Strange uncore readings

[srg]

Hi,

 

I have got a few questions regarding your application.

I was running a stress test with AIDA64 but then saw some strange values.

The uncore wattage was, at one point according to AIDA, at -10000W, which is quite unrealistic.

Now I haven't had any problems, crashes or whatsoever, from which I conclude that my system is considered pretty stable (already ran Prime in-place for 24h without errors before)

but I'd still like to know whether it's a bug, a hardware flaw or just a reading error.

I've already had problems with the clock and the usage dropping to zero and going up again, but solved it by googling

and closing Steam and Chrome plus not running other monitoring applications in the background. 

However, when this happened, I also wasn't running anything else so I'm kind of confused.

 

Also, I've already read in other threads, that if the application detects an instability it is supposed to turn read. Yet I didn't quite understand what exactly should turn red and whether there are any other signs (like the status saying the test stopped or something like that) of light instability, not severe enough to crash the program/system

 

Thank you in advance!

 

Regards

 

 

PS: Is it normal that, when also stressing the iGPU, the image is hanging for up to 3-4 seconds?

[Fiery]

1) The negative uncore reading is just a temporary reading error, you can simply ignore that value.

2) The System Stability Test window turns quite red when it detects an error. Here's a sample screen shot:

3) To stress the GPU, AIDA64 uses OpenCL code. OpenCL however is a layer that is quite closely tied to the video display driver right now, and so putting a heavy OpenCL stress on the GPU causes the video driver to slow down. That issue will only be solved when GPU manufacturers will finally decouple OpenCL from the video driver.

Regards,

Fiery

[grs]

Thank you very much Fiery!

(it's the OP, I simply forgot the credentials of my "old" account)

 

Would there be any way to prevent those reading errors from happening in the future? It still messes up the average value sadly.

 

By the way, by "video display driver" in point 3), you are refering to both, dedicated GFX drivers AND integrated GPU drivers [intel HD], if I got that correctly?

 

Again thank you very much for your detailed and quick response!

 

Regards

[Fiery]

Would there be any way to prevent those reading errors from happening in the future? It still messes up the average value sadly.

There's already a mechanism in place to filter out such values. Maybe you're not using the latest AIDA64 version of v5.00?

 

By the way, by "video display driver" in point 3), you are refering to both, dedicated GFX drivers AND integrated GPU drivers [intel HD], if I got that correctly?

If you've got multiple GPUs that support OpenCL, then AIDA64 will use all of them to stress the system.

[grs]

There's already a mechanism in place to filter out such values. Maybe you're not using the latest AIDA64 version of v5.00?

 

If you've got multiple GPUs that support OpenCL, then AIDA64 will use all of them to stress the system.

 

I'm currently on version 5.00.33, does that mean it actually should not happen to me, and more important, can I still ignore it?

 

Regards

[Fiery]

I'm currently on version 5.00.33, does that mean it actually should not happen to me, and more important, can I still ignore it?

You can ignore the negative values. We however will need to implement stronger workarounds to filter out negative/invalid readings

[grs]

You can ignore the negative values. We however will need to implement stronger workarounds to filter out negative/invalid readings

Thanks again, this is by far the best customer service I have ever encountered!