0
\$\begingroup\$

For performance tracing of intermittent degredation in performance I wanted to use the Performance Counters available in Windows 10 1809 under GPU Engine -> Utilization percentage. This particular laptop has NVIDIA GeForce GTX 1050 with 4 GB VRAM + 16GB shared memory.

The instance names look like this:

pid_1388_luid_0x00000000_0x00011372_phys_0_eng_8_engtype_Compute_1 pid_912_luid_0x00000000_0x00010F5C_phys_0_eng_2_engtype_Copy

While I know pid_xxxx is the process identifier, what I am trying to work out the meanings of remaning components.

1) luid_xxxx_xxx

In my case all instances start with luid_0x00000000 and have 3 different endings:

  • 0x00010F5C
  • 0x0001134D
  • 0x00011372

2) Next is phys_0, I assume this is first physical GPU, but I want to be able to programatically translate it to GPU adapater name.

3) Eng_x, on my machine this ranges from 0-12

4) Eng_type have the following. While mostly self explanatory, is the difference between Compute_0 and Computer_1 different cores on the GPU

  • 3D
  • Compute_0
  • Compute_1
  • Copy
  • GDI
  • Render
  • Graphics
  • LegacyOverlay
  • Security
  • VideoDecode
  • VideoEncode
  • VideoProcessing
  • VR
\$\endgroup\$
1
\$\begingroup\$

1) luid_xxxx_xxx

These appear to be locally-unique identifiers. They're 64-bit values that are unique within the scope of the machine they're generated one.

Next is phys_0, I assume this is first physical GPU but I want to be able to programatically translate it to GPU adapater name.

That seems correct to me. You should be able to use your 3D API of choice to enumerate the physical adapters available and correlate those with these counters in order to get a name for them, if you like.

Eng_x

This looks like an "engine" index. See below.

4) Eng_type have the following. While mostly self explanatory, is the difference between Compute_0 and Computer_1 different cores on the GPU

This identifies the type of GPU engine; a GPU engine is distinct unit of silicon that runs in parallel with others on the CPU. There are two compute engines on that GPU. GPU engines are made up of cores, they are not the same as cores.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.