MEM Temp and Autofan

[Nerdish]

So does “Target MEM Temp” in the Autofan settings actually working for anyone?

I set up a test to purposely trigger a high fan speed based on Mem Temp but it doesn’t seem to respond. For what it’s worth and if it makes a difference at all, I’m mining with 4x RTX 3080 Suprims. Check it out:

Test Script
1. “Core Temp” set to 70 (I set this unrealistically high so I undeniably know my fans aren’t spinning up due to core temp. I realize if my core temp actual reached that high my Mem Temp would be much higher and/or frying as Mem Temp is always higher with ETH / Kawpow mining thus triggering a fan spin up)
2. “Mem Temp” set to 50 (I set this unrealistically low so I undeniably know I’m “overheating” based on my settings and the fans should spin up.
3. “Min Fan Speed” set 40% (To keep airflow going and minimize the risk of frying something)
4. “Max Fan Speed” set 100% (As to not interfere with testing)

I found this excerpt below from the System Changelog and News - #35 by hiveos post.

Target MEM TEMP
This feature is only used by AMD Vega and Navi series cards.
This rule is triggered if the GPU core temperature doesn’t exceed the set target temperature, in turn not allowing the memory temperature to exceed the set threshold.
For example; When Navi series GPUs exceed memory temperatures of 104C, they start throttling. With this feature, you’ll be able to keep your cards at optimal temperatures.

Now all of that being said, with the test parameters set as described above my fans rest at 40%… I can’t confirm the actual Mem Temp as there is no representation of it in HiveOS but I’m fairly certain it should be well above 50 triggering the fans to spin up.

Anyone have some luck or reason to believe it works as advertised?

[givo]

Read again excerpt you found - it clearly states that Mem Temp is a feature used only by AMD Vega and Navi. Since you said at the beginning of the post that you are mining with RTX 3080 - this is of no use to your case.

[Nerdish]

Shoot, I was assuming / hoping it was meant towards any GPU with a Nvidia / AMD chip, not necessarily the entire GPU being from that manufacture. I wonder why this would be though since they use the same 10 GB GDDR6X across the board for the 3080, at least from what I can tell.

Not what I wanted to hear but thanks for responding anyway. At least I can stop wasting my time trying.

[DediZones]

At nvidia, there is no memory chip sensor, I believe, so we assume that if a gpu is at 50 ° the memory will be + 20 ° or more …

So start from the principle the more the graphics chip will be cooled the more your memory will be, it is also visible if you put your ventilation to 50% you have a bad hashrate, if you put 85% your hashrate will increase.

[HaloGenius]

@DediZones

At nvidia, there is no memory chip sensor, I believe, so we assume that if a gpu is at 50 ° the memory will be + 20 ° or more …

This is not entirely true. There is no way to read the temperature on RTX 30xx series yet. On windows there is an opportunity only for 3080/3090 which equipped GDDR6X