Is "GPU Memory Junction Temperature" accurate on AMD cards?

[Rugakot]

Hello,

I have a Sapphire Nitro+ 7900 xtx and I was checking the GPU temperatures and they overall look good, except for Memory Junction which is around 100-108C while gaming. There is no throttling of any kind on the card, the card works perfectly in all games and I have no issues at all of any kind.

I tried to look around the card (without dissasembling) and I can tell it has cooler pads on the 12 VRAMS. I still have over 2 years and a half left on warranty so I don't want to dissasemble. Also, I asked Sapphire and they said there is no AMD specification on Memory Junction temperatures so according to the logs my card works perfectly, so probably RMA for this is not even an option.

The GPU temps never go above 70C and HotSpot stays under 85C even on FurMark after 15 minutes. I also tried to look for other software that can give me the same information of this sensor and I couldn't find any, which makes me question if this Memory Junction temperature is accurate. On Furmark it hits 108C in 2-3 minutes, if I put the fans at 100% always it stays at 104C.

1. Could someone confirm? Or could the sensor be defective?
2. Does anyone know another software that also tracks this sensor so I can compare?
3. Does anyone else happen to have the same issue? I tried to look around and almost no one talks about Memory Junction temps on AMD cards, only gpu temp and hot spot.

Thank you,

 

[Martin]

The sensor isn't defective. Those temperatures as usually higher as it's measuring the internal temperature (inside silicon).
When the card exhibits such high values under normal use, it might be because the thermal pads are not properly seated. There are many articles on improving the VRAM cooling but all require disassembling the card and replacing/reseating the pads, so it might not be an option if the card is covered by warranty.

 

[Rugakot]

The sensor isn't defective. Those temperatures as usually higher as it's measuring the internal temperature (inside silicon).
When the card exhibits such high values under normal use, it might be because the thermal pads are not properly seated. There are many articles on improving the VRAM cooling but all require disassembling the card and replacing/reseating the pads, so it might not be an option if the card is covered by warranty.

Hello Martin! Thank you for your reply.

1. Do you have any information (link to an article, spec, whatever) on this how these temps work, especially for AMD cards? I only see it mentioned in 30xx series and GDDR6X as a concerning problem but 0 information on AMD cards or GDDR6 apart from HWiNFO readings.

2. Is there any other way to visualize this sensor apart from HWiNFO? I would like to compare readings to maybe another software or something. I tried to look at many but none of them have this sensor (GPU-Z, Hwmonitor, open hardware monitor, gpu shark, etc).

Yesterday I played more than 3 hours at 4k and the card was consistently at 106-108C, the tooltip of HWinfo says the max is 105C. I tried to do 15 minutes of FurMark today and it stays at 110C constantly, no throttling nor the fans ramp up.

3. Would this degrade my card over time? I hope at least if it breaks that it happens before my warranty expires. I prefer to have warranty intact instead of trying to fix a temperature that is not throttling my card or affecting it any way (from an end user perspective). What is your recommendation?

Thank you again for your time!

 

[LeDoyen]

For Micron at least, the max operating temp is 95°C or 105°C depending on the part number, same as GDDR6X.

 

[Rugakot]

For Micron at least, the max operating temp is 95°C or 105°C depending on the part number, same as GDDR6X.

Hello! Thanks for your reply.

I have GDDR6 Samsung, I think GDDR6X is exclusive of Nvidia.

Do you happen to know what is the operating temperature of this one? I tried to look for it in the Samsung website but I couldn’t find it anywhere.

 

[LeDoyen]

uuuuuuuh... i don't know if it's that particular part number buuuuut....