Troubleshooting request - 3080 FE reboots

B

brentsg

[H]ard|Gawd
Joined
Mar 8, 2003
Messages
1,746
I could use some eyes on this before I do an RMA. Any thoughts greatly appreciated.
  • Case: Meshify C with reasonably good cooling (2x 140mm front intake, 1x 120mm rear exhaust)
  • GPU: 3800 FE
  • CPU: 9900k (stock boost of 4.4, but allowed indefinitely)
  • Motherboard: Asus Maximus Hero Z370
  • RAM: G.Skill 32GB (4x8) - 3866 XMP
  • PSU: 1000W Seasonic Prime Titanium (separate PCIE to GPU)
  • OS: Win10 64 Pro Current
Problem: 1-2 spontaneous reboots per day when Folding@Home. No graphical glitches or similar.

Troubleshooting:
  • Clean DDU driver re-install
  • Boosted power limit to max, reboots within minutes
  • Power limit at stock, reboots randomly but just 1 or so per day.
  • Underclocked / undervolted to 1875 / 0.875V - Still reboots but less frequently.
  • Clean Folding reinstall
  • Computer perfectly stable in this config for years with previous 2080Ti / 2080 Super.
I run aggressive enough fan profiles and nothing is exceeding 65-67C or so.

The only other PC I can drop this in is a Ghost S1 so that's not really viable. I did look to see how hard I'm pushing my 1500VA / 900W Cyberpower UPS. Looks like I'm peaking around 750W due to PC + other connected eqpt.

Am I missing anything? I hate to go thru the RMA process if there's some chance I have a PSU defect or otherwise. It bothers me that I'm not seeing any graphical artifacts or similar.
 
If I were to guess, I'd say that Folding@Home needs an update for the new GPU architecture. Everything else seems fine.
 
DooKey said:
Reboots are usually PSU and sometimes RAM. This doesn't sound like a GPU issue to me. This current gen of GPU's from NV seem to stress the PSU more than before.
Click to expand...
This is one of my concerns. Obviously this PSU should be just fine with this load but here I am. I was hoping to get NV to cross-ship for me but that's not happening.

I'm not worried about RAM because I can independently test that and it's fine. Plus I've had the RAM/MB/CPU combo going fine for so long. The fact that I can almost immediately trigger the reboot by pushing the power headroom indicates PSU or GPU. The fact that it still happens with the card undervolted to ~280W gets me iffy on the PSU part.
 
brentsg said:
This is one of my concerns. Obviously this PSU should be just fine with this load but here I am. I was hoping to get NV to cross-ship for me but that's not happening.

I'm not worried about RAM because I can independently test that and it's fine. Plus I've had the RAM/MB/CPU combo going fine for so long. The fact that I can almost immediately trigger the reboot by pushing the power headroom indicates PSU or GPU. The fact that it still happens with the card undervolted to ~280W gets me iffy on the PSU part.
Click to expand...

Can you underclock your CPU/memory, etc. to lower the system power draw to see if that would help your video card? Might be able to isolate the PSU vs. the GPU.
 
brentsg said:
This is one of my concerns. Obviously this PSU should be just fine with this load but here I am. I was hoping to get NV to cross-ship for me but that's not happening.

I'm not worried about RAM because I can independently test that and it's fine. Plus I've had the RAM/MB/CPU combo going fine for so long. The fact that I can almost immediately trigger the reboot by pushing the power headroom indicates PSU or GPU. The fact that it still happens with the card undervolted to ~280W gets me iffy on the PSU part.
Click to expand...
The reason I'm thinking PSU is normally a GPU malfunction causes a bluescreen or a driver reset.

Why not head to a local store that carries PSUs and won't charge you a restock fee for returns? Buy the PSU and see what happens when you replace what you have currently?
 
The 1000W Seasonic Prime Titanium is a pretty solid PS. Could just be an iffy power cable - have you tried switching out your PCIE power cables to a different set and plugging them into alternative location sockets within the PS to rule out a bad power cable?
 
Supercharged_Z06 said:
The 1000W Seasonic Prime Titanium is a pretty solid PS. Could just be an iffy power cable - have you tried switching out your PCIE power cables to a different set and plugging them into alternative location sockets within the PS to rule out a bad power cable?
Click to expand...
Good call. As a matter of fact it could be a cable that isn't plugged in all the way on the GPU. This did happen after changing out the GPU.

