get ruined ;(
4 9800x3d's later STILL GETTIng CORE ERRORS! HELP PC HAS BEEN FULLY REBUILT 4 TIMES!
i'm guessing with the 64gb vs 32 gb kit thing. is the 64gb kit is dual sided so it's basically it looks like a 4 stick kit to the cpu. like back in some of the previous gens they were getting better performance with 4 sticks vs 2
I've heard that Star Citizen is supposed to be abnormally impacted by memory. Would have been interesting to see that one included in a test like this.
Several people have brought this up I'll try and take a look at it.
rule of thumb for the fclk is that its good to sync it in 2 to 1 mode when youre running 8000 Mts+ ram. Anything below just get the fclk as high as it can go, same for the ram.
please show your sub timings dude this what people want from an enthusiast channel.
Noticed XPO/XMP settings kept boot looping but manually OC Ram was the way to go and stable. Interesting having to use the OG way to get new tech working
That's wierd I have had expo and xmp work on all my am5, after the initial Ling start for memory training. Usually I just find you can improve the timings alot from expo xmp. But either way stable is stable.
There's a whole new marketing opportunity opening-up here. We could see memory advertised as "runs best with ...". Potentially COD memory kits could be a thing lol.
Tried 8000 cl40 expo on a x870e godlike from a QVL set and it wouldn't even post. Looks like I won't have any real performance issues just sticking with 6000 cl30 though.
Thank you for the job!
I'm literally doing the same testing today. 6000 cl28 vs 6400 cl32 are the two kits I have & are testing
Keep us posted
So I came to the conclusion that the 6400 cl32 kit I have is about 1.25% faster overall throughout my testing then my 6000 CL28 I bought to compare it to. I ONLY used EXPO timings and did not do any tuning. I also game at 4k so I was already against the GPU BN in MOST cases.
I would try to force the cpu bottleneck because if you're seeing that difference at 4k i bet it's higher while the cpu has a higher load.. However if your real world application is 4k at the end of the day it is what it is.
@@TheRealDlo Higher mclk with good timings is the way to go even over a lower mclk with excellent timings. Some games and applications are faster with lower latency but overall higher frequency is better
That's why I went with a CL 28 6400MT/s 64 GB kit. It's even stable when over clocked to 6700MT/s. I'm on a x870 motherboard.
I bought a kit Corsair 64 GB 6600CL32 to pair it with the 7800X3D on an Aorus B650 Elite AX.
My goal is to have better on Star Citizen than with my current Lexar 32GB 6000 CL30
I hope that the new feature at Gigabyte of "Turbo BOOST X3D" will allow an improvement in performance.
Otherwise, I will try to decrease the timings to have a kit of 6400CL30 or 6200CL28 Cross my fingers
People are bringing up star citizen alot. I've never played it maybe I'll take a look is there a easy way to benchmark in it?
@@JohnnyRage303 Being an Alpha and depending a lot on the performance "server", I think this game is complicated to bench. But with the imminent release of 4.0 it will become much more stable on the server side and I think there will be an interest to do comparative tests. That being said, the game needs 36 GB to not be "pinned", it is a fact already demonstrated. Otherwise, to answer more precisely, walking and doing the same route in the city of Lorville is perfect for making benchs
@@NEIHERUniversStarCitizen This game will quickly become the standard of measurement for "games requiring a lot of resources in CPU and DDR" in the future. Very soon, if it’s not already the case If the performances are good on this title, then all other current games will pass the bar of performance "more than acceptable", like CyberPunk2077
6400c30 is extremely solid as is, especially with tuned timings!
I've made some tests while fiddling with the memory timings and I can absolutely tell that a nicely tuned memory can outrun any of the shown examples. While 6000 cl28 seems good enough, I'd argue that 6400 cl30 FCLK 2133 is a real sweet spot, at least for X3D chips that still have higher CPU-to-memory latency
Can 6400 C30 - 2133 FCLK be doable on a M-die 64GB dual rank? That's what I have but I dunno what I can take it to on a 7000 series while trying to maintain a v1.4 on the ram.
