Slashdot Mirror
Vapor-phase Processor Cooling
Econolinecrush writes "If even exotic water-cooling isn't enough for your processor cooling needs, there's always vapor-phase cooling. The Tech Report has an interesting review of Asetek's latest Vapochill system, an admittedly pricey cooling option, but one that manages to fight off condensation while keeping even high-end processors running at sub-zero temperatures. A little extreme? Sure, but it's undeniably cool nonetheless." I haven't seen a cooling system this intense since my organic chemistry labs.
...is that processors actually have a limit to how much they can be cooled. To be cooling it below zero is a ridiculous waste of power, money, and will only shorten the lifespan of your CPU.
To run my Cray III supercomputer in my basement.
...Just the thing for running the latest vaporware release!
at last my 486 becomes useful again!
underclock. even out of the factory, CPUs are basically overclocked for all intensive purposes. I know thats not what you all want to hear. But every advance that the chipmakers make that should be able to reduce the heat coming off a processor ends up getting put into running it faster instead. I grant its good for the MHz race, but the MHz race in the long run isn't really that good of an idea.
In SOVIET RUSSIA... erm...NSA AMERICA, the Internet logs onto YOU!
If I'm not mistaken, vapor-phase cooling is just an ordinary refrigeration cycle.
Geeks, of course, always go for the term that would sound coolest on the bridge of the Enterprise.
...
and here is the tomshardware review http://www4.tomshardware.com/howto/20030224/index. html
I don't claim I know more than I know, and if you know you know more than I know, then by all means, let me know.
Interesting nonetheless.
However, I'm guessing electro-migration is thermally activated, or at least sensitive to the temperature; does the extreme cooling mitigate the increased field strengh?
Jon.
Cool yes, but not too cold. Keep the chip in spec for stability and long life. Colder is not necessarily always better.
Kryotech used to have a product that provided vapor phase cooling for Athlons. They sold barebones systems with case, motherboard, and refrigeration, and you provided the rest. They were achieving clock rates of about 20% higher than you could get with traditional cooling techniques at the time. Plus, they were officially sanctioned by AMD in the sense that using the Kryotech cooling system didn't invalidate your processor warranty.
Using Kyrotech cooling, you could get a system running at 1 GHz when the fastest official parts from Intel and AMD were topping out at 700-800 MHz.
Then AMD and Intel increased clock speed like crazy, nobody wanted to pay $2000 for a small bit of clock rate improvement, and Kryotech decided to stop selling consumer stuff.
Is that cooling your CPU to subzero temperatures will neither increase its lifetime nor incrase its performance. It is the overall design of the system that matters (I.e. Motherboard, and memory) I know plenty lamers whose AthlonXP 2600+ system runs slooower than my Athlon 1 Ghz system, which is indeed pretty fucking sad. And yes, I will re-iterate that having the the CPU running below the junction-temperature is bound to cause problems. So is strapping a 5 pound heatsink on your core - and cracking it in process like a dumbass. Otherwise - I guess its cool. I can't wait to see some one gut a fridge, use the innards for cooling the computer (Use propane instead of freon tho)
i still don't understand why people buy into this type of (dont want to say crap, but kinda silly products), immersion cooling is more efficient, and even, as not only do you cool your cpu (on the top specfically), you cool all sides of it, and the rest of the motherboard. You can use something expensive like pure h2o, or that plasma type liquid they used in one of the Crays. Or something as cheep as oil. They use oil to cool power transformers on telephone poles. You can't forget either tho, how cool a case with a window and immersion system would look, as evertying would really glow, and you might even be able to put fish in it (depending on the type of liquid used, and if the cpu fans are in use too...sushi anyone?)
The Cray X1 is a Very Neat machine.. It employs phase change cooling.
Cray has a very nice 45 minute video with lots of good clips of the fog spray, etc:
http://cray.com/company/video/x1.html
PICARD: Data, what is our damage status?