OP, pull your cables and reseat them again.
 
Supercharged_Z06 said:
The 1000W Seasonic Prime Titanium is a pretty solid PS. Could just be an iffy power cable - have you tried switching out your PCIE power cables to a different set and plugging them into alternative location sockets within the PS to rule out a bad power cable?
Click to expand...

Thanks, will swap out for the Seasonic cables that just arrived. I just added the adapter to the existing cables but good idea.. even if it's solid I could have an iffy adapter. Shit is so baffling when it's this sporadic.

Trying to avoid hitting a local MC because I'm immune compromised so just playing it extra safe.
 
kirbyrj said:
If I were to guess, I'd say that Folding@Home needs an update for the new GPU architecture. Everything else seems fine.
Click to expand...
I've been running f@h for days on end on my 3090 FTW3 Ultra, no issues. However, I will get a reboot or lockup if I overclock too high.

If the RMA/power cable swap out doesn't fix it, I'd wager that your PSU isn't handling the transient power spikes that these 30 series cards can produce (whether the PSU is aging, or just that model's caps can't hack it). I can't find the list now, but someone on the EVGA 30 series forum was compiling a list of PSUs that could handle it.
 
brentsg said:
Thanks, will swap out for the Seasonic cables that just arrived. I just added the adapter to the existing cables but good idea.. even if it's solid I could have an iffy adapter. Shit is so baffling when it's this sporadic.

Trying to avoid hitting a local MC because I'm immune compromised so just playing it extra safe.
Click to expand...

You said "seasonic cable that just arrived". Are you talking about the 12 pin seasonic made cable that they ship to you if you apply for it?
Also, you didn't tell us what driver you were using. The latest game ready driver is hot garbage. Call of Duty warzone was "DEV Error" crashing even on stock clocks. I had to use the first Studio driver to fix that. The latest hotfix driver fixed all the issues.
Can you download hwinfo, run a high load program (Heaven Benchmark with Extreme preset works excellent here) and have "Sensors Only" HWinfo window active, with the GPU voltages and voltage rails shown at full load?
 
arestavo said:
I've been running f@h for days on end on my 3090 FTW3 Ultra, no issues. However, I will get a reboot or lockup if I overclock too high.

If the RMA/power cable swap out doesn't fix it, I'd wager that your PSU isn't handling the transient power spikes that these 30 series cards can produce (whether the PSU is aging, or just that model's caps can't hack it). I can't find the list now, but someone on the EVGA 30 series forum was compiling a list of PSUs that could handle it.
Click to expand...

The PSU is just a couple years old but I'll search for that list.

Falkentyne said:
You said "seasonic cable that just arrived". Are you talking about the 12 pin seasonic made cable that they ship to you if you apply for it?
Also, you didn't tell us what driver you were using. The latest game ready driver is hot garbage. Call of Duty warzone was "DEV Error" crashing even on stock clocks. I had to use the first Studio driver to fix that. The latest hotfix driver fixed all the issues.
Can you download hwinfo, run a high load program (Heaven Benchmark with Extreme preset works excellent here) and have "Sensors Only" HWinfo window active, with the GPU voltages and voltage rails shown at full load?
Click to expand...

Yeah I am currently using some CableMod cables that I've used since I got the PSU, connected to the NV adapter. I received the 12 pin Nvidia power cables requested from Seasonic but have not yet installed. Dumb that I didn't try that.

I'm running 456.71. Is that good?

I have hwinfo installed and can run this after a conference call. I need to identify which bit of info you are calling attention to.

Generally I'll CPU and GPU fold at the same time but I've temporarily deactivated the CPU socket. I'll swap the power cable later today.
 
brentsg said:
The PSU is just a couple years old but I'll search for that list.



Yeah I am currently using some CableMod cables that I've used since I got the PSU, connected to the NV adapter. I received the 12 pin Nvidia power cables requested from Seasonic but have not yet installed. Dumb that I didn't try that.

I'm running 456.71. Is that good?

I have hwinfo installed and can run this after a conference call. I need to identify which bit of info you are calling attention to.