My team group X2 16GB (32GB) 8000MT/s kit in AIDA64 reads 63,805 MB/s with a latency of 59.9 ns, using tweaked timings. In games, I believe latency is more important than bandwidth. FCLK 2000 and UCLK 2000. Ram 38-46-38-48
I rum 9800X3D on 5.5Ghz@1.250V ,Kinston (Hynix) 32GB 6400Mt/s on CL28 1:1 on 1.480V (with fan top of stiks because if the ram reach 50C crash) and FCLK 2133 @ 1.260V. My latency is 66ns (all stock was around 90ns,stock CPU with only EXPO on 6400 was 80ns) I play alot to run FCLK on 2200 because the latency will drop to 62-63ns but unfortunately the system is not stable
The higher your soc voltage, the harder it is to stabilize fclk. But soc stabilizes memory frequency. So it's a balancing act. But your setup sounds awesome. Nice settings.
@@JohnnyRage303 Yep but there have no extra FPS from the CPU overclock. Only the ram overclock give more fps in 1% low.
Good evening, were you able to use how much curve on your processor?
Check out my other 9800x3d vid it has it all in there
thx for the info, good video!
Appreciate it, glad you liked it!
is there something wrong with corsair cl30 6000 kits? i have a 32gb vengeance rgb kit and looking at 64gb kits for Tarkov, and i like those dominator platinums, but i rarely see them in this kind of videos. is there a reason, except for price?
There's nothing wrong with corsair many times there xmp profiles are a little more aggressive which is good. You probably don't see them as much because they are more expensive and if you are tuning the kit anyways the brand doesn't matter as much. They are all the same hynix a or m die underneath.
@ Thanks for the reply, thought I missed something about them. How can I tell if it’s a M or A die?
I'm definitely interested in your 9950x results!
Thank you for sharing 😊 are you using nitro setting for 6400?
Man I am having a hard time with 9800 2x32 on a x870e strix. I’ve only gotten loose 6000 even close to remotely stable. Anything else either errors right away or hangs on a 2f or 46 I think. Pretty frustrating
I have 64 gb 6400 mhz cl32 kit. I couldnt run that 6400 1 to 1 mode. So i watched buildzoid videos and asked him my secondary timings. I tweaked a lot.
With my 9800x3d im running at 6200 cl30 fclk 2066 with tweaked secondary timings.
Pbo +75 total 5.3ghz
I have cinebench r23 multicore 24960 score.
Time spy cpu score 17440
I have 4080 super and with the shadow of the tomb raider utlra performance dlls im getting 406 avarage fps.
is it more likely that 6400 cl 32 or 6000 cl 28 will run natively 1:1?
i recognize your name from buildzoid's live vids. i remember thinking your name sounds possibly turkish
@@JamesC1981 Yup. I am turk. And i watch buildzoid streams.
Great idea, what is the rank of modules tested? I think the 64GB modules may have an edge thanks to being dual rank. The fabric is a big bottleneck of am5. Same read speed for ddr5 6000 and 8000 are crazy.
I remember with AM4, dual rank helps gaming performance even though other benchmarks like AIDA64 show no performance difference. Seems to still be true.
With my 7950X-3D the G.Skill Ripjaw S5 sticks have worked great. A 32 gig kit of 6400mhz cl 32, running on a B650I Gigabyte board.
6200 CL28 1.435V with very tight timings and DDR Nitro settings with 2200 FCLK completely stable (OCCT, Y-Cruncher, 24 hours in Stalker 2 no issues)
AIDA Read: 70190 Latency 64.8
Can run 6400 but with nowhere as good timings, 6200 tight gives me best results.
FCLK needs proper testing to see if stable or you will run into issues. Benchmate Y-Cruncher 0.8.5 PI-2.5b will quickly pick up issues, then longer tests can be run.
2066 for 6200 is better than 2100 and maybe 2133 but 2166 and 2200 overtake that advantage.
