I have 9800x3d with a 6000 CL30 2x32GB kit. I used your popular easy hynix ddr5 timtings. It now runs on 6200 CL26,36,36,28 with FCLK at 2200 and 1:1 Mode. VSOC is 1.2V So far it‘s stable and no programs are crashing. Did some stress testing for 30-60min but still have to to longer ones. Only thing my CPU struggles with is undervolting. Some cores won‘t to anything lower than -5/10, my max is -20. EDIT: Core 1 & 2 won’t undervolt with -10, had to decrease to -5. I had to increase SOC to 1.21, VDD Misc to 1.15 and it prime95 works for 2 hours as of now. If anyone is interested I will continue to update this comment and share my settings once I know it‘s stable enough.
@@Pullo83I‘m on Linux so no. I use the Prime95 equivalent and y-cruncher. For lower frequency workloads I use some games I know that struggled not to crash on more aggressive overclocks.
After the B850 boards have been released now, I'd love to see some 8000 tests with the 9800X3D on the ASRock Steel Legend/Riptide. They claim to support many 8000 Kits out of the box. :)
I've got a crappy 2x16gb set of Teamgroup 6000c30 running at 6400c28 1:1 2200fclk, nitro 1-2-0, pmu 0, swap apu, with a 9800x3d using an open box $200 X670E-F from Microcenter. I have a new kit of Gskill 8000c38 expo coming next week. I'm hoping I can do 6400c26 with the new kit.
I have a 2x24 gskill 8000 kit and the best I could achieve is 6400cl28 in 1:1 mode and 8000cl34 in 2:1. I could also kind of run it with 8400cl38 but it's not stable. Like I can play games but I can't pass some stress-tests, which means that if I'll play a game with similar loads I'll encounter some problems. But I also have the cheapest x670e mobo, asrock x670e pg lightning. I think it's worse than some good (or maybe even most of) b650e boards. Maybe if I'll get something like x870e taichi lite I'll get it stable but it's not guaranteed and also not worth it. if I knew for sure that it will work I would at least consider this option.
Hey BZ, nice vid. I just wanna say I had pretty much an identical experience with my 9800x3d. Random boot fails on mild settings, GDM causing errors, floppy error behavior, even basic expo profiles getting errors sometimes, etc. I suspect my motherboard (ASRock b650i lightning wifi) is the issue but this video is giving me second thoughts. Going to try with a different memory kit and mobo and report back.
Clear CMOS whenever stuff gets weird, failed FCLK overclocks often lead to FUBAR memory training and can only be fixed that way. With ASRock boards, disable "Memory Training Runtime Reduction". It's supposed to turn itself off when you input manual OC settings but it doesn't, causes partial (and failing) retrains when changing settings.
Same mobo here. I couldn't even run expo 6000 MHz. But 4800 MHz was fine. I checked with ASRock and in the process they asked for pictures of my RAM labels. This made me reinstall the RAM. Before taking them out, the RAM looked to be installed properly. Tabs were engaged and all. However, reinstalling allowed for EXPO. Strange! Guess the single tab system needs some more work
@@luckysuicidalalso sometimes(perhaps often) bios can corrupt so to solve that just reflash same bios via usb from bios, or even better reflash it via button on your mobo.
Hello, same for me, initially. Very floppy and inconsistent behavior. On some boots, it can run 20hours of Karhu, on other boots, error at 4 minutes. I change my EXPO kit corsair Platinum 64gb 6000 C30 to a new kit corsair Platinum 7200 c36 XMP overclocked to 6200 C28 with fclk 2200, works flawlessly even at 6400 C28. My 9800X3d scales better with M die single rank kit versus dual rank A die. I need vddq= vdd= 1,5 voltd for rock stability Vsoc 1,15 volts.
Can I request a video where you explain the training settings, nitro, all that stuff? There's not much info out there and I just don't really get it lol
I tired using the tightest settings because why not, MB stopped posting lol, I had to clear CMOS to get it posting again. Sadly as you said there are not any guides that explain each setting and what are the possible setup for different ram kits/timings. Good thing I used the profile saver so it was easy to get my previous settings back.
It doesn't need to be explained. You use 1-3-1 and call it the day unless you have a $500-1000+ motherboard and want to do days worth of testing. And if 1-3-1 doesn't work, you leave it on auto. Same with ProcCaDs and ProcDqDs and DramDqDs, those get set to 30Ω/34.3Ω/34Ω, period, the end. ASUS is the only one who got that right. MSI likes to jack DramDqDs up for no good reason. All that does is reduce stability with no gain. MSI also likes to set ProcODT too high. Usually it doesn't get set over 40Ω if even that. Almost always ProcODT is under 40Ω on Zen 3, 4 and 5. Usually in the 30Ω range.
Memory OC is probably among the most frustrating things. So many options, luck seems to be important and extremely difficult to tell where the stability problems originate from.
@@ActuallyHardcoreOverclocking OCUK are meant to be getting 1k+ stock this week, so you should be able to order either before the weekend or sometime next week.
Does his 111 SP score have anything to do with the problems he comes up against, I have the same setup with 114 SP. Also big thanks for these timings, have applied them successfully
8000/2000 is the best setup for my 9800X3D as far as I'm concerned. It was also extremely easy to setup. 1:1 didn't really provide benefits latency-wise and had some weird behavior when I tried tightening the timings as well.
Do you have settings for that? I am waiting on my 9800x3d but in the mean time I ordered some Patriot 7000cl32 ram kit. I guess these are top tier ram?
@@Alex-fo1rx I have a g skill a-die 7800CL36 XMP kit I'm running 8000/2000 at 1.05VSOC. The XMP settings for my kit are 1.45V VDD, VDDQ, and VDDIO which works perfectly on my X670E Carbon. Just for fun I've tried a lot of different timings ranging from CL38 to CL34 using BZ's recent 2:1 videos and had success. To start with, this video (G.skill 2x16GB DDR5-8000 CL38 EXPO optimized memory timings, MSI X870 Tomahawk) would be a good place to start but I don't have experience with Patriot memory. BZ has mentioned that Patriot kits seem to not behave well with Ryzen chips from his more recent experience with their kits for some reason. Before BZ had a 9800X3D I was using his 9700X because that's basically the same CPU in terms of mem oc compatibility outside of silicon lottery variance. Ideally, the XMP/EXPO profile and your motherboard will set the termination settings appropriately so you don't have to deal with that headache along with things like VDDP, etc. Would say in general see if 1:1 6200/6400 or 2:1 8000 works without a ton of hassle and choose whichever one you can stabilize and get the best performance from. With my setup, in memory latency benchmarks 8000/2000 is awesome and runs at low VSOC while being comparable in gaming (maybe better than 1:1 in my case since the latency is consistently better?) from the testing I've done swapping back and forth. Worth noting though that Ryzen 9000 in general seems to have higher latency values in things like AIDA than Ryzen 7000 (like my 7950X3D), but I don't really care about AIDA, I just use it for relative differences between settings along other tools like PyPrime which I think is better.