Generally I'll CPU and GPU fold at the same time but I've temporarily deactivated the CPU socket. I'll swap the power cable later today.
Click to expand...
I had crashes in games with the .71 driver I think they went away by going back to .55 although the longest i played in one session was about 2 hours

There is a new hotfix driver available
 
brentsg said:
The PSU is just a couple years old but I'll search for that list.



Yeah I am currently using some CableMod cables that I've used since I got the PSU, connected to the NV adapter. I received the 12 pin Nvidia power cables requested from Seasonic but have not yet installed. Dumb that I didn't try that.

I'm running 456.71. Is that good?

I have hwinfo installed and can run this after a conference call. I need to identify which bit of info you are calling attention to.

Generally I'll CPU and GPU fold at the same time but I've temporarily deactivated the CPU socket. I'll swap the power cable later today.
Click to expand...
Found it! (it popped back to the top recently): https://forums.evga.com/We-need-to-...RTX-3080-or-3090-and-is-WORKING-m3118557.aspx

It looks like the 1000W Prime Titanium is NOT on that list, unless it's actually a 1000W Prime TX? If so, the PSU should be good.
 
arestavo said:
Found it! (it popped back to the top recently): https://forums.evga.com/We-need-to-...RTX-3080-or-3090-and-is-WORKING-m3118557.aspx

It looks like the 1000W Prime Titanium is NOT on that list, unless it's actually a 1000W Prime TX? If so, the PSU should be good.
Click to expand...

Yeah I saw and posted there. I was corrected that the list is for shutdowns that require a PSU switch toggle to boot after. But I do see people posting about reboots in the thread so who knows. It seems closely related imo.

I also noticed the PSU is listed as good but .. I'll keep an open mind for the moment. My fear is RMA on the card, not get rid of my problem and add coil whine to my list.
 
I'd look at other things rather than the PSU itself...cable re-seat, maybe...but that PSU is top, top notch: https://www.jonnyguru.com/blog/2017/12/25/seasonic-prime-ultra-1000-titanium-power-supply/6/.

There's another dude on the forum here running a 750W Titanium PSU with a 3090 with no problems (i7 CPU). A good PSU is a good PSU and NVIDIA's recommends 750W or better.

EDIT: although I saw you're running essentially a miner when you're not gaming, so that changes the equation big time from "bursty" gaming activities...
 
brentsg said:
The PSU is just a couple years old but I'll search for that list.



Yeah I am currently using some CableMod cables that I've used since I got the PSU, connected to the NV adapter. I received the 12 pin Nvidia power cables requested from Seasonic but have not yet installed. Dumb that I didn't try that.

I'm running 456.71. Is that good?

I have hwinfo installed and can run this after a conference call. I need to identify which bit of info you are calling attention to.

Generally I'll CPU and GPU fold at the same time but I've temporarily deactivated the CPU socket. I'll swap the power cable later today.
Click to expand...

Did you get the hwinfo screenshots? And test with the official seasonic cable.
And 456.71 is terrible. I had nothing but games crashing on that driver. Warzone crashed at *stock* on that driver. Pure stock.
456.55 Studio driver was at least stable in Warzone, at +75 mhz core offset. I suggest 456.98 hotfix.

456.98 is great, combined with the official Seasonic 12 pin cable. I was able to use +105 mhz without any obvious problems yet (+90 is rock stable). Testing +120 now, since it's in 15mhz steps.
I have a Seasonic Prime PX-1000 PSU.
 
Falkentyne said:
Did you get the hwinfo screenshots? And test with the official seasonic cable.
And 456.71 is terrible. I had nothing but games crashing on that driver. Warzone crashed at *stock* on that driver. Pure stock.
456.55 Studio driver was at least stable in Warzone, at +75 mhz core offset. I suggest 456.98 hotfix.

456.98 is great, combined with the official Seasonic 12 pin cable. I was able to use +105 mhz without any obvious problems yet (+90 is rock stable). Testing +120 now, since it's in 15mhz steps.
I have a Seasonic Prime PX-1000 PSU.
Click to expand...

The only things I had time for tonight, update the driver to 456.98. And I've been tinkering with some benchmarks which were fine. I'll get it back to the original problematic state (aside from driver) and let it go for a bit.