Could NOT get FCLK stable higher than 2066 until raising 'VDD MISC' Voltage to 1.25V
ASUS ROG STRIX X670E-F GAMING WIFI
32gb 2x16 Corsair Vengeance 32GB CL30 1.4V sticks
Thanks for all of that info, this video was really just me trying to find what direction I wanted to go. I'm gonna get into tuning more in depth soon. This is some good additional info to steer my setup.
Thanks for this well done video. Any thoughts on how fast you could get 192gb ram to max out at?
I've only played around with 2 stick kits. I know with 4 you can't clock as high. I thought last gen with some of the bios updates people had gotten 6000 stable but I could be wrong.
I just picked up some 32g 6400 sticks. Wonder how those would do.
Just got my 9800X3D today and it memory overclocks better than my 7950X3D on my X670E Carbon. I have a 7800 CL38 XMP Adie Gskill kit that is on most of the AM5 mobos QVL but XMP or tuned didn't work on my 7950X3D at the rated speed. XMP and tuned work at 7800 and 8000 (haven't gone higher because I figure it won't work) on my 9800X3D. My 7950X3D was able to do tight timing 6200/2067, but I feel like this 9800X3D might be able to do 6400 which I will try soon. Pretty stoked to have 8000/2000 working with tuned timings just because I think its fun and low SOC voltage is dope.
I would love to see you tweak the ddr5 8000 kit for better timings. Memory bandwidth and latency might not matter that much for gaming but it does for AI inference from system memory!
My main goal was to look at the xmp kits and I did a little bit of primary settings to see what direction I wanted to move in the future. I am going to dive more into tuning these kits further in the future.
the best kits to buy is 2x16 8000 c38 because they are the best bins and always hynix a die and then punch in the timings yourself, if you don't want to tune it yourself and only enable xmp you should at least put trefi to 65000, be careful tho you should put a fan over the ram of it gets too hot because of the high trefi
i saw a screen of oced 8000 mhz kit to 8800mhz, fclk 2200 uclk 2200 mclk 4400 !!!
Thats nuts do you have a link? I might try a max effort speed build see if it anywhere near stable.
@@JohnnyRage303 google for "AMD Ryzen 7 9800X3D 3D V-Cache CPU Hits DDR5-8800 Memory Speed on ASUS ROG Crosshair X870E HERO With NitroPath Tech
"
So to have all stable i can just go for a cl28 6000mhz? heard a lot about higher frequenzies make problems on 9800x3d
What kit is the 64 6000 CL28 you have listed in the chart? It's not in the description.
The 3 kits listed are the 3 xmp or expo kits used. I changed the timings to get the 6000 cl28. It was not a factory sold setting.
So MCLK/UCLK synchronisation with FCLK doesn't seem to be optimal with the 9800x3D in gaming... that is unfortunate for those with OCD, I guess, but fortunate for the rest of us. Pushing the IF above 1:3 in 1:1 and getting better results isn't as surprising to me than breaking 2:1 synchronisation... I guess for applications beyond gaming and on non-3D processes sticking with 8000:2000 and tightening the timings might work out better though.
They advertised it was desynced since zen4. Most people go for the ratio because at 6000cl30 it works well
What do you think about the cl28 kit form gskill that's at 6000mhz? the neo royal one. Just wondering since you didn't review it in this video. It seems to me in this video lower latency is better but, just curious.
I just saw those were released, amd does love low latency. I had the royal kit on am4 and was happy with the product. I haven't tried them in ddr5 yet. Usually it's their top line so I would bet they are binned pretty well.
tCL doesn't impact performance much so you're better off getting a 6000c30 kit for like half the price. As long as it's hynix you're good to go.
@@IIHydraIIand How would I check if it's hynix?
@@carlosferjezus Any 6000c30/c28 or 7200+ kit should be Hynix.
@@carlosferjezusmost motherboard QVLs will specify the chip... if yours doesn't just look up the same kit number on another brand's QVL.
@johnnyrage just curious how come you didn't set the fclk to 2200mhz on the 32GB kit? Stability issues or no? I have the same 32GB kit you showed in the video and am running it @2200mhz getting 69k MB/s read, 92k MB/s write and 74ns latency. The only thing I had to do was set my soc voltage to 1.15 to get it stable after stress testing. I'm running a 9800x3d on a ASUS x670e MB.
