I posted earlier, but it's been rattling around in my head, so I'm gonna rattle some more thoughts out.
Noise: The a/c unit is noisy. I can't vouch for the garden pump. My original thought was a capillary system, which wouldn't work well with viscous coolants. Well, not at any kind of speed. You could get a capillary system to work, but you could spoon it faster. I'd think submerging the pump as well would silence it fairly effectively, not to mention the a/c unit.
Coolants: There have been several ideas for this tossed about, including freon and liquid nitrogen. The problem with those? Maintaining the coolant system, and keeping it loaded. Alcohol would work, except for the 'Great Balls of Fire' factor. What about synthetic motor oils? The ones with lower viscosity ratings, or geared for high heat environments, would be a better bet. The thermal transfer qualities are undoubtedly higher than those of the mineral oil.
'Son, have you seen those two quarts of 10w/30 I bought for your mom's car?'
"Uh, no, Dad, sorry."
One reply to my original post suggested an enamel casing. Taking this a step further, why not sheath the motherboard in clear flexible rubber or latex (Trojan condoms, eat yer heart out), and doing some fancy work around the expansion and memory slot, and then injecting coolant in one corner, and drain it from the other? I'd think this would lend itself better to a more managable self contained environment, that makes upgrades easier to cope with.
Alternatives to that still include submersing the entire motherboard, but using some kind of bus cable extender to give you use of the expansion slots, and some funky modification to 'ram expanders' to raise your memory chips out of the goop.
I'm still rather keen on the singular application of an enclosed reservoir and circulation mechanism cooling just the processor heat sink, in a kit just the right size to fit in an 5 1/4 inch drive bay.
Instead of using the pump and air conditioner, you might be able (a bit of work, though) to build a capillary system to use the adhesive/cohesive properties of the coolant to induce a sort of gravity driven flow of the coolant. I'm not sure that mineral oil would work well with that, though, but there are other non-conductive fluids out there that would work better. Someone mentioned the low thermal transfer qualities of mineral oil earlier, so that would be another issue.
Beyond that, I applaud Dr. Freeze for the innovation. It opens up a lot of ideas.
The obvious issues are component installation and upgrade, as shutdown, removal, cleaning, install, and reset of the containment system aren't exactly convenient.
What electrical issues are raised by extending expansion slot connections, ala bus cable extensions? I know that may not be a very good idea where SIMM/DIMM slots are concerned, but (overkill) a vacuum sealed slot cover would be good in keeping the coolant off the slots.
As a more focused approach, what about 5 1/4 inch drive casings for 3.5 inch drives, implementing a similar setup? I have a pair of 10k rpm scsi drives that get HOT.
Coolant flow over just the processor heatsinks is something that I see has been done, as well, but there's applications in that, as well. Come to think of it, that's not a bad idea for something you could plug into an empty 5 1/4 inch slot. small pump and coolant reservoir, transfer tubing to and from a top-sealed heatsink. Heat exhaust could be aided with a heatsink or cooling fan on the reservoir, running hot oil to the front of the chassis for external dispersion, drawing cooler oil from the rear of the reservoir.
Okay. I thought for five minutes. I'm gonna stop now. =)
oh damn. that was bad. Hang on, autoLART time. (Hit enter instead of tab after typing in the subject. I'm not dyslexic, but my fingers are.)
Anyway, CmdrTaco got mentioned over at userfriendly.org for the death of a site..
From userfriendly (so you don't kill my other favorite site by going an looking!):
CMDRTACO DOES IT AGAIN 00:04 PST 15 FEB 1999 This is too cool. The Peer 2 Peer launch was mentioned on Slashdot yesterday, bringing Arcterex's server into a screaming death spiral as the infamous "Slashdot Effect" sent thousands and thousands of clicks over to the wee box. Arc wants you all to know that it's fixed now.:)
They're all on-board devices, integrated into the motherboard. Whether they're supposed to be that way or not, I'd prefer they be on separate interrupts. I'm a purist that way. Coward. =)
Yup, it's an issue. I recently experienced the joy of installing RedHat on a brand spanking new out of the box Compaq Deskpro (P333, big fat hard drive, lotsa ram).
It sucked.
The boot process hung on the CDrom, whether I was booting from floppy, or the functioning CDrom itself. Turns out Compaq's out of the box config is cable select slave for the CDrom. It *works*, but it causes problems. Windows doesn't seem to care, but hey, I expect that.
On top of that, the on board Ethernet, USB interfaces, and the Video *SHARE* IRQ 11. Luckily for Compaq you can change this in BIOS, or I'd be building the next Mars Pathfinder out of this fat desktop case. In retrospect, I'm thinking this may not be such a bad thing. It's good for the clue quota, anyway.
Beyond these two glaring caveats, I'm pretty happy with the Compaq. I can routinely bring any machine to it's knees, but this one performs pretty well under a good heavy load of E DR.15, x11amp, bladeenc, Netscape, a teeming horde of Eterms, and numerous other X apps.
