GTX 970 Strix Hardware Changes?

[duncann]

Thought I'd post my observations about the 970 strix. I had to send back two of these to the place I bought them from (one from newegg and one from amazon) because of defects. One had occasional massive bouts of coil whine upon turning the computer on. Another had ramdom flashing dots on any open windows and far cry 4 started to crash immediately upon loading a save game.

One of the programs I use to get info about the card (temperature, voltage, etc.) is aida64. The first two of these cards had approximately the following while idle at desktop (windows 7 64-bit):

gpu tdp%: 10%
gpu vrm temp: ~39C
gpu vrm power: ~10W
gpu vrm current: ~11.50A

The replacement card from amazon has the following:
gpu tdp%: 10%
gpu vrm temp: 30C
gpu vrm power: 1.25W
gpu vrm current: ~1.5A

At first I thought these differences might be because of the driver. The first two I used max 344.75. The new card I used 347.09 immediately upon installation into the computer. So I decided to try 344.75 on my current card. Still get the same readings. This suggests to me a hardware change somewhere (probably vrm circuitry). I tried looking for any rev numbers on the card itself but couldn't see anything. Maybe I didn't look close enough. Bios versions on all cards were the same (I saved them all as a file using gpuz). There are 10 or 12 bytes difference between them when doing a binary comparison but damned if I know what the differences mean). Anyone else noticed anything like this? Or maybe post some your readings.

Forgot to mention I have the fans running at 36% while idle and that the first card's fans were pretty quiet, but the second and third cards have some pretty loud fans that seem to quiet down somewhat with time.

 

[EdZ]

If things are now idling at a lower power than before, then my first suspect would be that something was running in the background when you were first testing that was prompting the GPU to not drop down to it;s lowest power state.

 

[duncann]

Nothing has changed on the system though, except for uninstalling the old card (gtx 275 with 340.52) and installing the new card. Everything about the system is identical, including background processes. Also, the tdp% is the same in all three cases. If something was hogging gpu power, that number would be slightly higher in the first two cases. Also, the gpu clock and memory clock would reflect a different power state, which it didn't. In all cases, the lowest power state was the active one.

 

[Armenius]

AFAIK you can't get accurate software readings of the VRMs with NVIDIA cards. I would use GPU-Z instead and look at the core. VDDC should be around 0.84-0.85v while working on the desktop in the low power state. It does sound like the first card you had was a bad one, though.