Honestly I ran the 32gb kit at 2200 for the Aida and cinebench tests but was leaving the house for Thanksgiving. And most the feedback I've had people seemed to think I was crazy running 2200 not 2:3, so I limited the full gaming runs I completed mostly just due to time. Seems like on this video alot of people are more curious about the 2200 so when I get home I might do some more tests.
My tuned 64GB 6400 hits under 60ns in AIDA with a 7950x3d but that's with a 2dimm ITX board
Just show the ZenTiming screenshots for these kits, please!
If you raise the frequency of the Infiniti Fabric to 2200 MHz, is there a possibility of degradation after 2-3 years of use?
Interesting video, any chance you can run the FF XIV Dawntrail Benchmark with the 720p laptop standard preset with these kits?
Thank you.
What is the best 64gb 6000 cl30 kit in white?
I wonder if the larger memory chips help reduce the number of rows involved in the majority of read commands? Overall it might reduce the number of clock cycles per command required. An uneducated guess anyway haha
Im trying to understand why youre comparing a 6400 tuned 3 different ways to a 6000 and 8000 tuned one specific way obviously not at their peak
My first main goal was to just compare the xmp /expo performance for people that just want to plug and play. The second i was just playing around with settings as I went. the 64gb 6400mhz kit was the most flexible, my 32gb 6000 kit is the oldest and would regularly not post as i tried to add performance (thats the silicon lottery) I was unable to run as many profiles on it. The 8000mhz kit I could have explored more and I will in the future, I'm also finding these high end kits like high voltage you can push them as far as you are comfortable. But simply the 6400 got the most attention because that kit was very flexible in the time I was filming. I'm testing these kits again on my 9950x as well. I'm trying to keep it a little more standard since I've got that feedback.
@JohnnyRage303 thanks for the honest reply. Yes I and others are curious on the 8000 since AMD is flaunting it lately.
May be you will be lucky to find 2 48 gb sticks and play with it.
is expo 1 or 2 better should we let motherboard tweak it or completely ram profile?
Expo 1 is usually the faster profile. You want to at least enable that. You can find a little more performance with tuning. But it's up to you if you want to research that.
@ faster yeah but it runs fclk at 2200 for example if i run expo 2 i reach max ghz more often and fclk 2100
temp wise no difference
I am very likely an outlier, but I only play iRacing. I have a 9800X3D, Carbon 870 E Carbon. Not built yet. What memory do you recommend? Thank you.
Any 6000c30 kit for minimal tweaking 👍
So the G.skills 6000 64gb CL28 you are using is 4 dimms of 16gb correct??? Because g skills own site as well as Newegg and Amazon all say that G.skills doesn’t make 6000, 64gb, 2x32gb dimms. Please confirm?
a.co/d/28kIuhk
G.skill 2x32gb 6400cl32, was the kit I used
Yes they do I got a pair Neo trident z
@ do you have a link, because g.skills own site says they don’t. I would love to get the same ram you have
What about 4 sticks?
Never run 4, it's a nightmare to stabilise.
2200IF is King GG
if you can get there :/ mine does 2133 max stable. I can get 2167/2200, but the amount of tweaking voltages to get that stable there is 0 room to o.c. ram. Undervolting soc = flck stability, but lowers ram stability. 2133 with ram @6400mt/s is about where I'm fine setting the scales.
I recently upgraded to 9800x3d with msi x870 tomah.Atm im with gskills hynix-a 6400 with 2200IF!!!Man this is crazy really!
@@Dante_S550_Turbo i haven’t been able to run 6400 below vsoc at 1.25. Most of the people can easily run at 6400 at 1.20 vsoc. So for stable 6400 i need to run at 1.28 vsoc at least. At this high vsoc voltage fclk become instability. So i ended up with 6200 tighter sub and secondary timings. With lover vsoc at 1.16. And fclk 2200. It is rock solid stable. No vt3 errors no karhu errors.