Oh, and Redhat Support cheerily reports that Rasterman is doing well on his diet of bread and water, and that they gave him a bigger cage. =)
Slashdot Mirror
Noise:
The a/c unit is noisy. I can't vouch for the garden pump. My original thought was a capillary system, which wouldn't work well with viscous coolants. Well, not at any kind of speed. You could get a capillary system to work, but you could spoon it faster. I'd think submerging the pump as well would silence it fairly effectively, not to mention the a/c unit.
Coolants:
There have been several ideas for this tossed about, including freon and liquid nitrogen. The problem with those? Maintaining the coolant system, and keeping it loaded. Alcohol would work, except for the 'Great Balls of Fire' factor. What about synthetic motor oils? The ones with lower viscosity ratings, or geared for high heat environments, would be a better bet. The thermal transfer qualities are undoubtedly higher than those of the mineral oil.
'Son, have you seen those two quarts of 10w/30 I bought for your mom's car?'
"Uh, no, Dad, sorry."
One reply to my original post suggested an enamel casing. Taking this a step further, why not sheath the motherboard in clear flexible rubber or latex (Trojan condoms, eat yer heart out), and doing some fancy work around the expansion and memory slot, and then injecting coolant in one corner, and drain it from the other? I'd think this would lend itself better to a more managable self contained environment, that makes upgrades easier to cope with.
Alternatives to that still include submersing the entire motherboard, but using some kind of bus cable extender to give you use of the expansion slots, and some funky modification to 'ram expanders' to raise your memory chips out of the goop.
I'm still rather keen on the singular application of an enclosed reservoir and circulation mechanism cooling just the processor heat sink, in a kit just the right size to fit in an 5 1/4 inch drive bay.
Still thinking.
Instead of using the pump and air conditioner, you might be able (a bit of work, though) to build a capillary system to use the adhesive/cohesive properties of the coolant to induce a sort of gravity driven flow of the coolant. I'm not sure that mineral oil would work well with that, though, but there are other non-conductive fluids out there that would work better. Someone mentioned the low thermal transfer qualities of mineral oil earlier, so that would be another issue.
Beyond that, I applaud Dr. Freeze for the innovation. It opens up a lot of ideas.
The obvious issues are component installation and upgrade, as shutdown, removal, cleaning, install, and reset of the containment system aren't exactly convenient.
What electrical issues are raised by extending expansion slot connections, ala bus cable extensions? I know that may not be a very good idea where SIMM/DIMM slots are concerned, but (overkill) a vacuum sealed slot cover would be good in keeping the coolant off the slots.
As a more focused approach, what about 5 1/4 inch drive casings for 3.5 inch drives, implementing a similar setup? I have a pair of 10k rpm scsi drives that get HOT.
Coolant flow over just the processor heatsinks is something that I see has been done, as well, but there's applications in that, as well. Come to think of it, that's not a bad idea for something you could plug into an empty 5 1/4 inch slot. small pump and coolant reservoir, transfer tubing to and from a top-sealed heatsink. Heat exhaust could be aided with a heatsink or cooling fan on the reservoir, running hot oil to the front of the chassis for external dispersion, drawing cooler oil from the rear of the reservoir.
Okay. I thought for five minutes. I'm gonna stop now. =)
oh damn. that was bad. Hang on, autoLART time. (Hit enter instead of tab after typing in the subject. I'm not dyslexic, but my fingers are.)
:)
Anyway, CmdrTaco got mentioned over at userfriendly.org for the death of a site..
From userfriendly (so you don't kill my other favorite site by going an looking!):
CMDRTACO DOES IT AGAIN
00:04 PST 15 FEB 1999
This is too cool. The Peer 2 Peer launch was mentioned on Slashdot yesterday, bringing Arcterex's server into a screaming death
spiral as the infamous "Slashdot Effect" sent thousands and thousands of clicks over to the wee box. Arc wants you all to know that
it's fixed now.
They're all on-board devices, integrated into the motherboard. Whether they're supposed to be that way or not, I'd prefer they be on separate interrupts. I'm a purist that way. Coward. =)
Yup, it's an issue. I recently experienced the joy of installing RedHat on a brand spanking new out of the box Compaq Deskpro (P333, big fat hard drive, lotsa ram).
It sucked.
The boot process hung on the CDrom, whether I was booting from floppy, or the functioning CDrom itself. Turns out Compaq's out of the box config is cable select slave for the CDrom. It *works*, but it causes problems. Windows doesn't seem to care, but hey, I expect that.
On top of that, the on board Ethernet, USB interfaces, and the Video *SHARE* IRQ 11. Luckily for Compaq you can change this in BIOS, or I'd be building the next Mars Pathfinder out of this fat desktop case. In retrospect, I'm thinking this may not be such a bad thing. It's good for the clue quota, anyway.
Beyond these two glaring caveats, I'm pretty happy with the Compaq. I can routinely bring any machine to it's knees, but this one performs pretty well under a good heavy load of E DR.15, x11amp, bladeenc, Netscape, a teeming horde of Eterms, and numerous other X apps.
Oh, and Redhat Support cheerily reports that Rasterman is doing well on his diet of bread and water, and that they gave him a bigger cage. =)