Random Crashes while folding, not sure how to diagnose?

[bwalker36]

So I recently set up the F@H client and I Have gotten two crashes since yesterday morning, roughly a 24h span.  Now it is 100% possible I was getting these before and never noticed cause I never had something running the whole time.  My gpu is about 75c and CPU is 76-80C depending on when the air kicks on.  Neither are overclocked at the moment.  

 

The only problem I have had with this rig is my EVGA audio card, the drivers crash all the time and the LEDs are stuck pink.  So I'm initiating an RMA on that, I"m sure they will make me do some things before they grant it but once they do I'm ripping it out and using onboard to see if that resolves it. I'm not using the OC bios on my FTW 3080 so I feel like the 750W should be enough and if it wasn't I feel like it should crash much sooner than 12 hours later of running full tilt.  I did look in even viewer and there is a bunch stuff in ther I don't understand.  Here is a snippit:  I'm also concerned that there are a ton of other errors and warnings.  I filtered out warnings.  

 

 

 

[Avacado]

Are you folding on both (CPU and GPU)? Are you making sure to leave a few cores NON-Folding for other processes?

[bwalker36]

18 minutes ago, Avacado said:

Are you folding on both (CPU and GPU)? Are you making sure to leave a few cores NON-Folding for other processes?

Yes, and No lol.  

 

It actually just crashed again.  

 

Since I got to see it happen there is not crash or errors, its just power cycles.  Similar to losing power for a split second.  

 

Edit again:  I set the core usage to 10 so its running at about 75% CPU usage.  Will see how that goes.  

Edited June 23, 2022 by bwalker36

[Avacado]

Do you happen to have a back-up PSU to test with?

[bwalker36]

2 minutes ago, Avacado said:

Do you happen to have a back-up PSU to test with?

I don't but I also am interested in moving to an 850w or higher to really egage the OC bios on the 3080 ftw.  My understnding is this 750 would be stretched thin trying to pull 400w or whatever that bios is.  

[The Pook]

drop the power limit or limit the GPU to something like 0.7v or lower. Go away? 

 

I had to retire my Focus Plus 750W when I got my 3080 Ti. I was staying under 750W but Seasonic is too sensitive to transient spikes and made my system act like someone flipped the PSU switch.

 

I bitched about it on OCN in the 3080 Ti thread, fix was a new PSU. 

 

Please Wait... | Cloudflare

WWW.OVERCLOCK.NET

 

In order to use my 3080 Ti until I got a new PSU I had to set power limit below 50% (IIRC 40%). Locking voltage worked better (better performance) but I don't remember what voltage that was. 

[bwalker36]

4 minutes ago, The Pook said:

drop the power limit or limit the GPU to something like 0.7v or lower. Go away? 

 

I had to retire my Focus Plus 750W when I got my 3080 Ti. I was staying under 750W but Seasonic is too sensitive to transient spikes and made my system act like someone flipped the PSU switch.

 

I bitched about it on OCN in the 3080 Ti thread, fix was a new PSU. 

 

Please Wait... | Cloudflare

WWW.OVERCLOCK.NET

 

In order to use my 3080 Ti until I got a new PSU I had to set power limit below 50% (IIRC 40%). Locking voltage worked better (better performance) but I don't remember what voltage that was. 

I never had a problem until trying to fold as it its the first time the system was really pushing both at 100%.  Going to see how limiting the CPU works, then I might power limit the GPU to and that solves this incoming new PSU hah. 

 

Actually now that I think about it my PSU makes more sense.  I think the crashes are when the CPU is running, the GPU work unit completes then the new starts and that spike with the CPU already at 100% is tripping it.  Every time I have turned it back on I feel like a GPU work unit was just starting.

Edited June 23, 2022 by bwalker36

[Avacado]

9 minutes ago, The Pook said:

drop the power limit or limit the GPU to something like 0.7v or lower. Go away? 

 

I had to retire my Focus Plus 750W when I got my 3080 Ti. I was staying under 750W but Seasonic is too sensitive to transient spikes and made my system act like someone flipped the PSU switch.

 

I bitched about it on OCN in the 3080 Ti thread, fix was a new PSU. 