@@Alex-fo1rx Wrote out a decently long reply and for some reason YT didn't keep it I guess so that sucks and I don't want to write it out again. Basically 1.05VSOC, 1.45 VDD, VDDQ, VDDIO since my 7800 CL36 gskill XMP a-die kit uses those settings. For timings just start with BZs 8000 CL38 settings from recent videos. BZ has said that Patriot kits he's tested on Ryzen lately haven't worked well for some reason.
If your IMC can do 8000, it’s very easy to setup. If it can’t, it just won’t no matter how hard you try. My 7950X3D can do 6400 at VSOC 1.1V but it just won’t post at 8000.
Try to redeem this chip in another board: There was a rash of issues on this board and the Strix version with this bios breaking some EXPO profiles that previously worked. I personalty had different N63 performance depending on if I set EXPO in the Tweaker tab vs doing everything in the hex menus manually, but that could have been due to fabric error correction. (sharp waves in CPU usage, 1-2 second period) If I didn't touch the Ai Tweaker tab, which was the only place to configure XMP/EXPO, that test was steady. I don't have that x870e Strix MB anymore. If its doing the same thing then I guess we both got weird parts.
I have the same memory kit as you used in this video and mine can do 6400mhz cl28 with tight timings no problem, even cl26 if I had a RAM fan for temps, so I would definitely suspect the CPU as the cause of your pain :(. It hurts to lose to the silicon lottery, I know that feeling all too well from my Intel days...
I need on the same motherboard and CPU configuration with Corsair 2x48GB modules at 6000MT CL30 - 1.5V on VDD & VDDQ (SoC at 1.2V) BIOSv0706.. something messed up with AGESA or BIOS because i can run them at same speed with BIOSv0606 at 1.4V VDD & VDDQ (Soc at 1.1V)
I'm working on a 8347mt setup right now. Can do 8400 but not 1:1 fclk. So went down to 99.5bclk and it's stable so far hour of ycruncher. I'm gonna do a longer test tomorrow but it could be cool.
did you try adjusting VDDGs? should easily work at 950/1000mv I'd be very surprised if a chip couldn't do 2100 fclk at all, certain boards just default vddgs way too low in 2:1
dont use 1:1 fclk term ... you can't sync the fclk (it's never going to be 1:1 ... its more 1:4, 8400/4=2100 flck) ... because 1:1 is the ratio between MCLK and UCLK :)
That instability might be due to IF overclock. My 9800X3D can do 2133 IF with 6400 MT/s memory in 1:1 at 1.25v SOC, and runs mostly fine, for hours, but occasionally starts failing out of nowhere. Reducing the IF clock to 2000 is stable from what I can tell. Giving more voltage to SOC is not good long term, so I made peace with having IF at stock.
9:58-10:59 Maybe you have to adjust the ProcODT's. My 2x32 M-die ran fine at 6000 with my 7600X, but with my new 9950X I was getting an error in y-cruncher after a few minutes or hours. The 9950X sets the ProcODT's differently than the 7600X did. I set them to what the 7600X was running previously, and it worked for 60+ hours y-cruncher.
@@danielchristoph4576 I have an Asus B650E-E and I set these: Proc CS Drive Strength = 30 Proc CK Drive Strength = 30 Proc CA Drive Strength = 30 CPU On-Die Termination = 48 DRAM Data Drive Strength = 40 Rtt Nom Wr = 40 Rtt Nom Rd = 80 Rtt Wr = 120 Rtt Park = 80 Rtt Park Dqs = 80 If you have a different brand, basically I made it so ZenTimings looks like the 7600X: ProcOdt Pu = 48 (9950X on Auto was 43.6) ProcOdt Pd = 48 (9950X on Auto was 43.6) ProcCaDs = 30 (9950X on Auto was 30) ProcDqDs = Hi-Z (this I left on Auto in the BIOS) DramDqDs = 40 (9950X on Auto was 40) RttNomWr = 40 (9950X on Auto was 48) RttNomRd = 80 (9950X on Auto was 48) RttWr = 120 (9950X on Auto was 240) RttPark = 80 (9950X on Auto was 60) RttParkDqs = 80 (9950X on Auto was 60) I'm running 6000 MHz at 1.21v VDDIO,VDD,VDDQ and 1.07v SoC.
@@danielchristoph4576 I made it so ZenTimings looks like: ProcOdt Pu 48, ProcOdt Pd 48, ProcCaDs 30, ProcDqDs Hi-Z, DramDqDs 40, RttNomWr 40, RttNomRd 80, RttWr 120, RttPark 80, RttParkDqs 80. Maybe it was the new BIOS 3067 that set these wrong, not the 9950X.
@@Odysseas13LefkonaSt Thx for the details, I've allways wondered why DDR5 needs no more work on ProcOdt & co. Im gonna check my settings these weekend.
2x24 m die, able to get it passing CL 28, 6400, fclk 2133. Vsoc 1.27 didn't work lol, locked at 1.3. Timings are 28,38,32,31. Nitro 1 2 1, gear down mode enabled. I'm getting 60.0 latency. Can't break 24000 cinebench even though I can when ram is expo.
Maybe adjust power limits. Cinebench doesnt really care for memory performance, but it does care for IOD power consumption (includes memory controller).
I grabbed a sure to work G-Skill kit with EXPO when I moved to AM5 and a 9950X. Not much choice here in NZ, so I would have to import high performance kits. What I have is G-skill Ripjaws Neo. It's not very good. Low speed and high latency. Those Spek-tek G dies, which are reject or budget Micron ones I think. I'd be happy with 6400 with low latency and not having to pump lots of current into the memory. Not sure which way to go getting another kit. Just wait and see I guess. What I have works fine and I am not sure I'll notice 30 CAS over 36. It just bugs me sometimes. Once I get the reverse fans, cable combs, and a GPU upgrade I'll call the build done and go back to being OCD over electric guitar stuff instead. The other thing I'd like is to get the idle temps down. They sit at around 43C using my Lian Li 360mm AIO. If I enable PBO it's worse and will just about instantly hit 95C in Cinebench and reach high 70s in gaming or processing my audio workstation stuff. I am considering one of those AM5 offset kits. I dunno. I am not sure I am game to try a de-lid and liquid metal etc.