DATA: All shield, weapons, and propulsion systems are offline. We are being drawn into the nearby star Jhi-Quwo IIV. Hull temperature is reaching maximum tolerances. I estimate destructive hull breach and core detonation in 23 minutes...22:59...22:58...
PICARD: Geordi! Is there anything we can do?
GEORDI: Well, Captain, there is one possibility...but there may be risks.
PICARD: We're in a risky situation now! What is it?
GEORDI: I've run some analysis and we may be able to slingshot around the star and get to a safe distance -- if we can survive that long. Currently we will be unable to withstand the heliosphere of that star. But there is a chance...
PICARD: Yes?
GEORDI: One of the crew members has a vapor-phase cooling system installed on their workstation. If we can utilize a tachyon pulse modulator in combination with the vapor-phase cooler and route it through the warp core and to the shield arrays, we just might have a chance...
PICARD: Make it so! Who has this vapor-phase cooling system?
GEORDI: Well, sir, uh, *cough*worf*cough*
PICARD: What? Worf?
WORF: Grrrr...(string of Klingon insults and scowling) 0v3rc1o|<3rz r00l!!11!
PICARD: *head in hands* I guess this is it. Send a message to Starfleet, and tell Dr. Crusher to meet me in my ready room.
...
Unfortunately, reliability is exactly what the Vapochill does not provide. When it comes to reliability, there is often no replacement for simplicity. I never worry about the 486 I've got in my firewall because it doesn't need a heatsink or fan. I can be fairly certain, however, that the fan on my workstation processor will seize up someday and the chip may self-destruct.
These guys can't be more wrong about the reliability. This thing may shut itself off if it overheats, but if reliability were the number one concern, I'd underclock a new chip so that even if the cooling system does fail, it's not a catastrophe.
I believe that if the packaging of processor chips was made of this material, processor manufacturers could immediately forget all about heat, heatsinks, fans, and all that other BS.
To counter arguments against NASA, they always say that space exploration has furthered technology here on Earth in ways that benefit us all. I believe that this is something they should figure out how to manufacture less expensively in order to eliminate heat problems from multitudes of electronic products and save time to market since engineers won't have to worry about this anymore.
Personally, I'd prefer to see a whole lot more work going into silent computers. Processors and video cards are at the point right now where overclocking only yields benchmark improvements - bumping that chip up 15MHz only gives you a few points in 3DMark, or UT2k3.
For actual use though, it doesn't make any difference. 150fps, instead of 130. Or 0.5 seconds faster load times. Does anybody really care any more?
I'd be a lot more interested in spending money reducing the noise output of my machine. Give me passively cooled power supplies (instead of these 3 fan monsters). Cases designed purely for better noise reduction (Antec Sonata is heading in the right direction).
Having a frozen CPU running 20% faster than it was meant to might win you a few brownie points at a LAN party, but does it actually make any difference? I doubt it.
*shrug* each to their own. I'll be impressed when I see a 3GHz P4 or Athlon running without any noticable noise.
u guys ever actually know anything before sprouting off? or what?
a CPU is not a resistor, so don't expect the power-drain to be similar to be a linear thing like a resistor.
CMOS process (read about how they work) forms a channel for one of the two transistors, and to form the channel the power needed is in the nA (nano-amp) range. to maintain the channel, similar.
but when CMOS drains power is when the pair switches - i.e. both transistors are partially ON. current flows from power to ground through a relatively low resistance channel, wasted as heat.
so, for a modern processor, (or SRAM, or whatever based on CMOS), the steady-state power is negledgable, but scales pretty much linearly to frequency - you can count the number of transistors transisting to be another factor (so 100% usage of FPU at 3GHz is different power consumption than the same of the integer unit), but it's not worth the trouble.
Conclusion? yes reducing speed reduces power MOST EFFECTIVELY, besides process (manufacture) changes.
My life in the land of the rising sun.