 

Please Wait... | Cloudflare

WWW.OVERCLOCK.NET

 

In order to use my 3080 Ti until I got a new PSU I had to set power limit below 50% (IIRC 40%). Locking voltage worked better (better performance) but I don't remember what voltage that was. 

Thats where you went wrong. Asking a serious question on OCN.

[The Pook]

11 minutes ago, bwalker36 said:

I never had a problem until trying to fold as it its the first time the system was really pushing both at 100%.  Going to see how limiting the CPU works, then I might power limit the GPU to and that solves this incoming new PSU hah. 

 

Actually now that I think about it my PSU makes more sense.  I think the crashes are when the CPU is running, the GPU work unit completes then the new starts and that spike with the CPU already at 100% is tripping it.  Every time I have turned it back on I feel like a GPU work unit was just starting.

 

the issue isn't power consumption, it's high transient spikes caused by the 3000 series . I tried lowering my CPU OC too, didn't do anything. 

 

guess it's a way to test though 

 

3 minutes ago, Avacado said:

Thats where you went wrong. Asking a serious question on OCN.

 

but you're on OCN 

[Avacado]

1 minute ago, The Pook said:

but you're on OCN 

I don't ask questions there, it's purely for entertainment value. 

[bwalker36]

6 minutes ago, The Pook said:

 

the issue isn't power consumption, it's high transient spikes caused by the 3000 series . I tried lowering my CPU OC too, didn't do anything. 

 

guess it's a way to test though 

 

 

but you're on OCN 

 

My thought was that the spike when it kicks on with the CPU already running would trip it but maybe not i'm not smart lol

 

[The Pook]

as long as you're smarter than that weird Avacado guy you're probably okay 

 

from what I understand the quick spike is what causes the power cycle, the value it spikes to is irrelevant. if Seasonic OCP/OVP trips when you spike 200w in a ms it doesn't make a difference if it spikes from 100w to 300w or 500w to 700w, it should still trip. 

[bwalker36]

1 hour ago, The Pook said:

as long as you're smarter than that weird Avacado guy you're probably okay 

 

from what I understand the quick spike is what causes the power cycle, the value it spikes to is irrelevant. if Seasonic OCP/OVP trips when you spike 200w in a ms it doesn't make a difference if it spikes from 100w to 300w or 500w to 700w, it should still trip. 

That makes sense, thanks.  Did you ever try and get seasonic to send a replacement.  I may reach out and see what they do lol.  

[The Pook]

I did but they more or less said I didn't know what I was talking about. I sent them threads of people complaining about the same issue and then they just stopped responding. 

 

mine is just in my closet collecting dust, if I RMA then I'll just get the same model back anyway which doesn't really help anything. 

[bwalker36]

So capping my CPU folding at 10 threads seems to have removed my issue.  I'm just under 24hours without a crash and honestly it barely touched my PPD (didn't realize jut how much better GPUs are).  I think I'm still gonna RMA the PSU to see what happens, and buy a bigger one so I can finally play with the OC bios on this 3080 FTW and figure out how to OC alder lake.  Now begins the process of figuring out what PSU.

[Avacado]

22 minutes ago, bwalker36 said:

So capping my CPU folding at 10 threads seems to have removed my issue.  I'm just under 24hours without a crash and honestly it barely touched my PPD (didn't realize jut how much better GPUs are).  I think I'm still gonna RMA the PSU to see what happens, and buy a bigger one so I can finally play with the OC bios on this 3080 FTW and figure out how to OC alder lake.  Now begins the process of figuring out what PSU.

Iv'e got an extra Platinum 1000w Super Flower i'd be willing to sell at a good price.

[Sir Beregond]

On 23/06/2022 at 08:09, The Pook said:

as long as you're smarter than that weird Avacado guy you're probably okay 

 

from what I understand the quick spike is what causes the power cycle, the value it spikes to is irrelevant. if Seasonic OCP/OVP trips when you spike 200w in a ms it doesn't make a difference if it spikes from 100w to 300w or 500w to 700w, it should still trip. 

