RTX 3080 MSI Ventus Thermal Throttling

[smoorc]

So I have been testing out the new card I just received and decided to test some cryptocurrency mining (etherium).

There seems to be an issue with what I can only guess is either the memory or memory VRM overheating and causing thermal throttling as this is a memory intensive application. I have the card running at almost minimum core voltage. With a very large core underclock (1100MHz 760mV), with +850 memory offset.

Right around when the card gets to ~64 degrees (well below normal thermal limiting levels). The perfcap reason updates to thermals and the core clock drops right down to ~800. I don't see any temp sensors available which would explain the reason for the thermal throttling. But it is interesting to note that when I disable my custom fan profile. The fan speed goes straight up to 100%.

As long as I keep the core temp below the 63-64 degrees by increasing the fans speed a bit, then no thermal throttling behaviour occurrs.

Regular gaming and 3dmark benchmarks have not caused this even when the temp goes up much higher to 76. Which is why I am assuming its memory related and not going to be an issue for the majority of applications.

Has anyone else experienced this? Not sure if this is going to be a general issue or if I have a faulty card

ps. am getting around 98MH/s when not thermally throttled

 

[Ranzer87]

That memory overclock seems too high in fairness remember this generation of cards have a feature similar to ecc where rather then having errors and crashing with memory overclocks you'll get worse performance might be worth dialling that oc back and retest

 

[smoorc]

That memory overclock seems too high in fairness remember this generation of cards have a feature similar to ecc where rather then having errors and crashing with memory overclocks you'll get worse performance might be worth dialling that oc back and retest

yes this was correct turns out the memory clock was unstable. But am still getting the same thermal throttling even at stock memory clocks.

 

[XeNoNF50]

When I had the Ventus it wouldn't hit the power limit at 850mv at around 1800mhz in any activity fwiw.

 

[Ranzer87]

Glad I managed to help with the mem overclock speed... Haven't heard much about underclocking not hitting boost targets though,

Dont have the 30 series cards just yet but there was an option on rtx 2080 to change temp limits for downclocks using precision x, but it might be worth looking at msi afterburner to see if that's an option

Only other thing I can think of is either checking for a different bios/driver or setting a custom fan curve to just keep temps under that limit

 

[Mosson]

Did you manage to fix this? I'm getting a similar issue.

My lounge PC (small ATX HTPC case) - Palit 3080:
Can play games and mine at the same time without throttling (Palit 3080 +100/+1100 offset 70% PWR) [mines at 97Mh/s - while not in games - 65C].

Office Machine - Gigabyte Aorus 3080 Master:
The machine in the office Fanatec P400a, side panel off, 3 intake / 3 exhaust fans (for when side on) (+60/+1000 75% PWR) which can only mine at around 70Mh/s (after briefly running at 99Mh/s). PerfCap reason is THRM also despite running at 44C.
Stock settings yield the same results also. My power setting makes no difference as the thermalCap kicks in anyway.

*edit:* it wasn't always like this. It suddenly dropped one day and doesn't recover for long anymore.