@@eyluloktay9104I helped someone get to 1.22 vsoc @ 6400. It really depends on the chip
@@NoSkillsBro.Which ram did you buy?
Corsair CMK48GX5M2B6000Z30 is 48GB and has 6000 30-36-36-76 on the expo profile. Kind of a weird one.
I found a few cases where my framerate increased when I went from 32 to 64 on my current system. star citizen and MSFS both gave me a significant framerate increase weirdly. I got the 48gb kit on my dresser my 9800x3d is in the mail. so new system coming up.
You should show cl34 or cl36 8000mhz ram with tight timings. With my own cl34 M-die configuration, i can reach cl-34-44-44-80. I got the read speed of average 93K. FLCK 2200
I really hate to be that guy but tuning the sub timings isn’t that hard and without doing that this becomes a pretty misleading video. Title should be “if you are unwilling to work at your memory these are your best options” or something like that.
8000cl36 (fclk=2000) or 7600cl34 (fclk=1900) AIDA latency ends up around 5ns faster than 6000cl28 or 6400cl30 with about the same bandwidth on a single ccd CPU. Seems like all the 9000 series CPUs will do ~8200 with no problem and some will push 8400+
A big warning for x870e Asrock mobo users G.Skill even on the qvl list are giving Asrock mobo users major issues including not even booting up at all. Then the expo isn't stable if you are lucky enough to have a g skill kit that boots up. Some have reported weird glitches with voltage management. It'ssuppose to run at 1.1v/1.4v depending on default and xmp. Actually running on your system @ .9v/1.2v which is unstable and possibly the reason why they aren't working well.
I have the 6400MTs cl30 kit but I haven't tried these settings. I just set mine to 2133, I haven't tried either 6000 and cl28
look up buildzoids 6000 tight timings. it works for most people. Dropped my latency by 10ns in latency.
Has anyone have any experience with the CORSAIR VENGEANCE RAM
I have a CL30 32GB 6000 on my recent 9800x3d build and I've noticed some lock ups when restarting the PC randomly.
Should I return in and go for the G.SKILL Trident Z5 Neo RGB Series 64GB 6000 with expo?
Corsair vs g.skill corsair usually run slightly tighter timings at higher voltage. My example I have a corsair 6000cl30 that is 1.4v xmp with tighter timings and a g.skill 6000cl30 1.35 with slightly looser timings. If you're going to tune either brand will be fine. Just running xmp corsair might have a slight advantage, probably unpreceptable. I haven't had lockups with either.
Not sure what to think about The size difference. Would 64gb worth it over 32 to to gain Performance in Gaming?
having more ram does not gain you any performance what it allows you to do is have more ram. some games like star citizen wont run on 32GB for example and it should be a good example of whats to come in the future.
do u game with google open ,,,a few tabs
watch youtube or stream at the same time?
then you need more ram
honestly i linked 2 videos the other day in 2 other toutube channels about ram
a game like star citizen uses 27-34 gigs ram on 1440p
and uses 44-56 gigs on 4k
this all depends on areas
however star citizen is not even complete yet
when all the players load in and all the AI npcs i think its gonna be even worse
as off now you may see 1-4 people running around with about less the 10 AI npcs in city sections
they still want to add live AI that walks around ..does shopping and flying there own spacships
their aim is to make it look like real life
many people and npcs in areas
to look lively
like new york streets
no idea if they will ever get it right
just dont go less then 32gigs at the moment
games are growing hungry
and game worlds and data keeps increasing
stalker 2 uses 22+ gigs at 1440p
and u need space to run windows also
windows 10 and 11 can compress..to about a small 4 gigs in size on ssd or nvme ..but still needs 4-9 gigs to run itself
so dont go below 32 gigs
48 is better as it is A-die memory on ddr5
and 64 is also good..a little better future prooffing
vram for gpus
on stalker 2 ....8 gigs makes the game not run
and so look at 12 gig vram on gpus and up..to safe yourself
Generally yes. But some of the results were influcend not only by the speed, but also by the capacity. Did I understand him wrong on this?
his 64 gb ram is dual rank right?
