GPU memory overclock decreases mhz

[Anomander43]

Hi guys,

I'm experiencing a weird behavior which I can not explain so I was hoping someone has an idea.
Here's what my default card values looks like (ASUS TUF 3080 10GB).

If I overclock the memory with anything beyond 251mhz, it downclocks to 5001mhz

I am using the latest nvidia drivers 536.67.
From what I could find is that this is some "mid-power state". Card is not reaching 70 degrees and also this happens even without any load on it, literally just applying the setting. But even if put some load, then it stays like this and doesn't increase.
Any ideas what could be the cause of this and how I can "fix" it? Thank you all in advance!

 

[AusWolf]

My guess is that your VRAM is unstable above +251 MHz (or the driver thinks so) and it clocks back to protect itself. Or they might be some hard limit imposed by the card's BIOS.

 

[RJARRRPCGP]

With Nvidia, there's not usually such a low limit of OC'ing the VRAM, that's hard coded in the VBIOS like there is with AMD. Especially with VRAM. The VRAM can probably be OC'ed a lot on Nvidia.

With the core, it typically won't reset back, but if you're exceeding the hard-coded watt limit, then the GPU will just start to get throttled down, even with the slider up.

With Nvidia, you usually can select even a max VRAM frequency value that's unstable. Then if you see instability, then you lower the slider some.

I noticed that with AMD, there's often an artificial limit that is in place before any watt limits kick in and vice-versa for Nvidia.

 

[K7SEM]

Try Nvidia Inspector

 

[RJARRRPCGP]

The only thing I can think of, is GDDR6X. It looks like GDDR6X is (real frequency x 8) While the previous non-X is (real frequency x 4)

Thus it looks like you don't clock it as high as the non-"X" GDDR6.

 

[Anomander43]

Thank you all for the suggestions!

I've uninstalled, reverted to and older driver version and what not. At an attempt of an overclock, the drivers crashed and it was fixed for the following tests. However on the next OS restart it was back. Which led me to the thought that this was not a bug, but intentional and remembered that i was experimenting with something called "NVIDIA PMM Tool". Turned out the uninstall of that stuff was unsuccessful and I had to give it another go and this time it was properly removed, and now everything is back to normal. Learned something (should test properly when using stuff like that). Take care all!