Ram OC causing GPU instability?

[SnR3]

Hi guys I OC my ram to 3800 so I can set fclk to 1900 yesterday and everything looked just fine.
I followed the ryzen dram calc guide both safe and fast options posted normally and stress test did not reveal any instability with ram. I even later managed to raise it to 3866 and fclk 1933 temporary without issues again. I than decided to test it by playing some games. I even started OC my GPU to find some limits on it. At some point I started getting crashes with direct X error. No bsod no restarts just crashes of game. Decided to run some stress test and noticed during aida64 stress test GPU fails each time within 5-15 minutes. Is it actually possible that the ram OC can be causing such GPU behaviour? Or is it the fclk itself? I was unable to make the you stable again. I updated drivers while clearing previous 1st using ddu. Reverted any settings related to GPU oc. Set everything in drivers to default no help. Than I started reverting ram OC in bios. Safe settings did not help with issue. I started having concerns than maybe I just find out my GPU is failing (quite new unit just month old)
The issue has been solved when I reverted bios to optimal settings. Running aida64 as u read it and it's 1.5h and going with no issues. I'll try more demanding stress test overnight. Now I'm kind of puzzled on what to think about it. If you want I can supply info about my OC timings if that would help diagnose the problem. Listing my hardware below :

Gigabyte aorus x570 master (rev 1.1 with 31k beta bios)
Ryzen 5900x (no OC just PBO turned on) cooled with arctic liquid freezer II 420 (temps no issue)
Amd 6800xt MBA (GPU temps highest observed at 57° and 78° at junction)
G.SKILL Trident Z - F4-3600C17D-32GTZR 32 GB (16 GB x 2)
Sabrent Rocket gen4 1tb - 3 of those
PSU 1000w gold from Corsair.

I highly doubt temperatures are any issue as I've massive case and its really well cooled I've 420mm radiator on AIO and 4 120mm fans there is a lot of room and temperatures are really low on every component compared to any other benchmark temps u can find online.

Any help would be appreciated

That's settings in bios:

https://ibb.co/zbH1Zkn
https://ibb.co/1z6JPVJ
https://ibb.co/zX0XJc2

Edit: ok so FCLK is not an issue running fclk 1900 and ram at 3800 with default timings gives no faults so far during stress test. Will run some games now

 

[pc-guy]

Did you run memtest86 to see if the ram was stable? Also check the Viod and Vccd etc to see if they are set correctly by motherboard. Your Vsoc looks like 1.1V may need a bit more so Viod and Vccd should be around 1.05V. Probably yank the vram to 1.4v.

also leave all your timings to auto post and test. If all passed then start tweaking the primary timing.

Wouldn’t change any of the CAD bus timing or drive strength. Just leave it all on auto. Mucking about with it will probably cause more issues.

what I found about that calculator is that it has way too much settings in there.

 

[SnR3]

Hi thanks for the answer.
Yes I tested ram with memtest there was no errors.
I think I had set vccd even higher as I experienced some audio stuttering at 1.05
Leaving timings on auto fixed the issue with GPU instability I already mentioned it in 1st post but to be honest the whole point of ram OC is to try get those timings lower right?
Aida64 was showing ram at 57.6 ns witch was quite decent.
Ur saying u should put on auto all the drive related settings? Wouldn't it create some more issues?

 

[pc-guy]

Hi thanks for the answer.
Yes I tested ram with memtest there was no errors.
I think I had set vccd even higher as I experienced some audio stuttering at 1.05
Leaving timings on auto fixed the issue with GPU instability I already mentioned it in 1st post but to be honest the whole point of ram OC is to try get those timings lower right?
Aida64 was showing ram at 57.6 ns witch was quite decent.
Ur saying u should put on auto all the drive related settings? Wouldn't it create some more issues?

What I am saying is you should leave everything auto on the first go. So the system post and you can test that CPU and ram are stable at those clocks.

then start mucking about with the timings doing the primary first then secondary then the other settings.

it is a labour intensive process. Require loads of cmos reset. There is no one step short cut really. The calculator gives you guidance or ballpark mark where you can aim. It is not definitive. It really need a lot of attention and patience to get your ram tuned and working properly.

 

[Brizzles]

Problem is you've tightened your RAM , increased Fclk and o/c gpu all in one day, so the issue could be any or all.

You should do it in increments to make sure things are properly stable, i know from experience that you want it done quickly and what you deem stable in the excitement , is not always stable.

Start with Fclk , ram at XMP - get that stable for 24 hours - a single Cinebench run is not stable for instance. Cinebench multi, Prime , Gaming, anything else you think will stress it.

Then RAM, test after each change, or you'll never know which timing is at fault in case of an error. Again a quick pass of Memtest is not stable - I use Prime , TM5 full run and Memtest to stabilty test - this can take days for a full tune.

Lastly the GPU - that should be pretty simple after the above, as you know your system is stable.