The full path to fix the problem.

[naporitan]

1-108658-512-300x300.png (99.5 KiB) Viewed 1118 times

The latest information is available at the link below.

UPD: 09.09.2025
viewtopic.php?f=10&t=14883&start=80#p118877

UPD: 05.10.25
viewtopic.php?t=14883&start=90#p119293

viewtopic.php?t=14883&start=90#p119509

UPD: 16.10.2025
viewtopic.php?t=14883&start=90#p119507

[ChristophSmaul1337]

If you suspect your RAM kit to be a problem, you may run a simple test to rule specific things out. As you probably know, RAM stability is dependant on temperature. What can happen is that you've overclocked your RAM one day, let it train itself while the temperature was pretty low, and since there's "Fast Boot" enabled it will skip through some training on boot up, essentially "locking in" the once trained properties. When you're playing a game and the inside of your PC heats up more, due to the GPU dumping heat for example, the once stable memory might become unstable with elevated temperatures. To eliminate this specific cause, you may run a simple test:

1. Boot into BIOS
2. Restore default settings for now. After that, enable the XMP profile on the RAM. Do not tweak timings yourself and leave everything "as-is".
3. Find an option labelled "Fast Boot" or similar in your BIOS. Every BIOS should have such an option. It might be called differently depending on BIOS and motherboard manufacturer. Your BIOS might give you a search box you can use to locate the setting. Additionally, ASUS boards have something called "MCH Full Check", there might be something similar on other vendor's BIOSes. Try to locate this setting as well if your BIOS has such.
4. Once you found the "Fast Boot" setting, disable it. If you can find the "MCH Full Check" option, enable it.
5. Boot into Windows. Then, immediately reboot the computer again
6. Wait for Windows to boot up agian. Run a memory stress test of your liking for about 20 to 30 minutes. TM5 with the anta777 extreme preset is pretty good, but you can also start a game and play for a bit. The objective is to get the RAM sticks warmed up. 50°C and upwards would be ideal.
7. After ending the test / exiting out of the game, immediately reboot the computer. After the system has POST'ed, go back to BIOS.
8. Enable the previously disabled "Fast Boot" related settings again. Save and exit, boot Windows

This should now train the memory with high temperatures, so essentially under a worst-case condition. This will eliminate the problem coming from instability caused by elevated temperatures.

If this doesn't help, there's multiple other options what could be wrong. The overclock might simply not be stable, and stress tests might not be able to pick up on it. The overclock might become unstable under certain conditions other than temperature and, again, stress tests might not be able to pick up on it. It might be that the memory is stable in stresstest specific workloads, but a game's different characteristic makes it unstable, so it's not visible on a stresstest.

I think you can pretty much 100% rule out a hardware level defect of the memory modules, because if that was the case, a stress test would pick up on it and also your computer wouldn't be working at all, but constantly crash whenever your CPU tried to access a broken "segment" of memory.

For now, try the above steps with the RAM set you describe as "having problems".

[Slender]

voltage difference

[Fpsgamer]

Ram is a usual suspect on these kind of issues.
I had issues when I had a kingston kit on my 7600x untill I changed to a g.skill one.

I changed to 7800x3d a while back and I had stutter issues which I narrowed down to ram aswell.
Even tho I have the same ram that worked perfectly on 7600x, 7800x3d is more sensitive to ram.

[Kipperii]

What is the core cause of this RAM problem? I think it is worth digging deeper.

[Sensuki]

I've been trying to fix mine for years, I think I've finally had some success although not with the settings that I expected. It's the combination of settings that work together.

I am using KirbyOS and I have applied some other tweaks to the system as well, so a lot of the things probably help to reduce the general system /DPC latency which is quite low now - ran Latency mon and was averaging 1.8us DPC latency.

This is what is working for me

Windows 11 23H2 (KirbyOS)

edited because I got better results with the following

Multiplane Overlay disabled - MPO On is SUPER laggy for me. Night and Day difference on and off
HPET OFF in BIOS - With MPO on, I was playing with HPET On as it felt better. With MPO Off, HPET Off feels way better. Can feel the cursor delay that HPET On adds.
----------
BCDEdit Default settings (no tweaks). I played with Disabledynamictick for years, but came back to default recently and it's a much more consistent mouse feel
Fullscreen Optimizations disabled
Game Mode OFF - ON adds mouse lag for me
NVIDIA Low Latency Mode Ultra
Win32PrioritySeparation 2A
Monitor Overdrive OFF (at least on my monitor, ON adds latency)

[kriegsnake]

If you suspect your RAM kit to be a problem, you may run a simple test to rule specific things out. As you probably know, RAM stability is dependant on temperature. What can happen is that you've overclocked your RAM one day, let it train itself while the temperature was pretty low, and since there's "Fast Boot" enabled it will skip through some training on boot up, essentially "locking in" the once trained properties. When you're playing a game and the inside of your PC heats up more, due to the GPU dumping heat for example, the once stable memory might become unstable with elevated temperatures. To eliminate this specific cause, you may run a simple test:

I always had fast boot disabled in my bios, is your test applying to the case when it retrains itself after each boot?

@Kipperii, i thought you had a problem with your display unit, so somehow ram did a better outcome?

[ChristophSmaul1337]

I always had fast boot disabled in my bios, is your test applying to the case when it retrains itself after each boot?

Yeah so what that does is train your memory on every single boot. You're probably (maybe?) shutting down the PC when you go to bed, or go out to work, and when you return to the computer to turn it back on, the memory is at room temperature at that point, so it will train to what it thinks is safe and stable at that room temperature. As "Fast Boot" is constantly disabled, it will re-train itself at room temperature on every single boot. Now think about some edge case scenario where you might be just about stable with XMP at room temp but as soon as the sticks go above, let's say 40°C, they'll be come ever so slightly unstable. That's a realistic scenario in my book.

The goal with that test procedure I showed earlier is to have the memory train itself at pretty much the worst-case scenario, with temps upwards of 50°C or so. Then, you'd enable "Fast Boot" to kind of "lock in" the trained timings, so that the memory will always operate with the worst-case trained timings. It's specifically designed to not re-train itself every single time but stick with what it's been going to when temperatures weren't ideal. This is to create consistency in testing. It's not meant to fix an issue or be a solution. However, when troubleshooting a problem, it's always useful to single out as many variables as possible, and doing that procedure is a good way to rule out RAM - temperature related instablilty. Hope all of that makes sense.

[dervu]

So was it happening also with XMP off? That would be even weirder.

[Fpsgamer]

So was it happening also with XMP off? That would be even weirder.

XMP/EXPO off is usually not a solution since you most likely gonna lag without it, finding stable speed inside XPM/EXPO is.