I have a very basic question since im very new to manual oc-ing anything on my pc: how much performance can you actually get out of all this for gaming specifically? Is it actually worth the effort?
I have a TUF Gaming x670e Plus Wi-Fi. Do you guys think it would be worth trying a high-speed kit that’s not on my board’s QVL list? Never understood if QVL should be taken as “must be on this list” or just “we know the kits on this list work”
I think 6400 problem is because of terminations and drive strength. My 64GB Hynix doesn't post at 6400 on ProcDqDs Hi-Z but when I change that with brute force values in a range it boot to windows but not stable yet. I tried different VSOC (1.2~1.3) and different FCLK but fruitless. Auto timings and auto nitro and loosing primaries doesn't affect anyway. The only helpful way was changing terminations to only make it boot up.
I feel like you can get us CL24. On Jan. 3rd 2025 Videocardz reported that Gskill released 6000mt CL26 kits. I cant wait for you to get ahold of one. I think these are the actual high bin tight timing kits. Im running the same cl28 that you are in this video. This is a fantastic tune mine is able to go cl28 and 2200, 1:1 all other values copied from you. Overnight Memtest pass. Thank you. Huge uplift from stock.
How can I know what chip is in a RAM kit? most sites only say like X mhz, X Cl, or something like X mhz Cl x-x-x-x (18-22-22-42 for example). Last ram kit I got is a Samsung C-Die and it works and all but i'd love to have more oc room and wouldn't buy if I knew it was C-Die
Could you do a 6400cl30 or the famous 6400cl28? From a 6000mhzcl30 hinyx a kit is posible? Or its still better to use the famous timings you used in that video and thats it 😅
Did you have the same problems with 1:1 on the previous BIOS ? 0706 on just a regular 9700x was absolute garbage to work with, so going back to 0606 helped a bunch. Even the 7800C38 XMP profile wouldn't work before. Now I can at least run it and tighten timings some. I have not yet tried 8000 on 0606 since I'm convinced this IMC is just no good and I'm afraid it won't work
I have a similar issue with my 9700X with a 6000 CL28 kit on a HDV/M2, thing is now I'm able to boot even 6600 1:1 everytime, but errors are completely random, so annoying
I could stabilize my 9800x3d on a Gigabyte x870 at 6200/2066 but 6400/2133 booted and didn’t work. I think that how the 9000x3d are build with the connection through the cache affects their ability to run higher speeds.
hey buildzoid. Clicked this video so quick haha. Quick question. What are the safe voltages for VDDG CCD and VDDG IOD? I'm new to all this RAM tuning and I've read that 1.05 for both is fine but then again would like a confirmation from you just for peace of mind.
@@SeventhCircle77 I know. I can't see the voltages in either zen timings or hwinfo. I do have mine setup at 1.05mv for both and it did make my 6400 @ 28 38 38 with 2133FCLK stable. But then again I am not sure if running both vddg ccd and iod at 1.05mv is safe as a daily voltage.
1.05v is totally safe, I believe that's in the Auto range for MSI and ASUS boards. Higher VDDG CCD might have an impact on CO/PBO stability, also generally can go lower than IOD (50mv between both often sweet spots)
@@luckysuicidal Sweet. I will go ahead and do that. I only have my 9800x3d at an undervolt. Running at -25 on all cores in curve optimizer and its been stable so far. Hopefully dropping the IOD 50mv won't change that.
2:50 that happened to my 7800X3D when my CPU can’t really run 6400 with tight timings, but it *does* run 6000/6200 with pretty low VSOC. The CPU’s IMC is just bad.
I have 2x48GB 6000 CL30 (same brand) and would Love to see if you could get it to CL28. I tried it, but it crashed on me after only a few hours, so I reverted it to CL30 I used the EXPO CL30 settings and lowered just the Cl to 28. I think I should try messing with the other timings, but that is way over my current level Would love to see you messing with a 2x48 -> 96GB kit :)
Does anyone know how quick the pcb breakdowns for gpu's take to hit this channel after release? I need to help my friend with a good nvidia brand for the 5070ti and all I know is amd and buy powercolor. Its 750 msrp so I'd like to get this right if anyone can help. All the crap I see online is anecdotal and fanboy slop. I know asus is good but man that asus tax is huge. And I know you guys know your stuff. I wish evga was still around, would make this easy. Help?
Usually depends how fast PCB pictures show up. Techpowerup usually has good pictures as soon as they can and thats what BZ looks at if he doesnt physically have the card. The VRM isnt so much the issue unless you're buying bottom of the barrel (gigabye eagle) or the vendor does something absolutely stupid about the cooling design.
Same problem on z790 with i7 won’t post on 7000 cl30 only works with 6800 bet should be able this board to do 7600 also high voltage doesn’t work wont post.
Intel's IMCs on Raptor are more defective than the cores. Not one single official report nor have any of these loser tech outlets have admitted it, but it's out there. Ain't just me saying it. Intel's IMC degrades too, and in my view, faster than the cores and regardless of microcode. Yup.
Try the MSI board, my 9800x3d on MSI X870 Carbon with 36gb G,skill royal 6000 cl28 runs 1:1 6200 cl-28 - 36 - 32 - 28 Fclk 2066 quite happily at 1.45v Vsoc 1.15. Using bios 7E49v1A1A. be interesting to see if its an Asus board / bios issue or cpu issue.
Maybe unlucky 9800x3d lottery? Got 6200 26-37-30-28 working stable on my 9800X3D on A-die SR, with 2200 FCLK. Motherboard B650E latest BIOS. Mem VDDQ 1.6v (probably overkill). 6400MHz on 1:1 would error out, even with looser timings, indicating the UCLK didn't like going higher than 3100.
@@Jacko_486 Based on my tests, you essentially want "something" in 1:1, either FCLK=UCLK OR UCLK=MCLK. Having either will give you a performance boost. Having "both" in sync has diminishing returns, such that an increase in FCLK is more beneficial than the sync you suggested. I should emphasize, this applies for 6000 range kits, where UCLK=MCLK
@@Jacko_486synced UCLK/FCLK gives lower latency, but if you can do 2200 FCLK it evens out enough where you start getting more performance with the bandwidth increase only worth going for if your synced FCLK value is below 2100 though
I'm having a similar experience with my 9800x3d on a Taichi Light, with the weirdness and 1:1 issues, but on 2x32mb 6000 cl30 (fury beast). Even so, I have better timings than this running stable. Watching this vid has made me realize I would probably have done better going for 2x24 8000 sticks. I don't think I want to try 8000 with 2x32. Ironically, while you hated working on this chip the vids that feature it is uniquely useful.
