3 ms·

It's fascinating that I heard about homebrewers doing this kind of thing around a decade ago (although I believe they were keeping the hard drives un-submerged

It's fascinating that I heard about homebrewers doing this kind of thing around a decade ago (although I believe they were keeping the hard drives un-submerged rather than sealing them up) yet this is only now being considered "cutting edge" as a commercial option.

yungchin 14y ago
Yup, I was reminded of those too! Feeding the nostalgia: http://web.archive.org/web/20010623203101/http://www.hardwareoc.com/submersed.php http://web.archive.org/web/20010623203101/http://www.hardwar...
- mistercow 14y ago
  I like how the author of that completely did not understand how heat conduction works: Just submersing your components in a liquid doesn't do you much good as the liquid will be room temperature and your CPU will burn up just as fast as if it were in the open. Well no, because see, the thermal conductivity of air is about 0.026 W/mK and the thermal conductivity of mineral oil is at least 0.1 W/mK and oh god I'm trying to argue with someone 11 years in the past.
sliverstorm 14y ago
It's just a reminder that "commercially viable" is not the same thing as "possible".
- mistercow 14y ago
  Yet I wonder what really has changed that makes it commercially viable now. Pumping fluids around isn't a technology that has significantly advanced, after all. Is it just that Moore's law has ceased to apply to heat and power usage? Is it that CPU manufacturers finally got around to testing and falsifying the long-assumed-true hypothesis that servers need to be kept in ultra cool rooms to begin with?
  - sliverstorm 14y ago
    I spent some time reflecting on this, and I think it's because what the bottleneck is has changed over the years. For a while, I/O was the real challenge for servers. But hard drives have improved a ton, and many-drive systems have really improved a lot. Memory had its day, but the price has plummeted. Today, the limiting factor everyone is focusing on is core density, and cores generally produce more heat than any other component aside from high-end graphics and displays. Of course, power consumption and thus heat has been continually improving, but not by orders of magnitude. Core density, on the other hand, has been going up by a few orders. As a result, you've got to move more heat, and because density is what's going up, you've also impaired airflow. Take this with a grain of salt though, it's just speculation based on what I've seen & heard.
stevenrace 14y ago
The Cray-2 (1985) supercomputers featured 'liquid immersion' cooling - the cubes of processors were dropped into a bath of inert oil [1][2][3]. [1] http://en.wikipedia.org/wiki/Fluorinert http://en.wikipedia.org/wiki/Fluorinert [2] Seymore Cray speaking about liquid cooling http://www.youtube.com/watch?v=QTfi_gNPuh0&list=FLZwq3blS9UIN2JpCl_2V4jw&t=14m40s http://www.youtube.com/watch?v=QTfi_gNPuh0&list=FLZwq3bl... [3] Visible cooling fluid within a 90s era Cray at the NSA: http://www.youtube.com/watch?v=BH8X8w8a4f4#t=23s http://www.youtube.com/watch?v=BH8X8w8a4f4#t=23s

yungchin 14y ago

Yup, I was reminded of those too! Feeding the nostalgia: http://web.archive.org/web/20010623203101/http://www.hardwareoc.com/submersed.php http://web.archive.org/web/20010623203101/http://www.hardwar...

mistercow 14y ago
I like how the author of that completely did not understand how heat conduction works: Just submersing your components in a liquid doesn't do you much good as the liquid will be room temperature and your CPU will burn up just as fast as if it were in the open. Well no, because see, the thermal conductivity of air is about 0.026 W/mK and the thermal conductivity of mineral oil is at least 0.1 W/mK and oh god I'm trying to argue with someone 11 years in the past.

sliverstorm 14y ago

It's just a reminder that "commercially viable" is not the same thing as "possible".

mistercow 14y ago
Yet I wonder what really has changed that makes it commercially viable now. Pumping fluids around isn't a technology that has significantly advanced, after all. Is it just that Moore's law has ceased to apply to heat and power usage? Is it that CPU manufacturers finally got around to testing and falsifying the long-assumed-true hypothesis that servers need to be kept in ultra cool rooms to begin with?
- sliverstorm 14y ago
  I spent some time reflecting on this, and I think it's because what the bottleneck is has changed over the years. For a while, I/O was the real challenge for servers. But hard drives have improved a ton, and many-drive systems have really improved a lot. Memory had its day, but the price has plummeted. Today, the limiting factor everyone is focusing on is core density, and cores generally produce more heat than any other component aside from high-end graphics and displays. Of course, power consumption and thus heat has been continually improving, but not by orders of magnitude. Core density, on the other hand, has been going up by a few orders. As a result, you've got to move more heat, and because density is what's going up, you've also impaired airflow. Take this with a grain of salt though, it's just speculation based on what I've seen & heard.

stevenrace 14y ago

The Cray-2 (1985) supercomputers featured 'liquid immersion' cooling - the cubes of processors were dropped into a bath of inert oil [1][2][3]. [1] http://en.wikipedia.org/wiki/Fluorinert http://en.wikipedia.org/wiki/Fluorinert [2] Seymore Cray speaking about liquid cooling http://www.youtube.com/watch?v=QTfi_gNPuh0&list=FLZwq3blS9UIN2JpCl_2V4jw&t=14m40s http://www.youtube.com/watch?v=QTfi_gNPuh0&list=FLZwq3bl... [3] Visible cooling fluid within a 90s era Cray at the NSA: http://www.youtube.com/watch?v=BH8X8w8a4f4#t=23s http://www.youtube.com/watch?v=BH8X8w8a4f4#t=23s