GPU-Z 0.7.5 + R9 290. Throttling the card

[e1m0]

Hi!

My rig is as follows:
i5-3570k
Sapphire R9 290 reference
8GB ram
Windows 8.1 Pro

When I use GPU-Z my GPU starts throttling frequently (see the pictures).
Resulting this (white lines of the core and memory clocks are the throttles):
http://aijaa.com/NCvR9X

The exact moment of a throttle:
http://aijaa.com/pntucK

This doesn't happen, if I don't execute GPU-Z.

Before anyone says that this is temperature related, it is not. As my temperatures were (76-80°C) far inferior to my temperature target (88°C) and with out GPU-Z I don't experience the throttling.

This can be reproduced easily by just executing GPU-Z 0.7.5. when ever I'm stressing the GPU.
The throttling occurs even when down clocked (for example to 700MHz) and drops the core down to 300MHz causing fps drops etc.

 

[brandonwh64]

Could you test MSI afterburners hardware monitor to see if it does the same thing? Also let us know what driver you are using.

 

[e1m0]

Could you test MSI afterburners hardware monitor to see if it does the same thing? Also let us know what driver you are using.

MSI Afterburner (3.0.0 beta) doesn't create any drops and works correctly.
(AB screens, when GPU-Z is not running) http://aijaa.com/3QAlFu

If I run GPU-Z, afterburner of course shows the same drops that GPU-Z.
But if GPU-Z is not executed, there are no drops.

I'm using the latest WHQL 13.12 drivers (as can be seen from the AB).

 

[brandonwh64]

Hmm this may be a question for W1Z since he is the creator of the software. Give him a PM on this issue.

 

[W1zzard]

does it help if you turn off some sensors (click the arrow after name) ? are you running any other monitoring apps at the same time as gpuz?

 

[e1m0]

I tried every sensor, and at some point (when turned off all the VDDC in/out measurements) it seemed that the clocks didn't throttle anymore, but the GPU load didn't stay at 100% and did fluctuate alot, but was better than with those sensors on, even though the performance was quite similar to the situation where the clocks also throttled.

cgminer has temperature and fan measurement and Heaven 4 has clock and temperature measurements.
Those are the two programs that I have used to test GPU-Z or after burner.
Neither temperature nor clock sensors had any effect on throttling (if I turned them on/off) during the test I just made.

 

[W1zzard]

Is your CPU at 100% load while testing this? CPU, not GPU

 

[e1m0]

My CPU is not even close to 100% when testing this.
For example with cgminer it stays under 10% all the time, while GPU is 100% if I dont open GPU-Z (as soon as I open it the throttling begins).

I can't run further testes now because it appears that my connection to my computer that I was monitoring (from a distance) has failed after a crash when I executed GPU-Z as I tried to take a screen shot with CPU utilisation, AB and GPU-Z simultaneously....

 

[W1zzard]

the throttling happens only when using remote desktop?

 

[e1m0]

According to the data from GPU-Z I did experience throttling before remote desktop, but I didn't have time to verify the source and I didn't use GPU-Z that actively back then.
It was only today when I found out that it was GPU-Z that causes it (I started active monitoring of the temperatures and clocks yesterday, first I thought it was the GPU or PSU, but they are ruled now out as I could replicate the problem) and I'm sure of it as I could replicate the behavior just by opening GPU-Z while I had load on the GPU
I wont be able to do any further testes, as I can't hard reset my computer due to the distance. The computer apparently crashed and shut itself down without reboot. I'll be back to my computer in 2 weeks. Before that I don't believe that I can do anything about it.