@@QuaK3RRR 64gb DDR5 should always be dual rank afaik
Dude, read/write should be at 100k or so Aida64 on the 8000 kit. You have seriously messed something up.
Thats at the xmp settings. On amd that's what it's getting. Intel it would probably be much higher.
I'm only getting around 1050 cinebench scores.
Have you done any pbo tuning? That seems low compared to even factory reviews I've seen.
Can you do a video with all channels filled
I just got 2x32 6400 cl32 Patriot Viper Venom and for the price it was a great deal. Less than 2x of the cheapest 2x16 6000 cl30 kit
How is it going im thinking of getting them aswell
@@TheShaunydog Going to build the whole PC after Nvidia 5000 launches so I got no idea but it's Hynix A-die 2x32 so it's really good for the price
@@themarketgardener ahhh true, im doing same building new pc. Not sure if ill end up with a 5000 series card though, i feel like better to wait for super/ti versions and thats too long for me
i have same board and the same ram 6000 cl30, do you not have issues when putting expo? my pc refuses to boot with expo, i have to disable memory context restore for it to work and it takes ages to boot
No I've used the 32gb kit in a msi x670e, asrock b650e, and now the asus x870e. It's ran in expo on all of them. First boot is slow then it's trained and boots fast.
@ pretty odd, i have the same asus, 9800x3d and the same exact ram and without memory context restore disabled it wont boot at all.. do you have a bios beta version perhaps?
68ns thats high… go for low 40s, even your main timings are bad 🙈 you will see huge improvement with mem tweaking. fclk tuning is not so great on this chips. its good on the G chips. next year Ryzen Max+ will come out, and than the mem tweaking will be realy cool
Helo nice review but tell me x870e hero how is it ..boot time is laggy or fast...in x670e is 40 50 second...twll me how it is? Thanks
All am5 systems have a long initial memory training when you change your configuration but after that it boots quick.
@JohnnyRage303 hmm memory training is evry time when push to start pc? To boot in windows?!
Highly recommend the new Indiana Jones or Space Marine 2 as extremely CPU heavy titles.
i bought a corsair kit 16x2 6000CL30 should i go for 64 gigabytes or is it even worth it
Has anyone tried high speed ram test on Linux to see if it's a windows thing?
i have a 7900x but i got 6000 c36 xmp ram, is it okay for me to enable xmp? rn i have 5200mhz c36 timings
Yeah I would always at least enable xmp or expo
@@JohnnyRage303 xmp is stable on am5?
so witch one is the high end?
All 3 of these kits are "high end" I would say. I'm personally gonna pair the 64gb cl 6400 long term with my 9800x3d
@@JohnnyRage303 could please just give me the link the high end and expansive one :D
are u gpu limited most of the time?
at 4k it defo is
I just purchased this cpu with x870e aorus pro ice and 6000 mhz 2 sticks rams and CL 30 but dam everytime i restarting my pc led for RAM goes on ,after restarting will boot but the main issue is stuttering on desktop i cant even play games , the gpu is up to 99% utilisation at max freq , wtf is happening, i had intel before but I haven't reinstalled windows just uninstalling the intel chipset and all others, no EXPO enabled, the rams are on the QLV list of the mobo !!! It passed 2 day , i am going crazy......
1. Could you go to the Windows Device Manager and find out if the Display Adapter works properly?
2. Go to windows defender and check if memory integrity in on, if it isn't then look what driver is causing it to be off.
3. Try to boot into windows safe mode, then check the GPU utilization, if it isn't 99% anymore in desktop then it means you might have accidentally installed crypto miner
