Question RX 570 GPU Clock unstable and wrong temperatures ?

[Okye]

Hi all! I have a MSI Rx570 Armor 4GB, I made it a very slight undervolt, nothing weird. Since then, it works normally but when opening a game or some benchmark the graph raises a lot of temperature, it goes from (46 ° C) to (91 ° C) in a matter of seconds, when using a thermal camera the temperature of the core does not exceed (60 ° C), which tells me that the sensors of the graphics card are faulty. Also the GPU Clock starts at 1266Mhz but drops quickly to 360Mhz / 700mhz. After two minutes the video card stops giving video and the coolers work at 100% until I restart the pc and everything returns to normal.

PS: Discard that it is the PSU since it uses the same graphics but 8GB and it works perfectly. Also try flashing the BIOS but it remains the same.

Will you have some kind of solution?

Sorry for my English.

Edit:

CPU: Intel® Core™ i7-6700
Motherboard: MSI B150M Pro-VH
Ram: HyperX Fury 2x8GB HX432C16FB3/8
SSD/HDD: SSD KINGSTON SA400S37960G (894 GB)
GPU: MSI RX 570 ARMOR 4GB
PSU: Corsair VS Series VS600 600w (2 years of use)
Chassis: Corsair 110r
OS: Windows 10
Monitor: Samsung LF22T350FH led 22
Ambient room air temps: 24°C

Tested games: GTA V / IV, Forza Horizon 5, Warzone, CS: GO

Used benchmarks: Furmark 1.29.0, Superposition Benchmark. Here I have to make a point, in the benchmarks mentioned the gpu behaved as mentioned above, but I also used the benchmark that TechPowerUp GPU-Z brings, the latter the video card did not turn off, but the temperature rose quickly like the one increased fan speed. Thanks to the latter, I had time to check the temperature in the thermal camer.

 

[Lutfij]

Welcome to the forums, newcomer!

PS: Discard that it is the PSU since it uses the same graphics but 8GB and it works perfectly. Also try flashing the BIOS but it remains the same.
That's not how troubleshooting works. Make and model of your PSU and it's age? Make and model of the other GPU used and if you flashed the VBIOS, please mention the source of the VBIOS file as well as it's version.

On second thoughts, please list the specs to your build like so:
CPU:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
Monitor:

Include the ambient room air temps as well. Mentioning the names of the games and benchmarks also helps us two fold.

 

[Ralston18]

Update your post to include full system hardware specs and OS information.

Include PSU: make, model, wattage, age, condition? History of heavy gaming use?

PSU's are a critical component and must be considered as a possible suspect. There are other factors involved. Not just the GPU RAM.

Note: According to the manufacturing specs the 4 GB and 8GB GPU's list the same specs for power; Power consumption = 150 Watts, Recommended PSU = 450 watts.

Links:

https://us.msi.com/Graphics-Card/Radeon-RX-570-ARMOR-8G-OC/Specification

https://www.msi.com/Graphics-Card/radeon-rx-570-armor-4g-oc/Specification

[Do verify that I correctly identified the GPUs.]

The GPU sensors may not be faulty. The sensors may simply be responding to a temperature increase faster than the thermal camera can capture.

More information needed.

 

[Okye]

Welcome to the forums, newcomer!

PS: Discard that it is the PSU since it uses the same graphics but 8GB and it works perfectly. Also try flashing the BIOS but it remains the same.
That's not how troubleshooting works. Make and model of your PSU and it's age? Make and model of the other GPU used and if you flashed the VBIOS, please mention the source of the VBIOS file as well as it's version.

On second thoughts, please list the specs to your build like so:
CPU:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
Monitor:

Include the ambient room air temps as well. Mentioning the names of the games and benchmarks also helps us two fold.

Thank you very much for the comment, I forgot to specify my hardware. I already added it.

Update your post to include full system hardware specs and OS information.

Include PSU: make, model, wattage, age, condition? History of heavy gaming use?

PSU's are a critical component and must be considered as a possible suspect. There are other factors involved. Not just the GPU RAM.

Note: According to the manufacturing specs the 4 GB and 8GB GPU's list the same specs for power; Power consumption = 150 Watts, Recommended PSU = 450 watts.

Links:

https://us.msi.com/Graphics-Card/Radeon-RX-570-ARMOR-8G-OC/Specification

https://www.msi.com/Graphics-Card/radeon-rx-570-armor-4g-oc/Specification

[Do verify that I correctly identified the GPUs.]

The GPU sensors may not be faulty. The sensors may simply be responding to a temperature increase faster than the thermal camera can capture.

More information needed.

I used the test that TechPowerUp GPU-Z brings, the latter the video card did not turn off, but the temperature rose quickly like the one increased fan speed. Thanks to the latter, I had time to check the temperature in the thermal camer.

 

[Ralston18]

Do you have another known working PSU to swap in for testing?

