03-25-2023 08:33 AM
(sorry to double post, i realise i might not have posted in the right thread)
Hello, sorry for my poor english, i speak french. I build a lot of computers since 20 years, im 43 yo. Never had this much trouble with a new build.
I just build a 13900ks with a z790-e gaming, 2 pcie 4 ssd, and some 7600mt ram on the QVL list (F5-7600J3646G16GX2-TZ5RS), and nvidia 4090 that was in my previous build, and corsair 1200w shift power supply.
I started by flashing bios using usb stick to latest version (0816)
Posted to bios, loaded defaults. Then installed windows 11.
Then installed chipset, including intel ME + lastest firmware update to ME,
Installed drivers for audio, nvidia etc.
Installed steam + somes games etc.
Then after all that was done, i enebled xmp profile in bios, then posted fine and booted windows. Never had problems with xmp before, but did a big mistake there, i did not run memtest to be sure it was stable. Started playing games, all games crash to desktop. Checking realibity history and it crash with hardware error code 141.
Then to find the problem, did run memtest directly from motherboard and it give me errors. So i thought i just found the problem. Tryed another pair of ram thinking it might be defective, same problems. So i lowered the frequency of the ram. Instead of 7600, i used 6600 (with the same timing). Ran memtest all over the 4 pass (took a few hours) 0 errors. Was a bit sad i had to lower frequency but still high enough anyway. So went back to windows, playing games again, and they all still crash to desktop (maybe after 30min-1h of gaming). Never had bluescreen or lockup, just games stopping and back to desktop without any message. Again, realibity history say hardware errors code 141.
Online research say it is normally related to GPU drivers, so uninstalled nvidia drivers with DDU, reinstalled lasted drivers, that did not fixed anything. The video card was in my last build (amd 5900x) and never had any problems. So i decided to run ram stock (4800) and so far did not crashed games. I will need more testing, but so far anything i try with xmp enable, even low frequency, even if it pass memetest it crash games after some time. Last test i did was not using xmp, putted 5600 for frequency, and copied lowest timing from a 5600 kit i saw (28-34-34-89) and gamed for 2h without any crash.
I read that primary timing are not very important with ddr5, so im not very satisfied about that. Im a bit lost, and it eat lots of time testing all this. I was expecting memory officially supported by MB to work out of the box with a 13900ks... I was wrong but im worrying i have a bad cpu or bad motherboard.... ram is OK im pretty sure.
What do you guys think? any idea?
03-25-2023 09:44 AM
Quite a few people have had success solving memory problems using a pressure plate for the CPU. A few brands are available. The standard Intel CPU clamp can cause distortion which can lead to poor pin contact and memory issues are possible. If you don't want to do that, have a look at memory overclocking guides for DDR5 and look at increasing voltages to improve stability. You'll find suggestions in these threads as well. Memory stability issues come up very frequently, so have a good look around. Good luck.
03-25-2023 10:21 AM
Thanks you very much for your answer. yes im aware of the cpu braket, but i thought it was mostly for temperature. I use a h170 420mm aio cooler and temps are ok, but didn't knew it could cause stability issue not to use one. I saw a video from jay2cent that tighting the cpu cooler too much can cause issues, so i have been careful with that.
I know xmp is an overclock and it not guaranty to works, many thinks enter into account, the cpu memory controller is part of that. But normally reducing the frequency (keeping voltage and timing the same) should at some point be stable? I know passing memtest don't mean you will not have problems and don't mean its 100% stable, but jesus.... coming from a ram listed as compatible for the motherboard, dropping from 7600 to 6600... passing memtest (4 pass) without errors, but stil crashing every games is beyon me... I will need to test more, but i think if i don't eneble xmp and set frequency manually and voltage as expected by xmp, and primary timing, it may works (at 6600 at least). I wonder if enebling xmp and putting 5600 frequency would still crash the games....
03-25-2023 11:46 PM
If you're using Memtest86, this is really only a baseline for stability with DDR5. Use Karhu Ramtest for a more accurate picture of stability if still experiencing crashes.
You may need to tune the CPU System Agent (Extreme Tweaker Menu) and Memory Controller Voltage (Advanced Memory Settings)—values typically between 1.1 - 1.35 and 1.2 and 1.45 respectively. Test methodically in increments.
03-25-2023 10:54 AM
Yes, might help, I think it will a case of working through options and see what works best. Reagrding crashing in games, have you made sure that all drivers are up to date, including Windows MEI drivers and ME Firmware?
03-25-2023 10:57 AM
Yes 1st thing i did after instlling win 11 was installing chipset and ME + ME firmware (from the asus website for my MB)
03-25-2023 04:40 PM
Have you looked at Mokichu latest drivers forum? Top notch in the community and always has latest drivers not on website.
03-26-2023 05:51 AM
Thanks for the help. It just feel like something is wrong with xmp, cause my ram kit is 7600MT and running it at 6600MT with same timings (36 cas) pass memtest, but maybe would have failed other test, but crash games. Running it at 6600MT WITHOUT xmp enable with tighter 1st timings 30-38-38-28, (altrought 2nd timing are more important) didn't crashed games so far. Only tested a few hours, will have to test more to be sure but so far, if xmp is enable, it cause crash in games. Is it because 2nd timings are tighter with xmp then with auto? Maybe its the reason why it crash at xmp 6600 with looser 1st timing and it don't at 6600 without xmp with tighter timings.
03-26-2023 11:27 PM
Applying XMP will apply tighter memory subsets. Typically, applying the frequency and primary timings manually will result in some different parameters depending on what's programmed into the SPD for a given kit. So it's not that surprising you're seeing different results.