GTX 980 stable in furmark and gpu_burn but 'falls off the bus' in games

Graphics / Linux Linux
1

GTX 980 stable in furmark and gpu_burn but ‘falls off the bus’ in games within 10 minutes (usually much sooner) - specifically I’ve tested with GTA V and XCOM 2. I would appreciate any strategies to try if anyone has ideas.

Here’s what I’ve tried:
Upgraded to driver 510.47.03.
Tried various kernel parameters such as disable drm, set CPU state, disable modeset.
Underclocked using nvidia-smi to match nvidia reference power limit (165w) and nvidia reference clocks (3505,1215).

keywords
fallen off the bus falling falls no signal

2

I don’t know what it means ‘falls off the bus’

could you explain it in another way? i also have a 980 asus strix, my gpu has a much lower performance than windows.

for example overwatch on windows 180 fps stable but on linux it stays below 144fps
in kovaaks i have 450 fps in windows and in linux is below 320

my temperature is fine it doesn’t reach 55º C and the clocks are always on high performance

3

‘Falls off the bus’ is a reference to NVIDIA’s error message.

Feb 05 21:30:39 myhostname kernel: NVRM: GPU at PCI:0000:01:00: GPU-98d8b340-b34f-3042-f74f-49d5ebd02011
Feb 05 21:30:39 myhostname kernel: NVRM: Xid (PCI:0000:01:00): 79, pid=251, GPU has fallen off the bus.
Feb 05 21:30:39 myhostname kernel: NVRM: GPU 0000:01:00.0: GPU has fallen off the bus.
Feb 05 21:31:08 myhostname audit[239218]: ANOM_ABEND auid=1000 uid=1000 gid=1000 ses=2 pid=239218 comm="GpuWatchdog" exe="/home/myusername/.steam/debian-installation/ubuntu12_64/steamwebhelper" sig=11 res=1
Feb 05 21:31:08 myhostname kernel: show_signal_msg: 1453 callbacks suppressed
Feb 05 21:31:08 myhostname kernel: GpuWatchdog[239229]: segfault at 0 ip 00007fd2963ccb5f sp 00007fd28cadc400 error 6 in libcef.so[7fd292192000+6f56000]
Feb 05 21:31:08 myhostname kernel: Code: 89 de e8 54 36 8e fe 80 7d cf 00 79 09 48 8b 7d b8 e8 65 54 d1 02 41 8b 84 24 e0 00 00 00 89 45 b8 48 8d 7d b8 e8 61 65 dc fb <c7> 04 25 00 00 00 00 37 13 00 00 48 83 c4 38 5b 41 5c 41 5d 41 5e
4

Sounds like a problem with the psu, running furmark+gpu_burn will only stress the gpu. While gaming, the cpu will also draw more power.

5

It’s stable with all cores at 100% in mprime while running furmark extreme burn mode. I just re-pasted it and re-seated it, but that didn’t help. Crashed within 2 min of starting up XCOM 2. Same error - ‘has fallen off the bus’.

6

Tested with a 1000W PSU today. It didn’t help.

7

Did you already try to reseat the gpu in its pcie slot? Please check if you can lower pcie speeds to gen2 in bios to check for bus issues.

8

Yes I reseated the card after I removed it to redo the thermal paste. I just tried dropping my PCI-e speed to Gen 2 and then to Gen 1. I also disabled all my CPU power saving options and there was an option for powering down GPU cores at idle or something I disabled too. I think it was only related to iGPU, but just in case I disabled it.

None of that helped.