AMD R9 270x temps reported incorrectly/glitchy

[tgp1994]

Hi everyone,

I've been seeing an issue lately where my GPU temperatures are reported strangely.

Firstly, it seems like HWiNFO reports them at least 15 degrees celsius below where they're supposed to be. The screenshot below shows a concurrence between Valley Benchmark and the iTurbo software provided with my graphics card.

Secondly, whenever I start up a 3D application, HWiNFO starts dropping off the temperature readings, almost like a wire suddenly becomes loose. Please also see the screenshot - the almost "digital signal" portion of the temp graph is when I started running the benchmark software, although again this applies to any 3D application.

To be specific, my card is an HIS IceQ X² Boost H270XQM2G2M Radeon R9 270X. Playing devil's advocate here, I did have to replace the card nearly right when I built my computer due to overheating issues. I got a different one back awhile later and things seemed to be ok, although I've been seeing crashes in a few odd games. I've been trying to push my card to crash in the Unigen benchmarks, but can't get my card or drivers to crash with it pushing 60 Celsius.

Are both of these the result of HWiNFO?

[attachment=1305]

 

[Martin]

Please try to disable monitoring of the GPU VRM sensor (right-click over the sensor heading and "Disable Monitoring") to check whether that might be causing the issue. Please let me know the result.
Another possibility is that HWiNFO is conflicting with some other application measuring GPU parameters (maybe the iTurbo) when running concurrently.

 

[tgp1994]

Please try to disable monitoring of the GPU VRM sensor (right-click over the sensor heading and "Disable Monitoring") to check whether that might be causing the issue. Please let me know the result.
Another possibility is that HWiNFO is conflicting with some other application measuring GPU parameters (maybe the iTurbo) when running concurrently.

No dice, I'm afraid... I disabled the VRM sensor in HWiNFO like you asked and confirmed that the icon showed a red X, then I launched the benchmark software again only to see the sporadic temperature graph show up under GPU Temp.

I don't believe I'm running any other monitoring applications... I had only opened iTurbo to confirm that the reported temperature was incorrect, otherwise I don't run any other monitoring or overclocking applications.

 

[Martin]

Is the debug file you attached capturing the issue with sporadic GPU temperature? If not, please attach such a debug file that was created while the issue occured

 

[tgp1994]

Is the debug file you attached capturing the issue with sporadic GPU temperature? If not, please attach such a debug file that was created while the issue occured

That's correct - I tried running the capture while reproducing the problem, so hopefully that file contains something useful for you.

 

[Martin]

I'm sorry, but I can't see any problems with GPU temperature in the attached debug file.
Please create a new debug file while the problem happens. Also, please enable sensor logging during that time and attach the created sensor log.

 

[tgp1994]

I'm sorry, but I can't see any problems with GPU temperature in the attached debug file.
Please create a new debug file while the problem happens. Also, please enable sensor logging during that time and attach the created sensor log.

This was interesting... After letting my computer idle for awhile, I came back to run this test. I opened HWiNFO, enabled debug logging, opened the GPU graph, then closed HWiNFO to restart it. I opened it via shortcut, clicked yes on the UAC prompt, then after a few seconds, both monitors went completely black, mouse and keyboard lights went dark, and the CPU fan spun up to 100%. As far as I could tell, something in the hardware had crashed. When I attempted to reboot my computer via the hardware reboot button, the motherboard speaker would emit a keyboard error beep code. Strange... I powered off my computer and restarted it, no problems after that. I've seen my computer crash quite a bit (unfortunately), but never like that! I've attached the debug log which I guess had a small chance to gather some data before the crash, if you have any idea what happened there.

It seems like HWiNFO didn't want to automatically start on login this time, so I'm opening it via shortcut again to test. I could hear my hard drive being accessed quite a lot after I accepted the UAC prompt again, but after waiting about 15 seconds, nothing happened. No debug log, nothing. I tried running it again, got the UAC prompt, and just like the first time, both screens went dark. Rebooting my computer via the button results in a keyboard error beep code. I hard powered it down again to and restarted to eventually write this, but there was no debug log this time.

What do you think about all of this Martin? I'm assuming I might have better luck if I delete the config file, but I find it so strange that a program can seeming cause my BIOS to become confused, requiring me to power down the system to refresh it. I'm attaching the debug log from the first crash, if you think that might help.

Thanks so far!

 

[Martin]

It seems the issue is the GPU VRM sensor - when accessed, it can sometimes cause such problems. It should not happen, but I've seen this a few times. To avoid this, please disable GPU I2C Support in HWiNFO before starting the scan.
Then please create the new debug and sensor log files if the problem with sporadic GPU temperature persists.

 

[tgp1994]

It seems the issue is the GPU VRM sensor - when accessed, it can sometimes cause such problems. It should not happen, but I've seen this a few times. To avoid this, please disable GPU I2C Support in HWiNFO before starting the scan.
Then please create the new debug and sensor log files if the problem with sporadic GPU temperature persists.

Interestingly, I had left my computer logged on and left for a bit. When I came back, HWiNFO decided to start. Huh!

I disabled GPU I2C Support and enabled debugging again like you asked, and when restarted, HWiNFO was up in a matter of seconds. Enabling sensor logging and watching the GPU Temp graph, miracuously, it looks like it's working fine now! Temp reports concur with other monitoring software, and it isn't jumping like last time. I've attached the related debugging files just in case, but I think you've solved it Martin.

Thank you very much!

 

[Martin]

Well, some GPU VRM sensors can cause headaches
I'm glad that it's working now properly.