PC Graphics

[andreagtr]

Hi to all, after some years im back to team red! (And im so happy!!)

Ive buyed an used 6900xt under warranty, but it underperfoming and it had the hotspot temperature always at 110/115°c, no matter what kind of fan curve im going to set, or other things... the only way to keep it at 110, its to set power -10, and lowering a lil bit the frequencies...

The system it is a mini itx in a phanteks shift x case (back of the gpu are really close to the back of the motherboard with very poor airflow)

Ove tryed to keep out the case with very poor improvement...

My question is if it could be a thermal paste application problem, if I can replace it by myself without go out of warranty? Or I have to check something else???

Regards

[elstaci]

Your GPU card is basically overheating if the Hot Spot goes above 110c, it Maximum Operating Temperature.

Once it reaches and passes 110c the GPU starts to throttle and eventually will shut down if it reaches its thermal limit of 118c.

I would reset everything back to Defaults on the GPU card and remove the side panel of your PC and see if it runs cooler. If it does that indicate poor Air circulation inside the PC case.

One User mentioned that Mini ITX cases generally have very poor air circulation in them. So if you remove the side panel and the Hot Spot temps are lower then you know you need to somehow increase Air flow inside your PC.

[andreagtr]

Many thanks for your reply! It is ehat ive supposed to... the side pane are both removed, but the card had always the backplate close to the back of the mobo... (shift x its build to make loop systems only...).

If the problem persist with gpu out of case i had to replace the thermal paste? And i can keep the warranty?

Regards

[elstaci]

If your GPU card is still under Warranty I wouldn't do anything to the GPU card.

Any physical actions you do to the GPU card could void your Warranty. Best to open a Support Ticket and ask them if the GPU card needs to be RMAed to have it checked.

[andreagtr]

Im trying with gpu out of the case, all stock with fans at 100% (89%) at 50°c it reach 77°c on gpu and 110°c on hotspot during 3dmark, the gpu wont go over 2.2ghz more or less...

[elstaci]

Your GPU card won't never be as stressed as during a 3DMark stress test.

Does your GPU card Hot Spot still reaches 110c or over under normal conditions like when you play a game?

If it does then I suggest your open a Warranty Support ticket and see if it needs to be RMAed to be checked.

As you mentioned it is possible that the Thermal Paste or Pads are not as efficient as when you got it new anymore.

Let them know that you ran the GPU card outside your PC Case.

[andreagtr]

Ive tryed a repaste, seems no luck, on 3dmark it goes 79/112, i have to try it in games...

The warranty are almost over, a couple of months, then i have tryed...

I dont understand what could be the issue...

[elstaci]

I am sorry I just reread your Original Post and you mentioned that you bought a "Used" GPU card with Warranty?

It is possible you received a defective GPU card.

NOTE: The GPU fans are all running a maximum speed when it overheats correct?

Have tried to undervolt the GPU card to make it run cooler?

By re-pasting that Warranty may now be voided without getting a replacement. I don't know if you can get a refund or not if the Warranty is voided.

[andreagtr]

Youre right, ive choose this option because im pretto sure the warranty here its for original buyer only...

The unbelivable thing is the core temp goes 80

The junction sit at 60

And the hotspot at 112

This during game, with undervolt, custom fan and 2300 max freq...

How its possible it had a lower juntion temp than core???

[elstaci]

You are correct. If you purchased the used GPU directly from a User you have no Warranty since most Manufacturer's Warranty are Non-Transferable to a second owner. Only the original owner.

I though maybe your purchased a "Refurbished" GPU card that generally does have a very limited Warranty of a few months at best.

I don't know. The temperature that is important is the Junction Temperature or Hot Spot. That is the temperature that will cause your GPU card to throttle if it goes above 110c.

[andreagtr]

It could be a heatsink problem?

[elstaci]

Yes, it sounds like the heat sink is not making good tight contact with the GPU Chip or the heat sink fins are very dirty or has a oily film on it.

I needed to replace my Air CPU Cooler because my CPU was overheating. Found out that the CPU Cooler fins was all covered in a dried oily film from me cooking in my apartment.

Replaced it with a new 3rd party CPU Cooler and no more overheating.

[xlox]

in US it's not transferable but in Europe the guarantee follows the product .. depending where the gpu has been bought

but if the guy who sold you the card is ok , you can rma the card , and repaired (or new) card will be send to this guy who can send it back to you ... i had to do that one time because i didn't know guarantee was not transferable in US, and the guy i bought the card send it back to me as he received it, i just paid him the cost of the send

[xlox]

-delete-

[andreagtr]

Maybe ive found...

The vram pads are too long (or maybe the hole in the vrm frame its too little), the pads exeed the hole and this means too much thinkness between pcb and heatsink for gpu.

Replaced, with correct size, after 3 hours i had 45 on gpu and 64 on the hotspot!

Tomorrow i try full power but I think this was the problem 

[elstaci]

hey good troubleshooting. Did you also replace the pads when you re-pasted?

If you didn't then it is a good chance the original owner did and put in the wrong size pads.

Maybe that is why he sold it because it was overheating.

[andreagtr]

No luck, at stock freqiencies and pl it jump 115 on hotspot, gpu 80 (the heatsink stay cool, like the air the fans keep out of it...)

[InvadersMustDie1131]

Hello, so whats the normal hotspot temperature for that card while gaming? 90-95?

Thanks

[xlox]

can even be 108° if the card is oc ... some users complaining about cards going up to 110 but while using a "silence fan curve" ... so yeah depending of the conditions

my 6900xt gaming x trio oc could go up to 105 in intensive game , with a correct airflow in the case ( airflow as to be checked too in the case of high temps conducting to throttling (>=110°))