Question Need help in identifying a sensor for Vega 64.

[rahilpathan56]

Please see the image.

https://imgur.com/a/j7ErhbG

View: https://imgur.com/a/j7ErhbG

Memory temperature is not called as the 'hot spot', still I believe if it reaches 86 C might bottleneck the performance.
Which spot is it on the graphic card the sensor is? I would appreciate any help.

Thanks!

 

[Eximo]

I'm not really sure what you are getting at...

My understanding is that an array of temperature sensors exist inside the Vega and Radeon VII cards.

The memory is on the same substrate as the GPU, right next to it, so it isn't like you can do anything about it. Maybe try re-applying thermal compound to see if you can get the memory temperatures down. But if you are overclocking the memory, then a little higher temperature is to be expected. 86C is still well within spec, especially considering it is sitting right next to the GPU.

 

[rahilpathan56]

Eximo, thanks for your reply.

To elaborate my question, why only one of the temp is spiking. Rest of them are in 70's or 50's if they are under same substrate. It could be thermal paste I did reapply MX4 taking out the factory default.

https://imgur.com/a/Wq3TIGR

View: https://imgur.com/a/Wq3TIGR

Could you help me identify where could be this memory temperature would be if not under the die.
https://img.purch.com/amd-radeon-ve...NvbS9GLzkvNjkyNjEzL29yaWdpbmFsL1NjaGVtZS5qcGc

 

[Eximo]

With AMD they often used offset values, so that hotspot may be the difference to max. (Which would put max at 130C, which is reasonable for an absolute maximum) I'm pretty sure the hotspot is an average from multiple sensors as well, so it may be integrated at a different rate than GPU temperature (which itself may be a calculated value)

No doubt that memory temperature is from under the die. All that leaves is the Memory VRM temperature, which is listed (MVDD).

Gamer's Nexus did a pretty thorough investigation into the variations in height and cooler pressure on the Vega cards. You might consider looking into that. When you put it back together one of the memory chips might not be as tightly pressed against the cooler as you think. You have 2 HBM stacks, and only one Memory Temperature. They might be displaying the warmest or an average of the two.

I guess the basic question still remains. Are you experiencing throttling or problems?

 

[rahilpathan56]

Thanks for the explanation.
Offset makes sense, as I am experiencing lag after 20-30 min of heavy use on +50-60 Mhz on Core boost with power +20% up limit. I didn't even touch the mem clock yet as it is already at 86 C. I believe the GPU is limiting the performance to maintain the temps. But thanks, this tells me I have check the paste I applied and gives me something to dig into.