Gamers Nexus just recently put out a video on this and I guess the 30-series is super bad with its transient spikes and PSU manufacturers only expect it to get worse with the 40-series.

[Avacado]

17 minutes ago, Sir Beregond said:

Gamers Nexus just recently put out a video on this and I guess the 30-series is super bad with its transient spikes and PSU manufacturers only expect it to get worse with the 40-series.

I watched it too. Wasn't overly concerned as all of my PSU's are 1000/1600w plats. 

Edited June 24, 2022 by Avacado

[iamjanco]

On 23/06/2022 at 09:50, Avacado said:

Thats where you went wrong. Asking a serious question on OCN.

 

On 23/06/2022 at 09:55, The Pook said:

but you're on OCN 

 

23 hours ago, The Pook said:

they more or less said I didn't know what I was talking about.

 

 

On 23/06/2022 at 09:57, Avacado said:

I don't ask questions there, it's purely for entertainment value. 

 

  

[Sir Beregond]

13 minutes ago, Avacado said:

I watched it too. Wasn't overly concerned as all of my PSU's are 1000/1600w plats. 

Yeah, I am glad I decided to just grab a 1000W.

[iamjanco]

I've got a Seasonic 1300W Platinum in my current build and I've left the twisted triplet and added cap on the 12VDC line out of the cable I'm working on.

 

 

If need be, I've got a 1600W SuperFlower I can replace it with.

 

 

[damric]

One of the best things I learned about running folding@home, it's brutal. If there is even the slightest instability, it's going to let you know. Consider these monthly foldathons a very good stress test. It runs AVX very hard on the newer CPUs, and I've had a couple Ryzens that I had to turn off the boosting so that it would not crash. It's sensitive to memory overclocks in that it hates errors, particularly VRAM. For GPUs I have to dial way back the overclocks that are stable for 3D rendering, as they just aren't as stable crunching proteins. For the 48 hour foldathons, especially in the summer it's a good idea to tune for low power efficiency so you don't run up the light bill and heat up the room so much. Weak power supplies will also not manage so well. Make sure they have cool air feeding them.

[bwalker36]

38 minutes ago, damric said:

One of the best things I learned about running folding@home, it's brutal. If there is even the slightest instability, it's going to let you know. Consider these monthly foldathons a very good stress test. It runs AVX very hard on the newer CPUs, and I've had a couple Ryzens that I had to turn off the boosting so that it would not crash. It's sensitive to memory overclocks in that it hates errors, particularly VRAM. For GPUs I have to dial way back the overclocks that are stable for 3D rendering, as they just aren't as stable crunching proteins. For the 48 hour foldathons, especially in the summer it's a good idea to tune for low power efficiency so you don't run up the light bill and heat up the room so much. Weak power supplies will also not manage so well. Make sure they have cool air feeding them.

This was at stock too, well memory was set to XMP.  I am leaning heavily on those transient spikes and my PSU.  But AVX explains the heat on my CPU i was wondering how it was approaching 85c when I was letting it go full throttle.  

[Sir Beregond]

1 hour ago, damric said:

One of the best things I learned about running folding@home, it's brutal. If there is even the slightest instability, it's going to let you know. Consider these monthly foldathons a very good stress test. It runs AVX very hard on the newer CPUs, and I've had a couple Ryzens that I had to turn off the boosting so that it would not crash. It's sensitive to memory overclocks in that it hates errors, particularly VRAM. For GPUs I have to dial way back the overclocks that are stable for 3D rendering, as they just aren't as stable crunching proteins. For the 48 hour foldathons, especially in the summer it's a good idea to tune for low power efficiency so you don't run up the light bill and heat up the room so much. Weak power supplies will also not manage so well. Make sure they have cool air feeding them.

Well when I eventually move and my main PC moves to a room that is not my bedroom, will be looking forward to participating in these.

[damric]

1 hour ago, Sir Beregond said:

Well when I eventually move and my main PC moves to a room that is not my bedroom, will be looking forward to participating in these.

I don't blame you. My office is always hot because of the machines running. I'm working on a solution for that at least for my hottest rig. I'm going to install the radiator to blow out the window. I already got the long hoses and stuff I just need to put it together.