There have been some cooling problems with the 7900 XTX cards. It is early days now for this sort of problem to be diagnosed and bounded. Der Bauer is hard at work on it, but he doesn’t have much experience with heat pipes or vapor chambers…
For now, it looks like a single bad batch of cards with vapor chamber problems. Not a design issue. From past experience, sixty years ago at the family company, we were attempting to use heat pipes to cool a plasma.*
The three ways for a heat pipe to fail were:
- Too much working fluid (We tried various other fluids. One hundred-proof vodka worked pretty well.
- Too little working fluid. The magic formula was around a 5-molecule thick layer on the copper.
- Any contamination of the inside copper surface. Oxygen was the worst, so getting rid of the air in the water/working fluid, and cleaning the copper works best as a single step.
Der Bauer thinks that there may be a design issue with a fluid trap. Personally, I suspect a manufacturing error that allowed some air or oil vapor into the chambers through the vacuum system. Very low-torr vacuum systems are a pain to maintain. In this case, you need to heat and clean any oxides out. Then you need to inject the correct amount of working fluid without contaminating the parts of the vacuum system beyond the cold trap.
- We needed to (build systems for our customers to use to) expose litho plates for printing. Most customers used plates on the order of four to eight feet in area.* But some customers had plates the size of area rugs…
We were experimenting with a 10 to 20 kilowatt plasma torch. The electrodes looked solid but had internal channels. You guessed it, heat pipes a yard long sitting in (separate) washtub-sized water reservoirs that they boiled off. Designing a safe system to install in a printing plant ended up requiring too much floor space for the cooling system. So we ended up with pulsed Xenon low-pressure systems, where the average voltage and current were around 5 kw but the pulses were 150 A at 1-2 kV.
Current DUV or EUV litho systems are using similar or identical resists but on an area a few inches square. Lots of lenses mean they need about the same source brightness.