1 |
On 2012-11-11, Pandu Poluan <pandu@××××××.info> wrote: |
2 |
|
3 |
> And, IIRC, Seymour Cray likes to use some inert fluoride-based coolant to |
4 |
> dunk the components of his supercomputer machines. And he would even go to |
5 |
> lengths to design a "coolant fountain" that's not only functional, but also |
6 |
> decorative. |
7 |
|
8 |
Back in the 80's one of the local supercomputer companies (ETA |
9 |
Systems) built (and actually sold) machines which used CMOS CPU boards |
10 |
that ran submerged in liquid nitrogen. IIRC, they ran at around |
11 |
150MHz and achieved 10 GFLOPS which was pretty amazing at the time... |
12 |
|
13 |
However, the system software was crap. Like Cray, ETA was a CDC |
14 |
spin-off and AFAICT, all CDC system software was awful. In any case, |
15 |
the product was a commercial failure. I heard through the grapevine |
16 |
that maintenance was a headache, and lots of boards failed due to |
17 |
thermal stress when they were taken in/out of the LN2. |
18 |
|
19 |
> That's the only qualms I have Re: water-coolant. I always an afraid |
20 |
> of leaks. So, I always wimped out and use the thermal wick kind of |
21 |
> almost, but not quite, somewhat similar to liquid coolant ;-) |
22 |
|
23 |
One of the nice things about LN2 is that it doesn't make such a mess |
24 |
when there's a leak. :) |
25 |
|
26 |
-- |
27 |
Grant |