4. In the command prompt window, type SFC /scannow. Let it run and then reboot computer
5. It's highly recommended to take the bullet and nuclear option and just clean install the windows.
@ I found the issue, fir the stuttering was the rizer cable which is 3.0 and i have tested the ram modules and they are ok but still i have 0D and ram error code and when i have this error after i am restarting the pc the mobo its reading only one ram module i have to turn off to get her to read both sticks and i think could be because of the ssd which is 3.0 and i see on the motherboard diagram that there is no pcie 3.0 for the slot, they communicate directly with cpu through pcie 5.0/4.0 , both my ssd are 3.0 and one it's connected to chipset through 4.0 connection and the other one share pci x16 with gpu but the pcie x16 i see its only 5.0/4.0 , and the first ssd slot which is directly connected to cpu its empty, i have tried with only one ssd connected to chipset (last ssd connector ) and it seems tha was no 0D error or ram , in conclusion i am not sure if the ssd its the cause but definitely the rizer cable was one , i will buy a new 5.0/4.0 one to see if that its the issue
I have had problems in the past with this. In the bios PCIE channels are set to AUTO by default. you can manually set them to 3.0 / 4.0 / 5.0 per channel when you are running a bunch of different devices like this sometimes it can help to manually set them by device rather than leave in auto. But ultimately for me that was a temp fix until I just upgraded everything to PCIE 4.0 or higher.
@ The stuttering was because of the rizer cable which was 3.0 , i am going to order a 4.0 but the 0D and ram error still there , if i am not restarting the pc everything works fine
Are u using xmp ddr5 on an amd motherboard is that even possible.
I thought this is not working.
That’s why I ordered the gskill royal neo cl28 32gb
Xmp will work. It's enabled in bios via d.o.c.p. (direct overclock profile)
For future cs2 benchmarks plz use the fps benchmark workshop map 🙏
I'll take a look at it thanks!
@@JohnnyRage303it's the most played game on Steam. Most competitive eSport ever.
Test at 1080p. Do 3 runs at high to get that average, 3 runs at all low settings and get that average.
Competitive players really really want to have the best hardware possible.
Hi, i bought gskill 2 x32gb 6000 cl30 36 36 memory kit, i have a 9800x3d and strix x870e-e, I'm a casual gamer, should i just set everything on expo and im good? I don't wanna tinker so many things for small benefits.. But If it gains so much i could do... Can someone help?
It's a personal choice the expo profiles are huge over the base frequencies. You can always find a little more performance tuning. You just gotta decide if it's worth it to learn.
I have done cl28@6200 from cl36@8000 with my 7800x3d and really not a huge difference. So I just settled with 6400cl30😂😂 Gskill 2x24gb for sure.
Many people recommend using 6000mhz cl30 or cl32. but where I live the price of 6000mhz and 6400mhz have similar prices. Which do you think is better for me to take, and is 6400mhz plug and play?
@christiansalim Definitely cl32@6400 because most processors can do 6400.
@@CaptainCave-t3n is it plug and play or i need some tunning if i buy 6400mhz ?
@christiansalim Plug in play will be fine but tuning secondary timings will make it faster. There's many videos on TH-cam to show you.
@@CaptainCave-t3n oke thank u so much 🙏
aida latency test is very flawed. in most cases there si a latency penalty to not having the flck in that 3:2 ratio
is the 6400 running in a 1:1 mclk to uclk by default wheneer u go over 6000 its goes into 2:1
It's 1:1 3200mhz
@ nice. My cpu is almost stable at 6400. If I loosen out my timings it’s stable which is funny
I bought a 2x32GB G.Skill Trident Z DDR5-6400/CL32 kit, running on a Gigabyte AORUS Pro X with a 7950X3D.
I tightened the timings a bit, set FCLK to 2133, and left MCLK:UCLK at a 1:1 ratio.
Runs great. Read is around 80,000, and Latency is 78.3.
My system wouldn't boot with the default XMP profile and FLCK=2200. Be aware that only about 25% of AM5 CPU's can handle that FLCK=2200, while about 80% can handle 2133.