Been having super shitty latency with the 9800x3d, after a week trying to adjust timings with this exact ram kit minus the mobo... This vid is my last hope lol.
Update: Ran your benchmarks, they work, but like you said, it's a total disappointment, just not good enough. Returning my CPU asap. My old 7950x3d performs way better on the same mobo ram combo PS obviously not just returning it over this, this was the last straw, I've tried other sticks, EXPO and many manual timings, PBO and/or stock, and there's a new latency issue I've never had this bad with any other new gen CPU
very interesting that the 9800x3d seems to tentatively have worse memory controller than the non x3d parts?! my 7800x3d mc is also terrible. cant run 6200 w/ cl
X3D chips has a huge variation in terms of how far would they be able to be OCed my 7800x3d does 6400 cl28 1.54v and soc 1.26 while my friend did 6600 cl28 as will, but he told me he wasn't able to do it until he loosened the timings so much so he went back to 6400 cl28
"very interesting that the 9800x3d seems to tentatively have worse memory controller than the non x3d parts?!" No. Literally every single 9800X3D sells out due to the astronomical demand. Of those created by TSMC (for AMD), many will have a lousily binned IMC, so due to EVERY SINGLE ONE being sold (or blown up by a stupid kid installing it vertically), we will see more of the lousily binned ones. However, due to the greatness of the CPU over Intel's abortion, even a 9800X3D with a garbage-tier IMC would still be significantly better than any Core Ultra abortion. AMD does not deliberately bin WORSE memory controllers for X3D, no evidence has been shown to support that. In fact, they literally don't have the time to sort the damn binning. They SELL. EVERY. SINGLE. ONE. So we see more of the entire spectrum of the silicon lottery. Though I'm sure AMD will end up possibly pushing their limits for binning with the ungodly amount of demand they have. It won't mean they're going to put out products that won't post 6000 1:1 at 1.3V, that would make someone eligible for RMA and they won't do that either.
I'm having the exact same experience. Same mobo, same CPU, 2x32 rather than 2x16, but Hynix A-die nonetheless. Just completely unstable at random even with the most forgiving settings. Really disappointed by this CPU.
@@genjimain yea, my fully stable 8000 mhz setup with 505 doesnt work with 706. by "doesnt work" i mean if it boots, it works, but 9 out of 10 times it doesnt boot and stucks in training. let me know how it goes if you try 505.
@@ActuallyHardcoreOverclocking while I was trying to stabilise 6600/2200 (which I couldn’t)… it’s always my weakest core that throws out error in prime95 at 1.3vsoc maybe use curve shaper to add some voltages on top of your weak core and try 6400 again ? Just throwing an idea out there
Can we get a 2x32GB video pleeeeeeeeeeeeeeeeeeeeeeeeease
Yes, this pretty please!!
Pretty pretty please
Why not just look at 2x48? It's just Dual-Rank that matters.
@@ChrispyNut I didn't find any video from him with 9800X3D and 2x48. Is there one?
Yes please! :)
Neat, the timing of this video is perfect.
I have 9800x3d with a 6000 CL30 2x32GB kit.
I used your popular easy hynix ddr5 timtings.
It now runs on 6200 CL26,36,36,28 with FCLK at 2200 and 1:1 Mode.
VSOC is 1.2V
So far it‘s stable and no programs are crashing. Did some stress testing for 30-60min but still have to to longer ones.
Only thing my CPU struggles with is undervolting. Some cores won‘t to anything lower than -5/10, my max is -20.
EDIT: Core 1 & 2 won’t undervolt with -10, had to decrease to -5.
I had to increase SOC to 1.21, VDD Misc to 1.15 and it prime95 works for 2 hours as of now. If anyone is interested I will continue to update this comment and share my settings once I know it‘s stable enough.
hmm interesting. have you tested it in Karhu?
@@Pullo83I‘m on Linux so no. I use the Prime95 equivalent and y-cruncher.
For lower frequency workloads I use some games I know that struggled not to crash on more aggressive overclocks.
where did you get the "popular easy hynix ddr5 timtings"? :D cant find the video
After the B850 boards have been released now, I'd love to see some 8000 tests with the 9800X3D on the ASRock Steel Legend/Riptide. They claim to support many 8000 Kits out of the box. :)
I've got a crappy 2x16gb set of Teamgroup 6000c30 running at 6400c28 1:1 2200fclk, nitro 1-2-0, pmu 0, swap apu, with a 9800x3d using an open box $200 X670E-F from Microcenter. I have a new kit of Gskill 8000c38 expo coming next week. I'm hoping I can do 6400c26 with the new kit.
Dont forget to update in couple weeks how it went cause i am curious.
I have a 2x24 gskill 8000 kit and the best I could achieve is 6400cl28 in 1:1 mode and 8000cl34 in 2:1. I could also kind of run it with 8400cl38 but it's not stable. Like I can play games but I can't pass some stress-tests, which means that if I'll play a game with similar loads I'll encounter some problems. But I also have the cheapest x670e mobo, asrock x670e pg lightning. I think it's worse than some good (or maybe even most of) b650e boards. Maybe if I'll get something like x870e taichi lite I'll get it stable but it's not guaranteed and also not worth it. if I knew for sure that it will work I would at least consider this option.
Extremely lucky I got a garbage chip
Hey BZ, nice vid. I just wanna say I had pretty much an identical experience with my 9800x3d. Random boot fails on mild settings, GDM causing errors, floppy error behavior, even basic expo profiles getting errors sometimes, etc. I suspect my motherboard (ASRock b650i lightning wifi) is the issue but this video is giving me second thoughts. Going to try with a different memory kit and mobo and report back.
Clear CMOS whenever stuff gets weird, failed FCLK overclocks often lead to FUBAR memory training and can only be fixed that way.
With ASRock boards, disable "Memory Training Runtime Reduction".
It's supposed to turn itself off when you input manual OC settings but it doesn't, causes partial (and failing) retrains when changing settings.
