GPU OC Hang

[HoBGoBLiNzx3]

It's known and warned on ED's support page that GPU overclocking can result in crash issues. I previously had a MSI afterburner clock on my RTX2060 it wasn't very high, and I can't remember the actual numbers but...

I was OCing my i7-9700K (5hz @ 1.4 -2 I believe) stable, disabled all CPU power saving features. (Runs great)

So I decided to stress test my GPU on Kombustor.. 

I was able to get a stable +140/+1200 clock with no crashes, but decided on a safer clock because of DCS (VR might I add)

So I set the clock to +100/+1000. Single player booted instantly, and I was getting smooth 60fps (1.5 @ 90hz) with a 3960ER and 500BR. No SS anywhere.

No crashes, No problem!?

Except joining MP servers hangs on mission load splash... So I reduced the clock... nothing worked. Not even the smallest of increased clock speed or memory. Leaving me no choice really but to abandon the OC all together... Which is less than ideal, considering the stable nature of the clock, and the quality and performance that was achieved in SP. 

Now I'm well aware of the CPU load on MP but I'm failing to see the connection between GPU performance and SP/MP performance when it comes to at least clock speeds and memory????????

Can someone explain why this is?... 

and how to get the a stable OC to not hang/CTD in MP??????

 

[BitMaster]

Yes, this sounds a bit off to me as well but if it is reproducabe than there must be some kind of dependency of the two, or whatever module(s) involved.

Anyway, if you can enter MP with stock settings, do that, ALT-TAB out once in the cockpit and set Afterburner after loading into MP. That would navigate around the issue.

Not elegant but may work until you find the root cause.

 

If you know Afterburner, set the OSD to show all relevant data so you might get an idea what's causing this weird issue. If you dont know Afterburner well enough, learn it

*usually, what kills an overclock is too high temps. Better overclock through undervolting rather than only pushing the sliders to the right and hope for the best. That can work but you need a very good cooling ( not air usually then ). Undervolt it. There are good threads in here that also link to YT vids regarding this topic. In your case I would look into that.

Edited August 16, 2022 by BitMaster

[HoBGoBLiNzx3]

1 hour ago, BitMaster said:

Yes, this sounds a bit off to me as well but if it is reproducabe than there must be some kind of dependency of the two, or whatever module(s) involved.

Anyway, if you can enter MP with stock settings, do that, ALT-TAB out once in the cockpit and set Afterburner after loading into MP. That would navigate around the issue.

Not elegant but may work until you find the root cause.

 

If you know Afterburner, set the OSD to show all relevant data so you might get an idea what's causing this weird issue. If you dont know Afterburner well enough, learn it

*usually, what kills an overclock is too high temps. Better overclock through undervolting rather than only pushing the sliders to the right and hope for the best. That can work but you need a very good cooling ( not air usually then ). Undervolt it. There are good threads in here that also link to YT vids regarding this topic. In your case I would look into that.

 

Will definitely look into this. Yeah it is odd.. What is more weird is ED states that not only does OC afterburner cause problems but stock OC cards do too...!

[fagulha]

32 minutes ago, HoBGoBLiNzx3 said:

Will definitely look into this. Yeah it is odd.. What is more weird is ED states that not only does OC afterburner cause problems but stock OC cards do too...!

I needed to underclock my stock OC GPU because i found out that it crashes DCS and Elite Dangerous. The only two games that crashes if i leave it stock OC....

[Hiob]

How the processor or the ram is stressed exactly depends on the application. Running stable in one stress test never guarantees stable run in everyone.

GPU - overclocking isn‘t a problem usually, if you watch the temps. VRAM overclocking on the other hand is a big NO for DCS. I don’t know WHY exactly DCS is so sensitive about it, but I can confirm from multiple tests that it is. 

[HoBGoBLiNzx3]

53 minutes ago, Hiob said:

How the processor or the ram is stressed exactly depends on the application. Running stable in one stress test never guarantees stable run in everyone.

GPU - overclocking isn‘t a problem usually, if you watch the temps. VRAM overclocking on the other hand is a big NO for DCS. I don’t know WHY exactly DCS is so sensitive about it, but I can confirm from multiple tests that it is. 

No I totally agree but It was crashing at splash at +40/+400... When It was easily able to handle +100/+1000. I'm trying to figure out exactly that. WHY is DCS so sensitive to VRAM clocking.. 

[Hiob]

8 hours ago, HoBGoBLiNzx3 said:

No I totally agree but It was crashing at splash at +40/+400... When It was easily able to handle +100/+1000. I'm trying to figure out exactly that. WHY is DCS so sensitive to VRAM clocking.. 

Good Luck!

Just note that my 3080 is running on a custom waterloop (around 40° under load) and I can easily run the GPU with 2100 MHz when ambient temps are reasonable, but even +300 MHz on the VRAM don't work reliable. Since 300 and below are not even noticeable in performance anyway, I skipped the VRAM OC completely.

[HoBGoBLiNzx3]

3 hours ago, Hiob said:

Good Luck!

Just note that my 3080 is running on a custom waterloop (around 40° under load) and I can easily run the GPU with 2100 MHz when ambient temps are reasonable, but even +300 MHz on the VRAM don't work reliable. Since 300 and below are not even noticeable in performance anyway, I skipped the VRAM OC completely.

I'll likely keep my Vram untouched and just push the core +60 or so and call it a day. Then I don't have to worry about it. I'd push the core harder but it's on fan only and I can't see a 2060 getting any real noticeable frames that would make a difference. It does smooth out the 90hz on the headset though which is what I was after in the first place. 

The 120hz was a dream until 5 mins later when it was pushing 120 degrees. lol. I was actually getting better frames with 120hz than anything, and no ghosting, nodda. It's a little off topic but was the reason why I was OC it in the first place.

[Hiob]

17 minutes ago, HoBGoBLiNzx3 said:

I'll likely keep my Vram untouched and just push the core +60 or so and call it a day. Then I don't have to worry about it. I'd push the core harder but it's on fan only and I can't see a 2060 getting any real noticeable frames that would make a difference. It does smooth out the 90hz on the headset though which is what I was after in the first place. 

The 120hz was a dream until 5 mins later when it was pushing 120 degrees. lol. I was actually getting better frames with 120hz than anything, and no ghosting, nodda. It's a little off topic but was the reason why I was OC it in the first place.

Instead of setting an +60 offset to the core, you should undervolt via the curve editor. You can probably run +60 Hz AND lower your temps. Look at my other threads or search for a video on youtube on how to do it.