Posts Tagged ‘RTX 30 series GDDR6X temperature

The latest HWiNFO v6.43-4380 Beta version is available now with a number of improvements and new features, but the most interesting new addition is the added monitoring of GPU HotSpot temperature for NVIDIA GPUs. Recently this free tool added the ability for users of the new Nvidia RTX 3080 and RTX 3090 GPUs from Nvidia to monitor the operating temperature of their GDDR6X memory chips, giving access to the hottest operating temperature. The GDDR6X memory chips apparently start to throttle at around 110 degrees and they do run pretty hot under high loads such as Ethereum (ETH) mining for instance. The new feature is not focused on the memory, but the GPU and its operating temperature and more specifically what is the hottest spot based on the sensors inside the graphics processor.

HWiNFO v6.43-4380 Beta Changelog:
– Enhanced sensor monitoring on MSI H510 series mainboards.
– Fixed monitoring of +12V on some ASUS Z590 series mainboards.
– Improved support of LPDDR3/LPDDR4/LPDDR4X memories.
– Fixed a possible WHEA error/system crash during long-term monitoring of AMD RX 6000 series GPUs.
– Enhanced sensor monitoring on ASRock Z590 Taichi and Z590 PG Velocita.
– Consolidated AMD Navi monitoring, added Effective Clock and TGP Power for Navi21.
– Added sensor tool tips to provide more details about sensors and their values.
– Improved sensor UI responsiveness during some operations.
– Improved handling of disappearing/reappearing sensors in custom order.
– Fixed a possible hang when logging sensors with large sets of disabled items.
– Added monitoring of GPU HotSpot temperature for NVIDIA GPUs.

Tools such as MSI Afterburner report the operating GPU temperature in the form of value taken from a number of sensors inside the GPU and passed through a special averaging algorithm. The graphics processor is a large crystal with a lot of transistors inside and different parts of it can be under load while others may be idle at times during its normal operation. This results in temperature variation across the crystal and there are a number of sensors inside that are tracking this difference in temperatures and a special algorithm that averages the data into a single numerical value that you get as “the real” GPU operating temperature.

The HotSpot temperature measurement that HWiNFO does now report in the latest beta version is showing the hottest point as measured on the GPU by the sensors, this value is higher than the normal average temperature you get from other programs such as MSI Afterburner. Do note that it is normal to have a big delta between the hotspot and the regular average temperature reported (10-15 degrees is fine), this however may be very helpful in finding potential problems with the GPU or the cooler of the video card while mining (if more than 15 you should probably look into it). Especially if you are having problems with a GPU being unstable while mining, but otherwise seeming to operate normal and with low regular reported temperature. If there is a hotspot with a very big different there could be an issue with the thermal grease in some point of the GPU, or the cooler not properly seated or something else causing the problem.

A new tool to help diagnose possible problems with Nvidia GPUs that do not behave as expected while mining. We did a quick test on an ASUS RTX 3080 TUF Gaming where MSI Afterburner reports 43 degrees Celsius as average operating temperature for the GPU, while HWiNFO reports a hotspot with 56.8 degrees Celsius with the card at 100% load with about 230W power used and the fans at 100% while mining Ethereum (ETH). This measured is a delta of 13.8 degrees between the average and the hottest spot, the coldest spot is probably with even higher difference, so maybe it will be a good idea to also report the coldest spot as well and not only the hottest one… maybe in another update of HWiNFO.

To Download and Try the Latest HWiNFO v6.43-4380 Beta HotSpot monitoring feature…

One of the most problematic sides of using the latest Nvidia GeForce RTX 30 series GPUs for mining Ethereum is the fact that their video memory gets really hot and up until recently there was simply no tool to give you an idea how hot that is. Thanks to the latest HWiNFO version 6.42 you can now monitor the operating temperatures of the GDDR6X video memory of your RTX 3080 and RTX 3090 GPUs and take the appropriate measures to keep things cool for ensuring maximum performance and problem free operation on the long run. Have in mind that you might actually be quite surprised when you see the actual numbers being really high, but that is to be expected, considering the fact that even the surface of the backplate of the video card gets quite hot when we touch or measure it.

Just download and run the latest HWiNFO, you can even use the portable version, go to Monitoring and start Sensor status and then scroll down to the GPU data and see what you get reported as GPU Memory Junction Temperature data (right under the GPU Temperature). On the left image you can see the idle video memory temp we see on an ASUS EKWB GeForce RTX 3090 GPU and on the right the temperature of the video memory after we run PhoenixMiner with the tweaked RTX 3090 in order to give us 120 MHS hashrate for Ethereum mining. We get 36 degrees idle temperature, but when mining it bumps up to 92 degrees Celsius and this is a water-cooled GPU with the GPU temp going just about to 50 C under load.

A note regarding the new GDDR6X Memory Junction Temperature:
Just like in case of Navi, this is not the external (case) temperature, but internal junction temperature measured inside the silicon.
So don’t be scared to see higher values than other common temperatures, it’s expected. Also the limits are set respectively higher (throttling starts around 110 C).
EDIT: Adding that the value reported should be the current maximum temperature among all memory chips.

Checking the backplate of the water-cooled RTX 3090 with a FLIR thermal camera showed that at stock settings when mining you get around 72.8 C (the hottest spot) and with overclocked memory the hottest spot on the backplate where the memory chips are is 75.8 C. No wonder we got such high temperatures measured considering that the actual operating temperature of the memory is 92 degrees under the backplate. So, adding extra cooing fans on top of the backplate is definitely a must if you plan on using RTX 3090 for mining Ethereum. The same also goes for RTX 3080 when used for mining, though with the RTX 3070 and 3060 Ti the situation could be better due to the lower power usage they have and the GDDR6 memory they use (not the faster and hotter GDDR6X), especially when optimized for mining. We need to further explore the memory temps with these two, meanwhile if you check what temperatures you are getting with HWiNFO on your GPUs feel free to share results in the comments below.

Update: The just released updated HWiNFO v6.43–4380 Beta version comes with added support for Monitoring GPU HotSpot Temperature for Nvidia GPUs, so you might want to also check out this new parameter when using Nvidia video cards for mining crypto currencies such as Ethereum (ETH). This feature can be helpful in detecting issues with problematic GPUs that otherwise do seem to be operating fine, but are not running stable when mining, because of potential problems with proper cooing of the whole graphics processor due to bad thermal paste, not good contact with the cooler for the whole chip surface etc.

To download and try GDDR6X monitoring with the free HWInfo 6.42 diagnostic software…


top