Same mobo here. I couldn't even run expo 6000 MHz. But 4800 MHz was fine. I checked with ASRock and in the process they asked for pictures of my RAM labels. This made me reinstall the RAM. Before taking them out, the RAM looked to be installed properly. Tabs were engaged and all. However, reinstalling allowed for EXPO. Strange! Guess the single tab system needs some more work
@@luckysuicidalalso sometimes(perhaps often) bios can corrupt so to solve that just reflash same bios via usb from bios, or even better reflash it via button on your mobo.
Hello, same for me, initially. Very floppy and inconsistent behavior. On some boots, it can run 20hours of Karhu, on other boots, error at 4 minutes.
I change my EXPO kit corsair Platinum 64gb 6000 C30 to a new kit corsair Platinum 7200 c36 XMP overclocked to 6200 C28 with fclk 2200, works flawlessly even at 6400 C28. My 9800X3d scales better with M die single rank kit versus dual rank A die. I need vddq= vdd= 1,5 voltd for rock stability Vsoc 1,15 volts.
Can I request a video where you explain the training settings, nitro, all that stuff? There's not much info out there and I just don't really get it lol
Yeah nitro and procodt.. genuinely no clue where to even begin with that stuff and how they affect performance / stability
I tired using the tightest settings because why not, MB stopped posting lol, I had to clear CMOS to get it posting again.
Sadly as you said there are not any guides that explain each setting and what are the possible setup for different ram kits/timings.
Good thing I used the profile saver so it was easy to get my previous settings back.
Yeah, I agree! I understand frequencies, timings and voltages, but nitro and drive strengths are hard to understand for me
It doesn't need to be explained. You use 1-3-1 and call it the day unless you have a $500-1000+ motherboard and want to do days worth of testing. And if 1-3-1 doesn't work, you leave it on auto. Same with ProcCaDs and ProcDqDs and DramDqDs, those get set to 30Ω/34.3Ω/34Ω, period, the end. ASUS is the only one who got that right. MSI likes to jack DramDqDs up for no good reason. All that does is reduce stability with no gain. MSI also likes to set ProcODT too high. Usually it doesn't get set over 40Ω if even that. Almost always ProcODT is under 40Ω on Zen 3, 4 and 5. Usually in the 30Ω range.
@lePurpleDragon You can't possibly imagine how cryptic most of your comment is to me.
Memory OC is probably among the most frustrating things.
So many options, luck seems to be important and extremely difficult to tell where the stability problems originate from.
Get this man another 9800x3d!
I checked they're still completely sold out in the UK ATM
@@ActuallyHardcoreOverclocking The 9800X3D is back in stock in the US; it probably will be in stock in the UK soon.
@@ActuallyHardcoreOverclocking the pre order options deliver within 3 days currently, its a weird grey zone
@@ActuallyHardcoreOverclocking OCUK are meant to be getting 1k+ stock this week, so you should be able to order either before the weekend or sometime next week.
Does his 111 SP score have anything to do with the problems he comes up against, I have the same setup with 114 SP.
Also big thanks for these timings, have applied them successfully
8000/2000 is the best setup for my 9800X3D as far as I'm concerned. It was also extremely easy to setup. 1:1 didn't really provide benefits latency-wise and had some weird behavior when I tried tightening the timings as well.
Do you have settings for that? I am waiting on my 9800x3d but in the mean time I ordered some Patriot 7000cl32 ram kit. I guess these are top tier ram?
@@Alex-fo1rx I have a g skill a-die 7800CL36 XMP kit I'm running 8000/2000 at 1.05VSOC. The XMP settings for my kit are 1.45V VDD, VDDQ, and VDDIO which works perfectly on my X670E Carbon. Just for fun I've tried a lot of different timings ranging from CL38 to CL34 using BZ's recent 2:1 videos and had success. To start with, this video (G.skill 2x16GB DDR5-8000 CL38 EXPO optimized memory timings, MSI X870 Tomahawk) would be a good place to start but I don't have experience with Patriot memory. BZ has mentioned that Patriot kits seem to not behave well with Ryzen chips from his more recent experience with their kits for some reason.
Before BZ had a 9800X3D I was using his 9700X because that's basically the same CPU in terms of mem oc compatibility outside of silicon lottery variance. Ideally, the XMP/EXPO profile and your motherboard will set the termination settings appropriately so you don't have to deal with that headache along with things like VDDP, etc.
Would say in general see if 1:1 6200/6400 or 2:1 8000 works without a ton of hassle and choose whichever one you can stabilize and get the best performance from. With my setup, in memory latency benchmarks 8000/2000 is awesome and runs at low VSOC while being comparable in gaming (maybe better than 1:1 in my case since the latency is consistently better?) from the testing I've done swapping back and forth. Worth noting though that Ryzen 9000 in general seems to have higher latency values in things like AIDA than Ryzen 7000 (like my 7950X3D), but I don't really care about AIDA, I just use it for relative differences between settings along other tools like PyPrime which I think is better.
6400 1:1 is the same . I asked buildzoid this too
@@Alex-fo1rx Wrote out a decently long reply and for some reason YT didn't keep it I guess so that sucks and I don't want to write it out again. Basically 1.05VSOC, 1.45 VDD, VDDQ, VDDIO since my 7800 CL36 gskill XMP a-die kit uses those settings. For timings just start with BZs 8000 CL38 settings from recent videos. BZ has said that Patriot kits he's tested on Ryzen lately haven't worked well for some reason.
If your IMC can do 8000, it’s very easy to setup. If it can’t, it just won’t no matter how hard you try. My 7950X3D can do 6400 at VSOC 1.1V but it just won’t post at 8000.
Damn need this in 2x32 dual rank, always learning
Try to redeem this chip in another board: There was a rash of issues on this board and the Strix version with this bios breaking some EXPO profiles that previously worked. I personalty had different N63 performance depending on if I set EXPO in the Tweaker tab vs doing everything in the hex menus manually, but that could have been due to fabric error correction. (sharp waves in CPU usage, 1-2 second period) If I didn't touch the Ai Tweaker tab, which was the only place to configure XMP/EXPO, that test was steady. I don't have that x870e Strix MB anymore. If its doing the same thing then I guess we both got weird parts.
This but then with dual rank (2x32 is what most people use i guess) would be super useful.
this is what i need for my new build next month
Try it with the x870 Tomahawk
what is the impact of nitro settings (from 2,3,1 to 1,2,0 for example) vs let's say 2 ticks of tCL
same question with matched PHYRDL
Best OC video ever!