Ive seen a guy claim to have achieved 6400mhz cl28 with tje trident kit so im gonna try for that. If not that 6400 cl30 or 6200 cl28/6000 cl28
Worrying about 4K res on Time Spy CPU test is wrong. The CPU test is isolated not to care about resolution and rendering. If you look at just the CPU score and not overall, Extreme would be a better idea on any 8 core or higher core count CPU. For 8 core normal Time Spy still works (it stops scaling well past 8 cores). In general unless you are running a grossly wrong test (too light), all 3DMark graphics tests are purely GPU bound, and all CPU tests are purely CPU bound (with the caveat that older CPU tests do not scale past 8 cores very well. TS Extreme CPU test does scale, as does the CPU Profile test)
im getting the 9800x3d with the new meg god like with 64gb titanium 6000 cl 30 😁😁
im on 6400 cl30 2133fclk and have latency 66, when i go 8000 cl 36 2000fclk i get 65, sometimes 64 if i tune it tighter so idk about these results. BUT i am on A die 32gb
Also, tightening timings for gaming increases lows not the max or average that much, less dips=better
Hi i think yo not have eneabled suport for high bandwitch becasue my 8000mt 40 48 48 127 1.35V ram have over 63 000+ in aida and latency 66ns (in normal windows boot) Aorus x870e 9800x3d fabric on 2000. I try 6400 cs30 6000 cl 28 but best performance i have on 8000 40 anyway thx for test and you work
Try 4 sticks and get back to us. 😂
Should've tested this with pyprime for latency, Aida isn't the greatest benchmark
eh if using 9800x3d based on most videos i saw it’s really not worth it to go that deep into it 6000 cl 30 is fine
Hey jsyk theres a benchmark map on the workshop for cs2 thats all scripted to get rid of r2r variance
Rerun your tests with 1% lows in games that benefit from fast ram.
pls test 8400 mhz or even 8800 mhz :)
th-cam.com/video/WAFBjox2dEc/w-d-xo.htmlsi=2rxXdGoK7UqC_1iw
when I can get my hands on the CUDIMM 9600 I am gonna shoot for the moon with that. but initially I tried to OC my 8000 to 8800 and it was gonna take higher voltage than I wanted. I have RAM FANs in the mail maybe once I get them I will see with cooling if I can get it up there.
FYI, Counterstrike2 has a workshop benchmark to consistently test FPS.
Thanks I got it downloaded!
Wow on the game testing you dropped the fabric clock speed to make the 32 gig look really bad on purpose. Why didn't you keep the 22000 FCLS? Worse testing I've ever seen I'll wait for gamer nexus to make a proper and honest video.
I originally ran all the games at their 2:3 ratio and expo / Xmp settings, then I started branching out a bit to get a direction that I want to dedicate my time to, I've seen alot of back and forth on fclk freq and latency claims and I'm driving myself towards the setup I want and sharing my findings. By all means I enjoy those channels too. I'm just a dude playing around with hardware and sharing what I find. Sorry if you thought I was trying to skew the numbers.
@@JohnnyRage303 I can honesty respect that answer and overall it was a good video until you didn't post all the data halfway thru. It seams you are trying to get people to buy more ram than needed. Your running an expensive motherboard and the 80000 RAM is also really expensive. But the first half was very informative. I'll try to adjust my timing. Keep going I know you will improve with every video you make good job man.
Mfs benefits from 8000
I picked up that same Asus motherboard. My first and last Asus motherboard purchase. Pure trash. Grabbed a MSI Tomahawk and I have zero issues!
8400 is the best ram you can buy for am5
Dude sorry but u sont eeally know what ur talking about..u havent dine ebough testing with enough ram..and u never explaij how u got the cl with tuning..really unproductive video
Hey thanks for watching. My main goal was to really compare the xmp performance because I see alot of people just parroting what worked on the 7800x3d and i think a majority of users just enable xmp and call it a day. I do think the new generation has a bit more capability. I started to play a little bit with the primary timings just to get an idea of where I should focus in the future. Probably was a little disjointed in hindsight. What would you like to see different?
different games =different needs
data is used different ly across games
all dev programming
so one does better here and another does better over there
its all meh
hey, I'm also currently using the gskill royal neo cl28 32gb
Is it normal for a RAM to glow a different color of blue at the top with RGB. I changed the slots of my Asus x870e Crosshair but the RGB lights are buggy. If I unplug the whole thing and restart it, the RGB is clean. It bugs again when playing. I updated everything in the bios and windows
Did you run the 8000 kit on 8000? How come it score so low