New PC gpu and cpu underperforming

Ash_

Master Poster
I was thinking about going back a previous version on Nvidia drivers but I am not sure if that is going to work
It is normally conflicting drivers. You definitely completely removed them and clean installed them right?

I got this issue before, when i downloaded AMD drivers to stress test a 5700XT i was repairing and when i put my Nvidia back in, this was a similar issue, just not as bad.
 

Kickim

Bronze Level Poster
It is normally conflicting drivers. You definitely completely removed them and clean installed them right?

I got this issue before, when i downloaded AMD drivers to stress test a 5700XT i was repairing and when i put my Nvidia back in, this was a similar issue, just not as bad.
I used DDU program. Followed the steps as shown. No problem there
 

Kickim

Bronze Level Poster
1618172206774.png


I finished another run in Furmark and also had this app in the background to check the system while in load. What do you think of it? Is it looking good? The power drawn by the gpu seems a bit low but I might be wrong?
 

SpyderTracks

We love you Ukraine
The power use does seem low, I'd expect nearer 300w with Furmark as it's about the most intensive GPU load you can simulate.

But the clocks look good and the temps appear fine.

What would be more helpful though is to track the GPU usage over the period that Furmark is running and then outputting a report at the end.

You can do this with GPU-z and output to a file, then plot the graph



Start it logging just before you run Furmark and then end the logging just after you finish it, so we can see on the logs where the GPU power is started and finished. Run GPUz for around 10 minutes and make sure it's at full resolution.
 

scarpa1

Enthusiast
View attachment 24806

I finished another run in Furmark and also had this app in the background to check the system while in load. What do you think of it? Is it looking good? The power drawn by the gpu seems a bit low but I might be wrong?
I don't use HWInfo normally but downloaded it and ran a Furmark benchmark and your average GPU clock is way lower than mine. 1042mhz compared to 1875mhz

The min GPU clock is way lower too, yours 225mhz, mine was 1605mhz.

And what is GPU Hotspot Temp? as my max only reached 70c and current and mins were in the 40's. Yours are hitting over 100c.

Mine is a 3080 and liquid cooled and no idea if the GPU hotspot temp is important or not but those numbers are way higher than mine temp wise
 

Kickim

Bronze Level Poster
The power use does seem low, I'd expect nearer 300w with Furmark as it's about the most intensive GPU load you can simulate.

But the clocks look good and the temps appear fine.

What would be more helpful though is to track the GPU usage over the period that Furmark is running and then outputting a report at the end.

You can do this with GPU-z and output to a file, then plot the graph



Start it logging just before you run Furmark and then end the logging just after you finish it, so we can see on the logs where the GPU power is started and finished. Run GPUz for around 10 minutes and make sure it's at full resolution.
I will do that in a second

Also can you have a look on the gpu statistics on idle? It doesn't seem right
1618176624804.png
 

Ash_

Master Poster
Your card appears to be throttling. Perhaps it is having an insane thermal meltdown, too fast for the sensors to even log properly.

Looks like a complete card failure and you’ll need an RMA. Is it a gigabyte card?
 

Ash_

Master Poster
Where are you getting that from?
Clock speeds are correct at idle. As soon as it is under load, it goes to extremely low speeds. However, It also appears to be undervolting for some reason, but the voltage setting has already been confirmed as being set at 100%. I imagine a component has failed in the card.
 

Kickim

Bronze Level Poster
Your card appears to be throttling. Perhaps it is having an insane thermal meltdown, too fast for the sensors to even log properly.

Looks like a complete card failure and you’ll need an RMA. Is it a gigabyte card?
It's a Zotac. I'll have a talk with someone from pcs...hopefuly they can RMA it
 

Kickim

Bronze Level Poster
The power use does seem low, I'd expect nearer 300w with Furmark as it's about the most intensive GPU load you can simulate.

But the clocks look good and the temps appear fine.

What would be more helpful though is to track the GPU usage over the period that Furmark is running and then outputting a report at the end.

You can do this with GPU-z and output to a file, then plot the graph



Start it logging just before you run Furmark and then end the logging just after you finish it, so we can see on the logs where the GPU power is started and finished. Run GPUz for around 10 minutes and make sure it's at full resolution.
 

Attachments

  • GPU-Z Sensor Log 1.txt
    210.6 KB · Views: 232

Ash_

Master Poster
It more than doubles its power draw within a second and then the whole card appears to throttle.

Can you set your fans to 100% and re-test it and monitor it again.
 

Kickim

Bronze Level Poster
They were already at 100 but to be sure I left the firestorm open in the background. I mean. The fans are almost taking off i can't do more then that:))))
 

Attachments

  • GPU-Z Sensor Log 2.txt
    107.7 KB · Views: 227

SpyderTracks

We love you Ukraine
So it does appear that that Hotspot Temp suddenly hugely rises to 107c along with wattage within a very short space of time and fans ramp up to 100%, but clocks throttle right down to just over 200MHz

Definitely something wrong with that card.

At no time is it more than 50% used.

Just change the extension of the file to .csv and open in Excel, makes it much easier to read.
 
Last edited:

SpyderTracks

We love you Ukraine
Just as an additional, you can RMA just the GPU in your online account, no need to phone, if you link this thread in the RMA, that should give them the evidence they require to see that there's a GPU fault.

You can just RMA the GPU, they should then be able to send a replacement GPU for you to fit yourself next day shipping, DPD will drop off the new GPU and you give them the old one for them to return.
 

Kickim

Bronze Level Poster
Just as an additional, you can RMA just the GPU in your online account, no need to phone, if you link this thread in the RMA, that should give them the evidence they require to see that there's a GPU fault.

You can just RMA the GPU, they should then be able to send a replacement GPU for you to fit yourself next day shipping, DPD will drop off the new GPU and you give them the old one for them to return.
Thank you for your help. I will have a talk with someone tomorrow and also use the forum thread as a reference. Hopefully guided by someone on the phone I can use my informations properly this time in my RMA application
 
Top