> Though I'm not certain, it seems reasonable to think that the key > temperature is the junction temperature, which is much higher than > the air temperature. With more efficient cooling, that is,
intel/amd spec chips as "Tcase", which makes sense, since they don't really wanting you removing the heat-spreader. it's all about delta-T and thermal resistance. immersion means lower TR, so for a given max Tcase (say, 70C), your ultimate heatsink (atmosphere) can run warmer. in the video, they were dissipating into the room, which probably means about 50C. for an air-cooled system, you need intake temperatures of 15-20C to stay under Tcase at normal fan speeds. hot side will be around 35C, so it's up to your CRAC to get that back to 15C. warm-water systems are also based on the idea of reducing TR, so that the heat rejection side can run closer to Tcase. after all, if your cluster produces a megawatt of 60C water, that's pretty useful... _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
