Originally Posted by Psistorm
Nice!! Hey I'm curious, you mentioned how 1.3 on the classy tool can actually equate to 1.45 - I've been slowly increasing my voltage and clocks and benchmarking in between each change and it appears that my voltage stays near exact what I set it to in the classy tool - it's actually just .008v less than what I set it to (for example, I have all three cards set to 1.25 right now and they're at 1.242v).
I kind of want to set it higher but I'm on air for the time being and don't want to blow these cards up lol. I'm curious if GPU-Z is displaying the voltage wrong or something because I'm noticing a lot of discrepancies between GPU-Z and MSI Afterburner. For example, the aforementioned voltage issue where MSI Afterburner will display stock voltages but GPU-Z will display the correct voltages (what I set them to using the classy tool), temperatures display incorrectly (GPU2 was showing at 92C in Heaven using the Afterburner monitoring but only at ~82C when I checked it in GPU-Z) and last but not least the core clock only displays correctly on whichever card I am currently monitoring on GPU-Z (but all three core clocks show correctly on afterburner). When I switch over to another card on GPU-z it just shows it at base clock, but I know this is incorrect because as I was able to increase the core clocks to the point of instability, so MSI afterburner is definitely displaying the core clocks correctly (and I'm assuming the temperature as well)
Sorry for the long/hard to follow post
GPU-Z does not display correctly. There is a post on Kingpincooling by an EVGA engineer stating as such. There also have been countless post about this as well and nobody has gotten GPU-Z to show correctly consistently - especially at higher than stock. The higher you go, the more GPU-Z is usually off.
All TIC cards load higher by .05 volts. Some will load slightly higher, but none will load lower much lower (as in nobody with a DMM has shown lower). Also, higher frequencies will pull more voltage than lower frequencies. This is supposedly more pronounced at higher frequencies and voltages. Hence why some people are .15 over what is set. Heat will also play a part as well, as the hotter the GPU is, the more inefficient it becomes. This isn't much more voltage, but it is enough that you should be aware of.
The danger with increasing voltage in my opinion is the increased wear and tear on the regulation circuitry more than the GPU itself. MrTooShort lost a golden Titan because a VRM blew, not from the core cooking at 1.4 volts. With memory, I'm worried about the memory chips because there are so many of them, and only one has to go bad for your card to be a pretty paperweight. You also have less phases/components feeding it, so any one piece causing spikes makes up a significant chunk of the power delivery for the memory.
If you are on air, eventually adding voltage will lower stability. The GK110 core is very temperature sensitive. Stability at stock may tolerate 95C, at 1.3 volts it could drop into the 60s on some chips.
I monitor my clocks using PX. There is no way to correctly monitor voltages without a DMM. I use GPU-Z to validate my card for the owner's club.