Originally Posted by lutjens
I can't for the time being. There is a bug in the drivers that disables all GPU boost if both Displayport connectors from the UP3218K monitors are connected (ie: when running at 8K)...and restores the GPU boost when one of the two DP connectors is disconnected. Working on a fix with NVidia.
I highly doubt this'll work. Even back many years ago (like 10 years and more) it was necessary to alter the placement of SMD components on the PCB to turn GeForce cards into Quadros. I imagine NVidia has further refined their hardware lockouts since then...????
I don't mind doing that either. Soldering is something I'm very comfortable with - had no issues adding the stacked shunt resistors on these Titan V's already
Comparing the Tesla V100, Titan V, and Quadro GV100 (the pics of the Tesla and Quadro that I can find at least) they seem to be physically identical on the 'back' side as far as SMD's. I have not yet found any pictures of the 'top' side of the Tesla or Quadro but as most of the component pads are populated on the Titan V, I don't think the Tesla or Quadro can have much more than perhaps the SMDs to enable and utilize the 4th HBM stack which is disabled on the Titan.
If I really have to I might end up buying a pair of Quadro GV100's and just go that route but since I have no need for 64gb of VRAM and I love tinkering, I'd rather figure out what modifications need to be done to make NVLink work on the Titans.
Just need to find a helpful soul with a GV100 to pull the vBIOS with GPU-Z to confirm whether or not the BIOS is part of the mod necessary...