GPU Behaviour

[surfistche]

Hello tech lovers,

 

I am facing a very strange situation and I was wondering if I can get an advice on what the cause is or how to proceed.

 

Recently I bought a Sapphire Nitro+ 7900XTX. After the installation I noticed slight freezes, and decided to run some testing. I discovered that the GPU is producing VRAM errors. I am attaching the image for reference:

 

 

Spoiler

 

 

Now after contacting the retailer, I send my card back, and after a month long wait, I actually received an RMA. So far so good.

However, straight out of the box, I decided to do some slightly abusive testing on my system, to see if the new card will produce the same errors. I went for a few tests, which loaded my CPU, VRAM and GPU Utilization to 100% simultaneously. While no errors are to be seen for the moment, I noticed that my 420W rated GPU, while keeping it around that value for the most of the 30min testing, there was a time frame, where it actually hit almost double at nearly 700W. Attaching the second image for reference:

 

Spoiler

 

 

No overclocking of any sort has been done on the system. All the settings are pretty much fabric, with the exception of the use of X.M.P.

 

As shown on the above image, everything seems to be in order, except the time frame, where I had the 600+ power consumption. I would like to ask if that is normal, how can I further diagnose any possible power issues, and if I should opt for another RMA, or simply look into another manufacturer?

Thanksin advance,

Cheers!

[RONOTHAN##]

I will want to point out HWMonitor has a lot of issues with newer hardware and isn't really to be trusted. You want to use HWInfo for monitoring instead, it's a lot more reliable. That ~700W reading was almost certainly a reporting error in software, not the card actually hitting 700W. 

[Sjaakie]

If it works fine I would not worry about it. Modern high end gpu's can have big power spikes above their wattage rating. ATX spec accounts for that.

[surfistche]

8 hours ago, RONOTHAN## said:

I will want to point out HWMonitor has a lot of issues with newer hardware and isn't really to be trusted. You want to use HWInfo for monitoring instead, it's a lot more reliable. That ~700W reading was almost certainly a reporting error in software, not the card actually hitting 700W. 

HW Monitor was on par with OCCT though. I was following the graph there as well, and they were both showing the same usage.