Weird measurement values - who's at fault?

[Pattu]

Graphics card is a an RX Vega 56. Default settings, except power limit is at -25% (whatever that translates to in real values according to AMD)

The card crashes on high load (black screen, graphics card seems to turn off, PC is running, but does not send any video signals anymore).
I let GPU-Z run while running a high demand game leading up to a crash and examined the data.

In particular:

* in the last 3 lines CPU clock goes unusually high (>1500MHz, which according to the specs is above the limit of 1471MHZ)
* GPU load goes from around 70 to 84%
* GPU Chip Power Draw jumps from around 80W (although it has some 100W spikes) to 145W (might be max?)
* GPU hot spot jumps from 70°C to 511°C (yes, five hundred eleven degrees celsius, not a typo)
* SOC VRM Temp shows no value (just -)
* VDDC goes negative, -0.0437V (???)

Now I'm not sure how to interpret this data.
Certainly the temperature and voltage do not represent the real world conditions.

Is there a chance that GPU-Z is at fault here? Maybe the logged values are generated while the card is crashing resulting in this?
Or can I be sure it receives this data from the sensors before the crash? I read there are some sensor bugs in the AMD drivers - may that be the cause here?

 

[qubit]

If it's not something obvious like a driver or some software corruption, then you might have a faulty card.

What's a "high load"? Are you overclocking it?

 

[Pattu]

> If it's not something obvious like a driver or some software corruption, then you might have a faulty card.

Well, yes. At this point I'm just trying to figure out whether it's the drivers, the card, or as some others with the same issue have suggested the PSU.

> What's a "high load"? Are you overclocking it?

No, I mean demanding games (in particular Monster Hunter World on highest settings).

 

[Naki]

Fill in all system specs, please!
Not clear - what CPU is used, what mobo, which kind of RAM and how much of it, what storage devices?
How many other internal or external devices?

If you can, try the same GPU in another, known to work PC with a better PSU.
Also, try another, known to work GPU on this PC.

 

[Tatty_Two]

As a test, try increasing power from your current -25% to say -15% and test again, peak gaming power draw according to this sites original review of the card when released showed around 235w so for some reason it may be struggling with the power limit now.

 

[Steevo]

The CPU clock variation is normal.

It’s either too low power for the GPU frequency requested due to your low power setting, a faulty card, or a failing/underpowered power supply.

The incorrect values are probably corrupted values due to the system crash.

Try removing any overclock, return pier setting to normal and see if the issue persists.