0% GPU Utilization

Okay - I was a little premature. Now getting a reading (which is progress from just seeing 0%), and under load it nearly matches GPU-Z

On idle though, it bounces around all over the place - I've seen max readings of up to 38.4%, when doing nothing and its regularly reading 10-20%

Attached is a tiny screenshot of the difference, sorry I haven't formatted these, but both graphs are GPU utilization/load (y axis run 0-100%), x axis is time over 30s

Top graph is from GPU-Z GPU Load sensor, bottom is from HWiNFO GPU Utilization sensor over same time period (when idle)
 

Attachments

  • Capture.PNG
    Capture.PNG
    3 KB · Views: 4
I see... will check that.
But again I need a Debug File from that situation so I can analyze it. Please let it run during that idle period and attach. Would be also useful to know what approximate load values have been displayed by GPU-Z during that period.
 
Argh - back to BSOD's, even after deleting debug file.

After opening sensors - the last thing I'm seeing on the screen is it parsing through the sensors and getting to "GPU ATIx 26" (or something very similar)

I'll continue to try to resolve this and get the debug files posted... but to answer your question GPU-Z max'd at about 3 or 4% during that previous idle period
 
Please try to disable the GPU I2C Support if that maybe solves the BSOD problem temporary - you'll then loose the GPU VR sensor, but just to be able to gather the Debug Files...
 
Thanks for the debug file. I suppose when you sampled this file HWiNFO reported correctly 0% load. Is that so?
So could you please let it run for a few minutes (about 3) in idle so that it captures the peaks which you think are not correct ? Then please attach the new debug file.
 
Try these - ran for 4 mins or so... charts so usage over the last 3 mins. GPU-Z is top one HWiNFO is bottom one
 

Attachments

  • HWiNFO64.DBG
    882.5 KB · Views: 1
  • Capture.PNG
    Capture.PNG
    2.6 KB · Views: 6
Thanks, I'll analyze it. One more question - could you please attach a larger picture of the graph? The one attached is too small to see it precisely.
Or maybe if you have the raw sampled data (in CSV or XLS) which you used to create the graphs, that might be even better for comparison.
 
When checking the HWiNFO CSV log, I don't see any problematic spikes like on the graph - it's flat 0.
So on which values is that graph based ?
 
The little chart was from LCDHost - reading HWiNFO/Usages/ATI GPU (...)/GPU Utilization


It isn't on 0% due to me disabling the GPU I2C option is it?


Try these - a couple of 50% spikes in there
 

Attachments

  • HWiNFO log.CSV
    38.3 KB · Views: 1
  • GPU-Z Sensor Log.txt
    124.8 KB · Views: 1
Please check the newly released v4.11-1844 Beta. I have added few more improvements there.
 
Back
Top