4

I have a Dell Inspiron 3542 that sometimes displays short vertical lines when Windows 7 boots instead of the animated but logo, but they disappear immediately at the next loading step.

I wanted to check the GeForce 820M dedicated 2GB of RAM for errors, but I got different results for every single software and methods I tried.

OCCT

Newer versions of OCCT don't seem to provide graphics processing unit memory tests, but I ran the 3D tests anyway and I didn't get any errors.

Uningine Heaven Benchmark

I just ran that as a stress test, and I got no crashes or artifacts.

MSI Kombustor

MSI Kombustor ran fine, but it only used around 160MB of video memory, which I don't think fits as a RAM stress test.

Video Memory Stress Test

This program just didn't want to run the tests, it just gave me an error 300 when trying to run them.

I tried to set the 820M as the main graphics processor in the NVidia configuration panel, but the software just crashed and refused to run.

Video Memory Stress Test Bootable Image

I made a bootable media of Video Memory Stress Test to ensure that the issue wasn't produced by the operating system.

The test starts and reports every single try as failed, which seems suspisious, as I wouldn't expect the graphics card to work at all if all of its RAM was bad. It also didn't detect the GPU name and vendor properly.

Moreover, I tried it on my main PC with an RX 480 8GB, it reported the vendor correctly and didn't detect any errors, but it didn't seem to detect video memory beyond 2GB.

GpuMemTest

This software seems extremly simple. The test returned no errors. When monitoring the memory usage with GPU-Z it used most of the 2Gb, so it seems like it does its job.

I couldn't find a lot of documentation about this software.

MemTestG80

This software seemed promising at first, but it behaved weirdly.

I could easily test the video memory as long as I didn't run tests above 1.2GB. Any test on more RAM would immediatly crash the NVdia driver, and MemTestG80 would report that every single test failed.

MemTestCL

As its pretty much the same as MemTestG80, I didn't expect different results, but I was wrong.

MemTestCL couldn't run tests on the whole 2GB of RAM, but tests ran successfully on 1.9GB, without any errors. Monitoring with GPU-Z ensured me that the RAM usage was matching the amount MemTestCL was testing.

I have to stress test graphics cards quite often, and there are a lot of straightforward tools to do so. But there doesn't seem to be a definitive software to diagnose video memory.

In your experience, what is the most reliable way to test a graphics processing unit dedicated memory (on either Windows or Linux)?

6
  • Are you sure this notebook doesn’t have switchable graphics (Intel + NVIDIA)?
    – Daniel B
    Aug 16, 2018 at 7:56
  • @DanielB It also has Intel HD Graphics, but it doesn't have any dedicated memory. I ran MemTest86+ and it found no faults, the Intel Processor Diagnostic Tool didn't detect anything either. Aug 16, 2018 at 8:58
  • If it has integrated graphics, this is what’s used during startup. Depending on how exactly the artifacts look like it’s more likely the fault is either inside the display or in the display connection.
    – Daniel B
    Aug 16, 2018 at 9:45
  • But if the iGPU works fine in the OS then we shouldn't suspect it or the display. At boot time the way BIOS/UEFI releases control of the graphics to the OS (and its graphics driver) may result in all sorts of artifacts. I suspect this is the result of a legacy OS (Win7) booting in a modern UEFI machine. The artifacts occurs at that moment when "UEFI video" falls back to "legacy video" for compatibility.
    – user931000
    Aug 16, 2018 at 14:59
  • @GabrielaGarcia I also suspect that, as the original installation of Windows 7, installed with the BIOS in Legacy mode, worked fine. But I need to be completely sure that this isn't an hardware fault, as I'll be giving this computer away. Also, I forgot to mention that artifacts occur most of the time, but sometimes the Windows 7 boot logo displays fine. The semi-random nature of these artifacts don't ensure me that they are produced by the UEFI handing the video handling to the OS. Even if these artifacts are not an hardware fault, I'd like to find a way test dedicated video memory reliably. Aug 16, 2018 at 15:25

0

You must log in to answer this question.

Browse other questions tagged .