I have the same memory kit as you used in this video and mine can do 6400mhz cl28 with tight timings no problem, even cl26 if I had a RAM fan for temps, so I would definitely suspect the CPU as the cause of your pain :(. It hurts to lose to the silicon lottery, I know that feeling all too well from my Intel days...
buildzoid overclocking from the future: 09/17/2045 roflmao
I have m-die, what would be best for 9800x3d?
I need on the same motherboard and CPU configuration with Corsair 2x48GB modules at 6000MT CL30 - 1.5V on VDD & VDDQ (SoC at 1.2V) BIOSv0706.. something messed up with AGESA or BIOS because i can run them at same speed with BIOSv0606 at 1.4V VDD & VDDQ (Soc at 1.1V)
Strange, I had no issues running 1:1 6400 on a b650e-i with my 9800x3d and 24gb m-die.
whats a safe 24/7 daily vsoc voltage that won't degrade the chip in the long run?
Could ya make a 48gb gskill video for a 9950x in 1/1 mode
Would dual rank be a good idea for ryzen 9800x3d? Or should I go single rank with lower timings instead?
I'm working on a 8347mt setup right now. Can do 8400 but not 1:1 fclk. So went down to 99.5bclk and it's stable so far hour of ycruncher. I'm gonna do a longer test tomorrow but it could be cool.
did you try adjusting VDDGs? should easily work at 950/1000mv
I'd be very surprised if a chip couldn't do 2100 fclk at all, certain boards just default vddgs way too low in 2:1
dont use 1:1 fclk term ... you can't sync the fclk (it's never going to be 1:1 ... its more 1:4, 8400/4=2100 flck) ... because 1:1 is the ratio between MCLK and UCLK :)
@@PedroSombreireiro I'm syncing the fclk to the uclk.
@@luckysuicidal I have not tried playing with vddg much, I think I may be at an IMC limit. I'll give that a try tho.
That instability might be due to IF overclock. My 9800X3D can do 2133 IF with 6400 MT/s memory in 1:1 at 1.25v SOC, and runs mostly fine, for hours, but occasionally starts failing out of nowhere. Reducing the IF clock to 2000 is stable from what I can tell. Giving more voltage to SOC is not good long term, so I made peace with having IF at stock.
Try adjusting VDDG voltages to stabilize IF at 6400/2133.
9:58-10:59 Maybe you have to adjust the ProcODT's. My 2x32 M-die ran fine at 6000 with my 7600X, but with my new 9950X I was getting an error in y-cruncher after a few minutes or hours. The 9950X sets the ProcODT's differently than the 7600X did. I set them to what the 7600X was running previously, and it worked for 60+ hours y-cruncher.
What exactly did u change?
@@danielchristoph4576
I have an Asus B650E-E and I set these:
Proc CS Drive Strength = 30
Proc CK Drive Strength = 30
Proc CA Drive Strength = 30
CPU On-Die Termination = 48
DRAM Data Drive Strength = 40
Rtt Nom Wr = 40
Rtt Nom Rd = 80
Rtt Wr = 120
Rtt Park = 80
Rtt Park Dqs = 80
If you have a different brand, basically I made it so ZenTimings looks like the 7600X:
ProcOdt Pu = 48 (9950X on Auto was 43.6)
ProcOdt Pd = 48 (9950X on Auto was 43.6)
ProcCaDs = 30 (9950X on Auto was 30)
ProcDqDs = Hi-Z (this I left on Auto in the BIOS)
DramDqDs = 40 (9950X on Auto was 40)
RttNomWr = 40 (9950X on Auto was 48)
RttNomRd = 80 (9950X on Auto was 48)
RttWr = 120 (9950X on Auto was 240)
RttPark = 80 (9950X on Auto was 60)
RttParkDqs = 80 (9950X on Auto was 60)
I'm running 6000 MHz at 1.21v VDDIO,VDD,VDDQ and 1.07v SoC.
The CPU does not AUTO ProcODT, the motherboard does.
@@danielchristoph4576 I made it so ZenTimings looks like: ProcOdt Pu 48, ProcOdt Pd 48, ProcCaDs 30, ProcDqDs Hi-Z, DramDqDs 40, RttNomWr 40, RttNomRd 80, RttWr 120, RttPark 80, RttParkDqs 80. Maybe it was the new BIOS 3067 that set these wrong, not the 9950X.
@@Odysseas13LefkonaSt Thx for the details, I've allways wondered why DDR5 needs no more work on ProcOdt & co. Im gonna check my settings these weekend.
tightened up ddr5 6000 cl30 just a few settings. notice zero difference with 9700x. honestly, great success.
Can you do this with 2X32 GB kit too?
2x24 m die, able to get it passing CL 28, 6400, fclk 2133. Vsoc 1.27 didn't work lol, locked at 1.3. Timings are 28,38,32,31. Nitro 1 2 1, gear down mode enabled. I'm getting 60.0 latency. Can't break 24000 cinebench even though I can when ram is expo.
Maybe adjust power limits. Cinebench doesnt really care for memory performance, but it does care for IOD power consumption (includes memory controller).
did you have a look at the drive strengths? They look a little low to me
I grabbed a sure to work G-Skill kit with EXPO when I moved to AM5 and a 9950X. Not much choice here in NZ, so I would have to import high performance kits.
What I have is G-skill Ripjaws Neo. It's not very good. Low speed and high latency. Those Spek-tek G dies, which are reject or budget Micron ones I think. I'd be happy with 6400 with low latency and not having to pump lots of current into the memory. Not sure which way to go getting another kit. Just wait and see I guess.
What I have works fine and I am not sure I'll notice 30 CAS over 36. It just bugs me sometimes. Once I get the reverse fans, cable combs, and a GPU upgrade I'll call the build done and go back to being OCD over electric guitar stuff instead.
The other thing I'd like is to get the idle temps down. They sit at around 43C using my Lian Li 360mm AIO. If I enable PBO it's worse and will just about instantly hit 95C in Cinebench and reach high 70s in gaming or processing my audio workstation stuff. I am considering one of those AM5 offset kits. I dunno. I am not sure I am game to try a de-lid and liquid metal etc.
Sounds like my 9800x3d throws error after 1 hr at 6400 nitro 121 but it does 8200 h16a and 8400 h24m 🤦
lol
I have a very basic question since im very new to manual oc-ing anything on my pc: how much performance can you actually get out of all this for gaming specifically? Is it actually worth the effort?
