Lenovo G710 Very weird GPU issue

[KvnTM]

Hello Friends,

I'm puzzled about this laptop.. It has a dedicated Nvidia 820m GPU and BSoD's
randomly. The GPU works fine with 3D Mark and FurMark. Memtest reported no
issues and the CPU is fine. So far so good.

Now the weird part begins:
Furmark reports the GPU as a GT710m. GPU-Z as 820m and the Nvidia Control
Panel sometimes as 820m and sometimes as 720m. I can't install the official
drivers from nvidia for either the 720m or 820m. It always says that the installer
can't find a compatible GPU. But Windows installs the 820m drivers by itself.

When running 3D Mark or Furmark no BSoD's appear. But it instantly crashes
when closing the internal "benchmark" in GPU-Z with
VIDEO_SCHEDULER_INTERAL_ERROR. This typically tells me that the GPU is
dying. But no other Benchmark causes issues, even when running it for hours.

One thing I noticed: When stressing the CPU it thermal throttles to around
0.8GHz after some time. The frequency does not normalize after that, even when
the CPU sits at 50C. It just stays at 0.8GHz. Restarting the Laptop has no effect
either. Only completely shutting it down once restores the normal clocks.

I'm confused to hell... Did someone experience this behavior?
Screenshots were taken via TeamViewer. I don't have physical access to it at the
moment but I can do remote tests.

Sidenote: I analyzed the minidumps of all crashes and it was never caused by the
Nvidia driver. Always something random kernel or windows related. Only when
forcing it to crash via the GPU-Z benchmark it fails with the usual nvxxxxx.sys file.

[piernov]

Entry-level GPU are often the exact same GPU rebranded over several generations so it's not surprising to see different names. Here GPU is GF117 and apparently it can be 620M/625M/630M/710M/720M/810M/820M. The official NVidia driver package for any of those should still install properly.

Maybe you have a power related issue rather than GPU issue itself.

[KvnTM]

Originally Posted by piernov

Entry-level GPU are often the exact same GPU rebranded over several generations so it's not surprising to see different names. Here GPU is GF117 and apparently it can be 620M/625M/630M/710M/720M/810M/820M. The official NVidia driver package for any of those should still install properly.

Maybe you have a power related issue rather than GPU issue itself.

Thats pretty stupid but hey Nvidia isn't known for logical product names, right?

Anyways.. I'd also imagine that it is a power issue. But why doesn't it crash when everything is under stress? Temps are also irrelevant. Sometimes it crashes instantly after powering on and sometimes after hours.

Kinda reminds me of the Late 2013 15" Macbook Pro's without a dGPU...

[SMDFlea]

What bios version does it have. 8ECN95WW has a Vbios update

https://download.lenovo.com/consumer...s/8ecn95ww.txt

Update N15V-GM VBIOS to N15V_GM_75_17_84_00_FF.

[caspian]

What times do BSODs appear during ordinary usage of the laptop?
What does happen if you restart the laptop just after a BSOD?

[KvnTM]

Originally Posted by SMDFlea

What bios version does it have. 8ECN95WW has a Vbios update

https://download.lenovo.com/consumer...s/8ecn95ww.txt

Update N15V-GM VBIOS to N15V_GM_75_17_84_00_FF.

I‘d assume my mom (her laptop) didnt update the BIOS so i assume it’s the first released one. Will do that! I havent thought about that..

Originally Posted by caspian

What times do BSODs appear during ordinary usage of the laptop?
What does happen if you restart the laptop just after a BSOD?

Pretty much randomly but seems to be only happening when the gpu drivers are loaded. Didnt occur while running memtest overnight. She mostly browses the web on Chrome or does Text editing stuff.

The GPU-Z benchmark crashes it instantly when closing it. It does not crash while stressing the GPU or CPU.

It works normal after restarting. Minidumps are pretty much useless since it crashes at random things which are not related to the GPU. Only the GPU-Z thing crashes the nvidia drivers.

[KvnTM]

Updating the BIOS didn't fix the issue.
I still can't install the official drivers for it. I can only install a very old driver package from 2015.

I disabled hardware acceleration in Chrome a few days ago and it seems to be more stable now which makes sense because it doesn't use the GPU anymore.

[emestee]

I googled for curiousity and it appears the discrepancy between 820m and 710m being reported is normal - you can find any number of screenshots of furmark showing the same thing for other people.

[caspian]

If you test the GPU using a live bootable CD, you may get better error messages. this is because of the fact that no OS will be involved between the test software and the GPU.

[KvnTM]

Originally Posted by caspian

If you test the GPU using a live bootable CD, you may get better error messages. this is because of the fact that no OS will be involved between the test software and the GPU.

Any suggestions on a live CD with nvidia GPU support and benchmarks? Usually linux distros dont come with nvidia or any discrete GPU drivers iirc.