intel / intel/xpumanager

The data coming out of xpu-smi are incoherent when the load is high

Open
#147 16 comments 0 reactions 0 assignees View on GitHub
Dominant language
C++
Stars
194
Forks
34
PR merge metrics
No merged PRs in 30d

Description

xpu-smi --query-gpu=index,utilization.gpu,memory.used,memory.free,power.draw

This is for the 3rd Intel B70 cards
index, utilization.gpu (%), memory.used (MiB), memory.free (MiB), power.draw (W)
2, 0.00, 29028.4921875, 3627.5078125, 54.89
2, 1.01, 29101.18359375, N/A, 56.84, N/A, N/A
2, 11.80, 29099.05078125, 3556.94921875, 99.01
2, 20.59, 29101.140625, N/A, 165.63, N/A, N/A
2, 20.57, 29099.046875, 3556.953125, 172.79
2, 20.65, 29101.14453125, N/A, 163.05, N/A, N/A
2, 21.62, 29099.046875, 3556.953125, 139.88
2, 20.78, 29101.14453125, N/A, 162.70, N/A, N/A
2, 20.45, 29099.046875, 3556.953125, 176.42
2, 21.06, 29101.14453125, N/A, 176.58, N/A, N/A
2, 20.95, 29099.046875, 3556.953125, 177.38
2, 20.85, 29101.15234375, N/A, 188.36, N/A, N/A
2, 20.29, 29099.0859375, 3556.9140625, 176.76
2, 21.37, 29101.15234375, N/A, 158.71, N/A, N/A
2, 21.52, 29101.18359375, 3554.81640625, 157.06
2, 22.22, 29099.046875, N/A, 94.33, N/A, N/A
2, 11.11, 44.3671875, 32611.6328125, 97.91
2, 11.11, 42.23828125, N/A, 83.29, N/A, N/A
2, 11.11, 44.36328125, 32611.63671875, 57.20
2, 11.11, 42.234375, N/A, 56.28, N/A, N/A
2, 11.11, 44.36328125, 32611.63671875, 55.98
2, 11.11, 42.234375, N/A, 55.57, N/A, N/A
2, 11.11, 42.234375, N/A, 56.63, N/A, N/A
2, 11.11, 42.234375, N/A, 56.23, N/A, N/A
2, 11.11, 42.23828125, N/A, 56.38, N/A, N/A
2, 11.11, 233.68359375, N/A, 55.88, N/A, N/A
2, 11.11, 233.68359375, N/A, 55.96, N/A, N/A
2, 11.11, 233.6875, N/A, 56.44, N/A, N/A
2, 11.11, 233.67578125, N/A, 55.90, N/A, N/A
2, 11.11, 233.6796875, N/A, 56.02, N/A, N/A
2, 11.11, 233.67578125, N/A, 56.50, N/A, N/A
2, 11.35, 443.3359375, N/A, 55.86, N/A, N/A
2, 23104788116198.93, 954.37109375, N/A, 56.65, N/A, N/A
2, 5.01, 2873.48046875, N/A, 99.49, N/A, N/A
2, 1.19, 2939.25, N/A, 59.02, N/A, N/A
2, 0.00, 3086.87109375, N/A, 98.56, N/A, N/A
2, 4.41, 29070.33984375, N/A, 98.96, N/A, N/A
2, 0.00, 29070.33984375, N/A, 67.38, N/A, N/A
2, 0.00, 29070.33984375, N/A, 58.40, N/A, N/A
2, 0.00, 29070.34375, N/A, 55.79, N/A, N/A
2, 0.00, 29070.34375, N/A, 56.03, N/A, N/A
2, 0.00, 28942.33984375, N/A, 56.65, N/A, N/A
2, 0.00, 28942.33984375, N/A, 55.89, N/A, N/A
2, 0.00, 28942.33984375, N/A, 56.12, N/A, N/A
2, 0.00, 28848.33984375, N/A, 55.07, N/A, N/A
2, 0.00, 28848.34765625, N/A, 55.79, N/A, N/A
2, 0.00, 28848.33984375, N/A, 56.17, N/A, N/A

Contributor guide

Open the contributing guide

Research direction

Start by reproducing the reported xpu-smi --query-gpu=index,utilization.gpu,memory.used,memory.free,power.draw command on the 3rd Intel B70 card under high load. Compare repeated output rows and determine the expected behavior for missing values and implausible utilization or memory readings; done means the query produces coherent, consistently formatted measurements.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
cli
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Active
Clarity
Needs clarification
Newbie friendliness
38/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.