Quote:
|
Originally Posted by HAL-9000
This is interesting twist on Sterling cycle that I have never heard of before. Very cool (literally!). The one thing that looks like it could be this technology's downside is in their graph predicting CPU die temps for a 100W CPU vs. power input to their hypothetical PCT.
Namely, to keep said CPU "cool" at fifty degrees C, would require about 200 watts of input power from the wall into the PCT. Ouch! A pelt crushes that. But still a very interesting read, thanks for posting it.
|
Interesting observation. At a mere 40 watts more, the graph shows 0C die temp. The interesting part is the packaging size. True a 240 watt pelt has the potential to keep a die the same temp, but only if the hot side is cooled sufficiently by a watercooling loop. IMO, the additional power and footprint of that loop must be considered along with the pelt for a more fair comparison. Also, a pelt is still considered a dangerous option, fire hazard etc, unless some very extensive failsafes were built into the system. Hard to imagine millions of dollars in servers with 200+ watt pelts on them.
Also, this seems to have been researched through Intel's IA64 department, which was a while ago. Nothing seems to have come of it but that could very well just be legal issues.