Sujet : Re: Mrs. vallor's Linux workstation is online
De : recscuba_google (at) *nospam* huntzinger.com (-hh)
Groupes : comp.os.linux.advocacyDate : 23. Jan 2025, 21:48:05
Autres entêtes
Organisation : A noiseless patient Spider
Message-ID : <vmu9u5$1nd8n$3@dont-email.me>
References : 1 2 3 4 5
User-Agent : Mozilla Thunderbird
On 1/23/25 1:13 PM, rbowman wrote:
On Thu, 23 Jan 2025 07:17:52 -0600, chrisv wrote:
Yeah the water coolers are also now getting common. Personally I feel
the power consumption is getting out of hand, if water cooling is
needed. But then I'm not in need of a high-end CPU really using all of
its cores, either.
Years ago we used water cooling on an prototype machine. It was a 50 KW
Colpitts oscillator powered by a very large Eimac tube, not a computer.
There must be some corollary to Moore's law where power consumption is
inversely proportional to device complexity and doubles every two years.
More complicated thermal management systems become more common as one moves up in total power and system density. Plus requirements play a part too: a simple rule of thumb is that a distilled water system has ~twice the heat transfer capacity as the same design after replacing the water with antifreeze.
For forgoing liquid cooling, I can recall attending a DARPA conference on a "DARPA Hard" technology initiative they had for thermal management stuff, as I was working on a heat spreader that used Thermal Pyrolitic Graphite (TPG) to get rid of a liquid cooling loop under the chip; if memory serves, it was capable of roughly 200W/cm^2 extraction, which was good for getting the source heat concentration out of the guts and back to where we had a more generous space claim to move it off platform.
And from a Moore's Law perspective, we were investing in chip efficiency to get the original heat source load down ... was successful in doubling the efficiency and was closing in on triple when the customer lost interest due to non-tech factors. It will come back around, eventually.
-hh