 

[Okye]

Do you have another known working PSU to swap in for testing?

Yes, I used a certified Gigabyte Technology P550B 550w PSU but it works the same. That is why I also discarded the PSU from the beginning.

 

[Ralston18]

This PSU?

https://www.gigabyte.com/Power-Supply/GP-P550B#kf

Non-modular - correct?

And you did not reuse any splitters etc. from the original PSU connections

However as I understand the bigger picture PSU wattage went down from 600 watts (original PSU) to 550 watts (test PSU).

Try a higher wattage, more efficient PSU.

Not ready to rule out a potential PSU related problem.....

 

[Okye]

This PSU?

https://www.gigabyte.com/Power-Supply/GP-P550B#kf

Non-modular - correct?

And you did not reuse any splitters etc. from the original PSU connections

However as I understand the bigger picture PSU wattage went down from 600 watts (original PSU) to 550 watts (test PSU).

Try a higher wattage, more efficient PSU.

Not ready to rule out a potential PSU related problem.....

If that's exactly it, I'll try a higher voltage PSU.

Even though I tried a video card the same but 8GB and it worked perfectly? Is it possible for the consumption to change so much between two video cards of the same model?

 

[Ralston18]

The PSU must be able to reliably fulfill the the power demands of the host system. All components - not just the GPU's.

FYI:

Best Power Supplies of 2021 - Top PSUs for Gaming PCs | Tom's Hardware

Not with the immediate intent to purchase a new PSU.

Objective is to get a sense of the bigger picture and use to the provided calculator links to size a suitable PSU.

Add some wattage to have room for growth.

 

[Okye]

The PSU must be able to reliably fulfill the the power demands of the host system. All components - not just the GPU's.

FYI:

Best Power Supplies of 2021 - Top PSUs for Gaming PCs | Tom's Hardware

Not with the immediate intent to purchase a new PSU.

Objective is to get a sense of the bigger picture and use to the provided calculator links to size a suitable PSU.

Add some wattage to have room for growth.

I tested my video card on a colleague's pc, installing the corresponding drivers, but the same thing happened with a difference that this time the PC turned off completely. Another difference that I noticed is that Radeon Software Adrenalin notified me that the GPU exceeded the temperature limit while using the TechPowerUp GPU-Z test, this on my pc did not notify it. In the GPU-Z or Msi Afterburner the temperatures continue to skyrocket senselessly, as soon as the test started the temperature went from 46 ° C to 90 ° C in less than 2 seconds. I forgot to bring the thermal camera but when I touched the heatsink, it did not seem that the gpu is at the temperatures it showed.

I add the specs of the hardware where I tested it.

CPU: AMD Ryzen ™ 7 3700X
Motherboard: ASUS ROG Strix X570-E
Ram: Crucial Ballistix BL16G36C16U4B 2x16GB
SSD / HDD: Western Digital WDS100T2B0A 1TB x2
GPU: EVGA GeForce RTX 2070 XC BLACK EDITION
PSU: Cooler Master Silent Pro Gold 800W RS-800-80GA-D3
OS: Windows 10

 

[Ralston18]

If the GPU is failing on two different systems then I think that the GPU has failed in some manner.

Are you able to try re-applying GPU thermal paste and otherwise improve cooling for the GPU?

 

[Okye]

If the GPU is failing on two different systems then I think that the GPU has failed in some manner.

Are you able to try re-applying GPU thermal paste and otherwise improve cooling for the GPU?

At first I also thought it would be a thermal paste problem, so I replaced the thermal paste with Arctic Mx-4. I also thought about changing the thermal pads but the GPU has only been in use for 4 months and it seemed unnecessary to me.

Is it possible for the GPU to lower the Clock performance to protect it from the temperature it provides?

 

[Ralston18]

Try changing the OC in measured increments. Go slowly and determine if and when the temperature increase first occurs.

I found the following review that should provide some help with determining what is happening.

https://www.legitreviews.com/evga-geforce-rtx-2070-xc-gaming-graphics-card-review_208403/14

Wattages?

 

[Okye]

Try changing the OC in measured increments. Go slowly and determine if and when the temperature increase first occurs.

I found the following review that should provide some help with determining what is happening.

https://www.legitreviews.com/evga-geforce-rtx-2070-xc-gaming-graphics-card-review_208403/14

Wattages?

The 2070 GPU is my colleague's, I put it in the specs just in case. In that system I installed my video card which is the RX570 Armor which is the one with the fault. But I understood, I will look for a recommended review to help me with that.

 

[jasonf2]

There was a pretty in depth article on thermal cameras being used to measure temp a few years ago. https://www.tomshardware.com/reviews/accurate-infrared-thermal-measurements,4453-3.html .
Make sure that you do a read through here before making assumptions. The emissivity and transmittance of the surfaces being measured wreak havoc on accurate testing unless you have a benchmark worked up. It isn't as simple as point the camera at the card.