I started folding when they added the COVID-19 models. It makes me feel like I'm doing something productive, however small.

Sincere thanks for all the responses everyone. Will get screenshots and cable swap tomorrow.
 
Seasonic support's recommendation was to somehow use 3 PCIE power cables with my FE.
 
brentsg said:
Seasonic support's recommendation was to somehow use 3 PCIE power cables with my FE.
Click to expand...

Waiting for your results when you have time. I'm curious if either the 12 pin cable, the hotfix driver, or both, completely fixed your card.
Also curious what your voltage rails look like. Remember to show it during full load (hwinfo64 has a "minimum" and "maximum" table under the voltages).
 
Falkentyne said:
Waiting for your results when you have time. I'm curious if either the 12 pin cable, the hotfix driver, or both, completely fixed your card.
Also curious what your voltage rails look like. Remember to show it during full load (hwinfo64 has a "minimum" and "maximum" table under the voltages).
Click to expand...
HWiNFO.JPG
HWiNFO_stock.JPG


First image is with raised power headroom, other is stock. Both are heavy load, fan is fairly aggressive.

I haven't had time and I was planning to redo this because I forgot to restart the monitoring app so my columns other than "current' are not useful. I was feeling good on the hotfix driver and then it rebooted this afternoon. Next up, power cable swap. After that will probably downclock memory.. I have to start doing mental gymnastics trying to figure out why other things away from PSU/card/cables/drivers might be the issue. I've also had others reach out with the same issue.
 
Last edited:
During gaming, my 3090 maxes out at 350W.

I see 375W there for your 3080. Still doesn't point to GPU...my system has a 9900KS and a 1000W PSU (Corsair AX1000 - by Seasonic, as well so very good). No issues. I also don't do anything other than game, though...
 
sk3tch said:
During gaming, my 3090 maxes out at 350W.

I see 375W there for your 3080. Still doesn't point to GPU...my system has a 9900KS and a 1000W PSU (Corsair AX1000 - by Seasonic, as well so very good). No issues. I also don't do anything other than game, though...
Click to expand...

The power target max for the 3080 FE is 370W.

Installed 457.09, raised the power limit and rebooted in like 5 seconds. Will swap the power cable after all my conf calls today.
 
Last edited:
Falkentyne said:
Waiting for your results when you have time. I'm curious if either the 12 pin cable, the hotfix driver, or both, completely fixed your card.
Also curious what your voltage rails look like. Remember to show it during full load (hwinfo64 has a "minimum" and "maximum" table under the voltages).
Click to expand...

Ok I finally got time to mine my old PCIE cables out and swap in the Seasonic 12 pin cable. I don't love it because it doesn't look good with my other cables but if it's functionally sound then I can worry about asthetics later. Now I just need to let it grind until I feel that it's stable.

In the meantime I pushed the power slider up and let it run Port Royal stress test for a long while. I'm attaching HWiNFO for that run, just look at the max column.

Port Royal Stress - Power 115.JPG
 
brentsg said:
Ok I finally got time to mine my old PCIE cables out and swap in the Seasonic 12 pin cable. I don't love it because it doesn't look good with my other cables but if it's functionally sound then I can worry about asthetics later. Now I just need to let it grind until I feel that it's stable.

In the meantime I pushed the power slider up and let it run Port Royal stress test for a long while. I'm attaching HWiNFO for that run, just look at the max column.

View attachment 294261
Click to expand...

The voltages look the same but it looks like the new cable helped stability?
 
Falkentyne said:
The voltages look the same but it looks like the new cable helped stability?
Click to expand...

Time will tell. I've been optimistic before!

So far so good, about 18 hours or so without a crash. Knock on wood but it might have been the stupid adapter.
 
Last edited:
Thanks for keeping us posted. So it was the NVIDIA adapter? I just ordered the Corsair cable...haven't had any problems but at the very least it will be aesthetically better...
 
DooKey said:
Reboots are usually PSU and sometimes RAM. This doesn't sound like a GPU issue to me. This current gen of GPU's from NV seem to stress the PSU more than before.
Click to expand...
I came here to say this. Changing the PSU usually fixed my random reboots. I hope this helps!
 
You must log in or register to reply here.
Back
Top