Blog

Oct 11, 2023

Optimal CPU & GPU Temperature for Gaming

Long term effects of consistently high temperatures? from cooking the “main”

Long term effects of consistently high temperatures?

from cooking the "main" silicon? probably not, at least not for a very long time in a consumer use case.

cooked at max temp 24/7/365 (say for example in a computational cluster or other system that would constantly utilize the GPU at or near 100%)….

You might start to see a standard deviation from day 1 performance at 10-15 years that could be attributed to solid-state diffusion of defects, impurities, GCR or other radiation damage, etc. in the silicon (which could very well include the memory or other microcontrollers on the GPU card and not the GPU itself)

This is almost certainly true, even with the most modern fabrication technology like 3 or 4nm*

Anything faster and it would be widespread enough that it would be common knowledge.

*remember: the whole "process node" thing with regard to the number of nanometers no longer has any relationship to anything like "gate pitch" or "feature size" in reality.

All that means is that the TRANSISTOR DENSITY of the current process node is >6x 45nm (for the case of 7nm) or >15x for 3nm.

There is not a gate, or fin, or metal trace, or whatever that is 7nm… (or half of whatever feature or metric)

Even that is misleading, because what you actually obtain in your commercial product, if you were to grind it and put it in an AFM Or SQUID or whatever would not have anything near the "maximum capability" of the current process node. Those marketing terms…. 7nm or 3nm are more meaningful to the businesses that buy foundry services (eg AMD or Qualcomm) than to a regular person or end user of the packaged silicon.