Low GPU memory clock readings?

[Maxiaid]

Hi. I'm trying to display the VRAM clock speeds in a sensor panel, but the amount that AIDA64 reads is about 8 times smaller than what the Nvidia App shows.

When AIDA64 says 50 MHz, Nvidia App says 405 MHz; when AIDA64 says 101 MHz, Nvidia App says 810 MHz; when AIDA64 says 1337 MHz, Nvidia App says 10701 MHz; etc.

There is correlation between the two, since they both go up and down at the same time and with a fixed rate.

Card is an RTX 4090. At first I thought it might be counting the speed for each individual memory chip, but then I checked the PCB and there's 12 chips (2GB each) in total, so x8 it doesn't make sense to me.

Why could this be, and is there a way to make the reading match what the Nvidia App states? Thanks in advance.

[LeLoT3]

I'm facing the same annoying issue.

I think it could be related to the last update. Before my 4080 was showing 22400MHz, now it shows 1400MHz

 

One user on Reddit told me this to understand the numbers:

 

Quote

1400MHz is the clock speed of the VRAM, GDDR6X transfers 16 times per clock, so 1400MHz GDDR6X transfers at 22.4Gbps (the effective speed). It can also be shown as 22400MHz effective speed, however note it a calculated estimate of performance and is not running at 22400MHz.

If it is listing as 11.2Gbps it could be because it is not calculating the final doubling.

It is similar to how DDR5-6000 has a clock speed of 3000MHz but an effective speed of 6000MHz (or more properly designated 6000MT/s).

Graphics memory is much faster per clock cycle.

 

Still doesn't help with the issue or bug, but I understand the number now.

It would be nice for AIDA64 fix this bug.

[Maxiaid]

Thanks for clarifying, @LeLoT3.

I don't know if it was working as intended before the last update, I wasn't monitoring this parameter before.

Still, my ratio is 8x instead 16x, so the same reasoning must apply, only halved for some reason.

Hope the devs fix this.