Page 4 of 5

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Sat Jan 16, 2021 6:12 pm
by Kjetil
Running 441.66 om my 2080TI(24/7 folding)

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Thu Jan 21, 2021 9:28 am
by VxJasonxV
I quit GOG Galaxy a few nights ago, ultimately no change. I shutdown both Galaxy and Folding, uninstalled Galaxy, opened Folding back up, no change.

The reason why I never answered the question about my PSU is that I just didn't know what it was. According to my documentation it is a 750W 80+ Gold PSU, no mention of the manufacturer. The reason why I didn't know what it was is because my case (Corsair Carbide 275R) has a PSU lane enclosure, and I haven't done any research to see how to open it up. I assume it's quite simple but I hate cable management and really don't want to go mucking around more than I have to, so I haven't.

I can see the video card power cable coming out from the PSU enclosure's vent, exit, thing, and the second 8-pin connector (for SLI, I assume?) also plugged into the second 8-pin port, filling all 16 pins with the same single cable. I suppose the next step for me is to change cables, assuming I have a second cable on the PSU. I believe this isn't common unless it's rated for SLI? And I also don't think 750W is rated for SLI? I'm quite out of the hardware details game. If I have two cables are you suggesting that I plug the first output of each of them in?

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Thu Jan 21, 2021 10:29 am
by ajm
A PCIe power cable (with 6 or 8 pins) can only provide a certain amount of power, up to 75 W for a 6pin cable and up to 150 W for an 8pin cable (with a good PSU). And the PCIe slot of the motherboard also provides max. 75 W to the Graphics card. If your Graphics card can draw more, it needs to have a second (or third) 6 or 8pin PCIe cable.

All connectors of the Graphics card must be populated for it to simply work, as being recognized by the system. Otherwise, the PC won't post or (if you have some other Graphics) will show some error. But that is just the start. Afterwards, the power also needs to be there when needed. And this may be the cause of your problems. Because running FAH definitely will require the max. amount of power the card can use.

Now, if I understood correctly, you have plugged both PCIe connectors of the Graphics card with the same cable, that happens to have two connectors. It might be okay, but it also might be insufficient, that is, if the cable provides 150 W, or two times 75 W, and your GPU awaits two times 150 W, it will crash during power intensive tasks.

This has nothing to do with SLI. Each installed Graphics card must have its own power delivery satisfied, with or without SLI. SLI will only allow to bundle several GPUs for computing specific tasks, and only when the used software is programmed for it. SLI is pretty munch dead by now, as Nvidia stopped developing it.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Thu Jan 21, 2021 10:41 am
by VxJasonxV
I shouldn’t have even mentioned SLI, to be completely clear I don’t have two video cards.

But yes, all 16 ports are plugged in via 2 8-pin connectors off a single PSU cable.

Guess it’s time to tear things up and look for a second cable. There is no cable unused coming out of the PSU enclosure, if I do have a second cable it’s likely that it is wrapped up and sitting unused in the enclosure.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Thu Jan 21, 2021 10:51 am
by ajm
You can access those cables from behind. The PSU shroud of the Carbide 275R is open on the rear side. Just take off the back panel and you will figure it out easily.
And, yes, you should find another 8pin PCIe cable in the material that came with the PSU.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Fri Jan 22, 2021 3:27 am
by VxJasonxV
https://i.imgur.com/xHJWS8o.jpg (Warning: HUUUUUUUUUUUUUUUEG image. Man, I should have given this thing a quick air blast while I had it open…)

Highly appreciated, ajm. Fingers crossed for a week of uptime. If I make it that far I'll re-enable CUDA too.

I have the picture on hand but if for any reason I need it again, but in case it matters, my PSU is MPY-7501-ACAAG. Couldn't actually see any logos, but it's purportedly a CoolerMaster 750 (but not the -US version?).

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Fri Jan 22, 2021 6:14 am
by ajm
:D Glad it worked out so far!
Do let us know!

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Tue Jan 26, 2021 6:21 am
by VxJasonxV
Just sat down at my computer and it was once again sleeping, and in the logged out state. :/ No CUDA for me. I'm at a loss of what else to do. I could keep using Afterburner to limit the power of the 2080, it survived under 40% power, that might allow it to actually continue to complete works. All of this is predicated by the fact that I want to keep folding. afaik the log outs aren't affecting anything, if I needed something to run persistently even when locked / idle I would probably just exit Folding for that night. But I do want to keep folding in general, it's also the general degradation over time that bothers me.

I posted a log from something that showed it continuing to draw max power, right? And I am confident I'm not having any physical hardware problems, because nothing else of mine in this room, on that same plug, is failing. In order to totally rule that out I should check my EFI to see what the "wake on power loss" setting is, set it to "remain off", but the fact that I can watch it happen with no other device effected makes me doubt this is relevant.

