NVenc usage leads to driver-reset and reboots

[donald24]

Hello there,

my current rig is the following,

i7-8700k
2x16GB 3200Mhz
Nvidia 1080TI FE -stock clocks-
Windows 10 x64 1809 -patched-
latest Nvidia driver 436.02
NOD32 AV


Been trying to use NVenc with 1080P-fast-profile, but when running like five minutes max, first the picture of my two monitors freezes, one goes to "no signal" and the computer resets and coldboots.

Activity-log shows nothing interesting:
https://pastebin.com/hXCHaSwz

The system is behaving perfect when encoding without hw-accel. Temperatures, powersupply are well controlled and sufficient. I can 3D-game for hours, and encode movies as long as I don't activate the NVenc-mode.

I've tried hw-accel on [Removed Dubious Product Reference]-HD-Video Converter and I can use it without crashes, so I am out of ideas.

I switched to my second Win10-install and it behaves the same. Any ideas?

[tlindgren]

latest Nvidia driver 436.02
Been trying to use NVenc with 1080P-fast-profile, but when running like five minutes max, first the picture of my two monitors freezes, one goes to "no signal" and the computer resets and coldboots.

436.02 is known to be rather buggy and was quickly replaced 436.15 on the 27th. That version fixes some of the bugs but is still nowhere near as stable as say 431.60 so you may want to test using 431.60 even if I've not heard anything specific about NVenc related bugs on either of the 436.x releases.

Since the machine reboot it's very likely to either a hardware or driver issue (99+% chance basically).

NVEnc (and NVDec) is dedicate hardware not used for normal functions so the gaming test doesn't test NVEnc unless you're live streaming with hw encoder, this might be worth trying to see if that too causes issues? Not sure if the other video encoder you mention uses NVEnc but even if it does it doesn't rule out hardware or driver issues, it could require specific circumstances to trigger.

If switching to a more well-used driver version doesn't help the obvious way forward is to start swapping components (starting with either PSU or GFX card). If you can't do that it's sounds like it's going to be a hard nut to diagnose, perhaps try running both FurMark (start first) and GIMPS Prime95 Stress Test (the "max heat" option) to see if you can provoke the crash that way and possibly get some insight.

[Devon Malcolm]

if you can play 3D games and ncode movies for hours i would like you to change the "output mode" to "advanced" and try again.