I’ve had a question about the GPU Silicon Die for a long time and I was wondering about it’s silicon die. On the GPU Silicon Die, I know it has the “Cores”, but it also has mem controllers on it and other things. Does everything on the DIE operate at 1.1v, or is other voltages being supplied to it to support the different domains or parts of the DIE? I heard a guy say since the GPU chip has to communicate with the Video Bios, it has to operate at the same 3.3v as the bios chip of that video card. So is there multiple voltages that the GPU DIE operates at since it communicates with various things, like the VRAM, the HDMI, the VBIOS and the PEX, etc?
Hi there @rideshareelliott, welcome to the NVIDIA developer forums.
Running different parts of a complex modern processor at different voltages is quite common and NVIDIA GPUs are no exception. In addition to that there is the use of DVFS, Dynamic Voltage and Frequency scaling, to adjust to any specific needs.
However the detailed architecture of how any of these concepts are applied in NVIDIA GPUs is something that cannot be shared in a public forum, I am afraid.
Thanks!
Okay, thank you Markus. I don’t want to relay any proprietary information. Only if, generally speaking, GPU dies or maybe somewhere on the package, that it’s responding to these chips at the same voltages, or maybe it has a voltage regulator to keep high voltages from damaging the die, like using the Dynamic Voltage and Frequency scaling to adjust the voltages as not to damage the die directly. Just curious, because I know many people are of the thinking that the die only receives .9v or maybe 1.1v to all areas of the die and it only operates at one voltage and it must be transferred through the GPU PWR and Return pins. The questions were asked simply to see if maybe they could check to see if a GPU chip had actually failed and if they could check such a thing through these voltages between the Chip and it’s corresponding parts.