So, what else is left besides seeing if I can still act on warranties, and how can I prove with a reasonable degree of certainty which component (video card, PSU, hopefully nothing else) is actually to blame?

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Tue Jan 26, 2021 6:40 am
by ajm
I looked up the thread and didn't find anything about the CPU. Could it be that it's not sufficiently cooled or that the cooling is not properly installed? How about the CPU temperatures (idle, under load)?

Otherwise, all I can think of now is to swap the possible culprits (PSU, GPU, CPU, RAM) to isolate the issue. Is this a prebuilt PC? If so, maybe your vendor would help you with that.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Tue Jan 26, 2021 8:23 am
by VxJasonxV
There's a lot of anecdotal evidence in both directions that it has to do with temperature. It's been pretty cold here and I had my window fully open in below freezing weather. Since the system was sleeping up until a few hours ago it was cool along with the rest of the room. Warmed up from use while I made that post, and then crashed again shortly afterward while I was on the phone for just over 30 minutes, admittedly this was "from hot". I had locked my screen out of habit which is when Folding starts up. All this after 3 days of uptime and Folding cranking for that whole period of time though? Would be very strange that this is purely temperature related.

It is a prebuilt, from a vendor I had never dealt with before, it was recommendation from a local LAN center. I haven't investigated any support dealings from them yet.

I haven't done anything to measure the temperature of the CPU, I doubt I even have software on my system for that. You saw the general state of dust in my photo a few posts back. Present but not caking. I know I've seen temperature stats in the EFI but I don't think I have any software for this in the OS itself.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Tue Jan 26, 2021 8:29 am
by ajm
You can install HWMonitor: https://www.cpuid.com/softwares/hwmonitor.html
It gives you the present temp and the min and max temps for all connected sensors in your PC. That will tell us rapidly whether there is a temperature issue.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Wed Jan 27, 2021 4:56 am
by VxJasonxV
I see a bit of an issue with HWMonitor, that I have to be there to take a reading. Restarting the program reinitializes the maxes, and saving is a user-invoked option.

This isn’t insurmountable since I’ve mentioned that this happens when logged in just by allowing Folding to run. When I’m done here I can just turn it loose and take periodic snapshots.

I will also get a baseline of Desktop idle and while playing a game here in a few. I appreciate your continued insight.

Mostly unrelated; I’ve thought of getting an outboard video capture recording device. Had I ever done that, recording all this would be quite a bit easier since it wouldn’t fail when the OS resets in whatever form it does.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Wed Jan 27, 2021 6:24 am
by ajm
We only need HWMonitor to see whether your temps are okay. You don't have to let it run the whole time for this.
It's just a matter of minutes, looking how the temps evolve at idle and under load. If the temps are okay, we can look elsewhere. If there is a temp problem, we can try and solve it. That's it.

I'm not familiar with video capture devices and I don't know if we could gather useful troubleshooting infos with it.
If you need it anyway to record your gaming, go ahead. But I wouldn't buy one just for troubleshooting.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Wed Jan 27, 2021 9:28 am
by VxJasonxV
Oh, I see. The "Save" just dumps a hardware profile, doesn't save the metrics. Good thing I looked at it before I just expected it to save a snapshot of current values. Then the best way to get them is to document / screenshot the window? Good thing I have a display in portrait…!

https://i.imgur.com/ESTPhsX.png

Despite both cores running right now, not even the video card's blower is dialing up its speed as it does fairly often. I'll come back in an hour or two if I'm still awake / after I hear the blower starting to do its thing.

I did look at my UEFI, and I can control the power of the system's fan by 4 levels: Silent, Standard, Performance, and Full Speed, and I'm on Standard. All of my fans are spinning, doing their thing, no whines, nothing particularly troubling I can see.

Apparently I have liquid cooling? Which I know nothing about. Always interested me, but never had it in a computer previously.

https://i.imgur.com/lAX1ulK.jpg

Anyway, yeah, pretty high maxes already. I'll see if they're any worse after I hear the blower kick in.

Re: F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

Posted: Wed Jan 27, 2021 9:47 am
by ajm
You can click the little square on the left side of the elements to hide everything but the temperatures. So you have a clearer view of what you want.

Here, the CPU seems to be well cooled. You have a so-called AIO water-cooling. There is a pump in the block on the CPU, that circulates water through the tubes in the radiator at the front of the PC.

But I cannot make sense of the two pictures I have of this radiator. On https://imgur.com/xHJWS8o it looks like you have a radiator on the inner side. And on https://imgur.com/lAX1ulK it looks like you have the radiator of your AIO on the outer side, or ventilated only on the upper side. ??? Can you make a picture of the whole radiator?

But most importantly, the 2080ti is very hot, at 86°C for 80% of the max. wattage. Can you check below the Graphics card whether all fans are indeed turning? Maybe one is blocked by something.
It also would be a good idea to take it off and give it a good blow.