I have a TUF Gaming x670e Plus Wi-Fi. Do you guys think it would be worth trying a high-speed kit that’s not on my board’s QVL list? Never understood if QVL should be taken as “must be on this list” or just “we know the kits on this list work”
I think 6400 problem is because of terminations and drive strength. My 64GB Hynix doesn't post at 6400 on ProcDqDs Hi-Z but when I change that with brute force values in a range it boot to windows but not stable yet. I tried different VSOC (1.2~1.3) and different FCLK but fruitless. Auto timings and auto nitro and loosing primaries doesn't affect anyway. The only helpful way was changing terminations to only make it boot up.
Really helpful video. With this ram kit, can you run these timings at 6000 tcl 28 on most 9800x3ds, do you think?
I feel like you can get us CL24. On Jan. 3rd 2025 Videocardz reported that Gskill released 6000mt CL26 kits. I cant wait for you to get ahold of one. I think these are the actual high bin tight timing kits. Im running the same cl28 that you are in this video. This is a fantastic tune mine is able to go cl28 and 2200, 1:1 all other values copied from you. Overnight Memtest pass. Thank you. Huge uplift from stock.
How can I know what chip is in a RAM kit? most sites only say like X mhz, X Cl, or something like X mhz Cl x-x-x-x (18-22-22-42 for example). Last ram kit I got is a Samsung C-Die and it works and all but i'd love to have more oc room and wouldn't buy if I knew it was C-Die
Could you do a 6400cl30 or the famous 6400cl28?
From a 6000mhzcl30 hinyx a kit is posible?
Or its still better to use the famous timings you used in that video and thats it 😅
Did you have the same problems with 1:1 on the previous BIOS ?
0706 on just a regular 9700x was absolute garbage to work with, so going back to 0606 helped a bunch. Even the 7800C38 XMP profile wouldn't work before. Now I can at least run it and tighten timings some. I have not yet tried 8000 on 0606 since I'm convinced this IMC is just no good and I'm afraid it won't work
Is micro stutter still a thing ? On 9800x3d, when a game doenst fit in the cache ?
Is it also worth trying on dual rank kits? If not - what to change?
Is RAM active cooling recommended for this setup? I saw you used it for ddr5 8000 cl34
I have a similar issue with my 9700X with a 6000 CL28 kit on a HDV/M2, thing is now I'm able to boot even 6600 1:1 everytime, but errors are completely random, so annoying
Would be great if you could test this cpu on a msi board! 🙏🏻
I could stabilize my 9800x3d on a Gigabyte x870 at 6200/2066 but 6400/2133 booted and didn’t work. I think that how the 9000x3d are build with the connection through the cache affects their ability to run higher speeds.
hey buildzoid. Clicked this video so quick haha. Quick question. What are the safe voltages for VDDG CCD and VDDG IOD? I'm new to all this RAM tuning and I've read that 1.05 for both is fine but then again would like a confirmation from you just for peace of mind.
Depends on the setup. Am5 950mv should work ok in most cases. I'm not sure what the auto is.
@@SeventhCircle77 I know. I can't see the voltages in either zen timings or hwinfo. I do have mine setup at 1.05mv for both and it did make my 6400 @ 28 38 38 with 2133FCLK stable. But then again I am not sure if running both vddg ccd and iod at 1.05mv is safe as a daily voltage.
1.05v is totally safe, I believe that's in the Auto range for MSI and ASUS boards.
Higher VDDG CCD might have an impact on CO/PBO stability, also generally can go lower than IOD (50mv between both often sweet spots)
@@luckysuicidal Sweet. I will go ahead and do that. I only have my 9800x3d at an undervolt. Running at -25 on all cores in curve optimizer and its been stable so far. Hopefully dropping the IOD 50mv won't change that.
I'm running at 1.55 vddcand 1.35 vddq
2:50 that happened to my 7800X3D when my CPU can’t really run 6400 with tight timings, but it *does* run 6000/6200 with pretty low VSOC. The CPU’s IMC is just bad.
do you know why when i set speed to 6400 the phyrdl desync? while 6200 doesn't
@ probably unstable timings? I haven’t encountered that yet in 1:1 mode.
didn't switching to M-die help you with your timings?
@@catonturtle3202 Yeah it did. Still, it was a bit loose compared to what I have now which requires an extra 0.5V for the VSOC
@@PowellCat745 maybe cus gdm disabled?
This the video I was waiting for, I have the same exact problem with my 9800x3d 1 to 1 mode is extremely hard to get stable
I have 2x48GB 6000 CL30 (same brand) and would Love to see if you could get it to CL28. I tried it, but it crashed on me after only a few hours, so I reverted it to CL30
I used the EXPO CL30 settings and lowered just the Cl to 28. I think I should try messing with the other timings, but that is way over my current level
Would love to see you messing with a 2x48 -> 96GB kit :)
I hope this works on G.Skill M die kits as well because I don't like selling a kidney just to get my hands on a better binned A die kit lmao.
They will have the optimized times for those ram g skill for only the 6000 cl28 ?
Please do more 48gb kits!
it's not the cpu, it's the bios...
one simple question, have you tried bios version 505?
bios 706 has biiiig problems with this board
Is B650 out of the question for 8000?
I got a G.Skill Trident Z5 Neo RGB 32GB DDR5 6400 CL32 kit, would this timings work if paired for 7800X3D?
That kit did not work at expo with my 9800x3d x870e taichi, tried even 6200 at expo and still wouldn’t work
dont even have a 9800x3d but i can already tell this is going to be a great watch for those that do lol
Need dual rank, I'm using 6200cl30-38-34-30 with solid 65ns 1.52v (9950x and x670e Taichi Carrara)
Mostly you can just use the single rank timings with higher VSOC. Btw 6200CL30 at 1.52V seems a bit high.
Does anyone know how quick the pcb breakdowns for gpu's take to hit this channel after release? I need to help my friend with a good nvidia brand for the 5070ti and all I know is amd and buy powercolor.
Its 750 msrp so I'd like to get this right if anyone can help. All the crap I see online is anecdotal and fanboy slop.
I know asus is good but man that asus tax is huge. And I know you guys know your stuff.
I wish evga was still around, would make this easy. Help?
Usually depends how fast PCB pictures show up. Techpowerup usually has good pictures as soon as they can and thats what BZ looks at if he doesnt physically have the card. The VRM isnt so much the issue unless you're buying bottom of the barrel (gigabye eagle) or the vendor does something absolutely stupid about the cooling design.
Same problem on z790 with i7 won’t post on 7000 cl30 only works with 6800 bet should be able this board to do 7600 also high voltage doesn’t work wont post.
Intel's IMCs on Raptor are more defective than the cores. Not one single official report nor have any of these loser tech outlets have admitted it, but it's out there. Ain't just me saying it. Intel's IMC degrades too, and in my view, faster than the cores and regardless of microcode. Yup.
Try the MSI board, my 9800x3d on MSI X870 Carbon with 36gb G,skill royal 6000 cl28 runs 1:1 6200 cl-28 - 36 - 32 - 28 Fclk 2066 quite happily at 1.45v Vsoc 1.15. Using bios 7E49v1A1A. be interesting to see if its an Asus board / bios issue or cpu issue.
why trp @36?
isn't it an A die?
@@just_OsaMa23 sorry didn't have my glasses on
Maybe unlucky 9800x3d lottery? Got 6200 26-37-30-28 working stable on my 9800X3D on A-die SR, with 2200 FCLK. Motherboard B650E latest BIOS. Mem VDDQ 1.6v (probably overkill). 6400MHz on 1:1 would error out, even with looser timings, indicating the UCLK didn't like going higher than 3100.
Would'nt it be better to set fclk at 2067? Because 6200/3 = 2066.66
@@Jacko_486 Based on my tests, you essentially want "something" in 1:1, either FCLK=UCLK OR UCLK=MCLK.
Having either will give you a performance boost. Having "both" in sync has diminishing returns, such that an increase in FCLK is more beneficial than the sync you suggested.
I should emphasize, this applies for 6000 range kits, where UCLK=MCLK
When I say FCLK=UCLK, I mean 1:2* sync not 1:1
@@Jacko_486synced UCLK/FCLK gives lower latency, but if you can do 2200 FCLK it evens out enough where you start getting more performance with the bandwidth increase
only worth going for if your synced FCLK value is below 2100 though
0706 bios is super buggy it just refuse to train my CMK96GX5M2B6000Z30 memory expo, i had to go back to 0606 bios lol
I'm having a similar experience with my 9800x3d on a Taichi Light, with the weirdness and 1:1 issues, but on 2x32mb 6000 cl30 (fury beast). Even so, I have better timings than this running stable. Watching this vid has made me realize I would probably have done better going for 2x24 8000 sticks. I don't think I want to try 8000 with 2x32. Ironically, while you hated working on this chip the vids that feature it is uniquely useful.
Been having super shitty latency with the 9800x3d, after a week trying to adjust timings with this exact ram kit minus the mobo... This vid is my last hope lol.
Update: Ran your benchmarks, they work, but like you said, it's a total disappointment, just not good enough. Returning my CPU asap. My old 7950x3d performs way better on the same mobo ram combo
PS obviously not just returning it over this, this was the last straw, I've tried other sticks, EXPO and many manual timings, PBO and/or stock, and there's a new latency issue I've never had this bad with any other new gen CPU
I wish AMD had more RAM Ratios between 6000 and 6600.
My 9800x3d is exactly like this, unlucky :/
Lmao fills me with sadness, I actually lold
very interesting that the 9800x3d seems to tentatively have worse memory controller than the non x3d parts?!
my 7800x3d mc is also terrible. cant run 6200 w/ cl
Super weird. I wonder if there was some sort of intentional trade off on AMD's part, or if buildzoid actually just lost the silicon lottery.
X3D chips has a huge variation in terms of how far would they be able to be OCed
my 7800x3d does 6400 cl28 1.54v and soc 1.26
while my friend did 6600 cl28 as will, but he told me he wasn't able to do it until he loosened the timings so much so he went back to 6400 cl28
"very interesting that the 9800x3d seems to tentatively have worse memory controller than the non x3d parts?!"
No. Literally every single 9800X3D sells out due to the astronomical demand. Of those created by TSMC (for AMD), many will have a lousily binned IMC, so due to EVERY SINGLE ONE being sold (or blown up by a stupid kid installing it vertically), we will see more of the lousily binned ones. However, due to the greatness of the CPU over Intel's abortion, even a 9800X3D with a garbage-tier IMC would still be significantly better than any Core Ultra abortion.
AMD does not deliberately bin WORSE memory controllers for X3D, no evidence has been shown to support that. In fact, they literally don't have the time to sort the damn binning. They SELL. EVERY. SINGLE. ONE. So we see more of the entire spectrum of the silicon lottery. Though I'm sure AMD will end up possibly pushing their limits for binning with the ungodly amount of demand they have. It won't mean they're going to put out products that won't post 6000 1:1 at 1.3V, that would make someone eligible for RMA and they won't do that either.
I thought 9800x3d would have been well binned to do well in both modes. I guess not...
AMD doesn't bin the IMC at all from what it seems, just the CCDs.
Yeah, that is unfortunate.
It would be nice to buy a premium product and know that you really got the best.
YESSSSSSS
I'm having the exact same experience. Same mobo, same CPU, 2x32 rather than 2x16, but Hynix A-die nonetheless. Just completely unstable at random even with the most forgiving settings. Really disappointed by this CPU.
Try bios 505. latest bios (706) kinda buggy with memory
@@nebiforever Really? I'll try it now. Thank you.
@@genjimain yea, my fully stable 8000 mhz setup with 505 doesnt work with 706. by "doesnt work" i mean if it boots, it works, but 9 out of 10 times it doesnt boot and stucks in training. let me know how it goes if you try 505.
21 minutes? seems short for a BZ video.
I had bad experience with the 7800x3d but the 9800x3d for me has been atleast average in terms of memory frequency. Must be luck of the draw.
Asus released the Apex for AMD...
Not a cpu issue.
VDDP at 1.10V and you get most stuff stable like 6400 prolly even 6600 😂
Yeah I also think vddp might help stabilise but then again I’m running 6400/2133 with 0.95 vddp I guess each chip is different
been there done that didn't help.
@ did you have any -CO on the cpu?
@@ActuallyHardcoreOverclocking while I was trying to stabilise 6600/2200 (which I couldn’t)… it’s always my weakest core that throws out error in prime95 at 1.3vsoc maybe use curve shaper to add some voltages on top of your weak core and try 6400 again ? Just throwing an idea out there
@@fleecee you do ram overclocks first numb nuts
It's 2045, why even bother with such old tech?
just disable GDM, AM5 is not like AM4
Another AMD video I'll skip this one
MSI motherboard is better~
First?
have you tried running your 7800x3d in 2:1 8000 with mobo+kit which work at 8000 with zen5?