View Full Version : Microchannel Cooled CPUs?
MacRumors
Oct 7, 2003, 06:48 AM
EE Times (http://www.eetimes.com/semi/news/OEG20031006S0028) reports that a new startup company, Cooligy, has been working on new active cooling systems for upcoming CPUs.
The system works by using active water cooling over hundreds of microchannels through which water is cycled. The cooling requirements of upcoming 90 and 65nm chips are said to be more challenging for traditional "passive" cooling systems.
Of note, Cooligy claims to have already developed prototype systems with Intel, AMD and even Apple. They expect first samples to be available "next quarter to makers of high-end systems like workstations and 1U servers."
arby
Oct 7, 2003, 06:56 AM
Could it be that the promised 3Ghz G5 will come with next gen water cooling? Apple seem to have learned their lesson after those noisy G4's ...
MongoTheGeek
Oct 7, 2003, 07:00 AM
Gee, what 1U servers are there out there...
I guess thats an ETA on the g5 XServe.
mac15
Oct 7, 2003, 07:07 AM
sounds 'cool' haha. and the pathetic joke of the week winner is (drum roll) mac15 for his craptapular 'cool' joke
Jerry Spoon
Oct 7, 2003, 07:17 AM
Originally posted by mac15
sounds 'cool' haha. and the pathetic joke of the week winner is (drum roll) mac15 for his craptapular 'cool' joke
Don't worry about that lame joke. Consider it water under the bridge.;)
mac15
Oct 7, 2003, 07:48 AM
haw dude thats even worse :)
formatc
Oct 7, 2003, 07:50 AM
"The small size, light weight, and excellent thermal performance of the Cooligy system allow tighter packing of components on the circuit board and higher reliability of individual chips as well as the entire system."
"The method was successfully modeled and prototyped in cooperation with Intel, AMD, Apple, and DARPA."
"The Intel test produced the highest performance Intel had ever seen from any cooling technology."
http://www.cooligy.com/technology.html
Originally posted by MongoTheGeek
Gee, what 1U servers are there out there...
I guess thats an ETA on the g5 XServe.
I doubt this tech is for the G5 Xserve. The G5 is a 130nm chip. And this tech is only sampling next quarter. Prob won't be actually in anything for a bit longer.
arn
paulc
Oct 7, 2003, 08:34 AM
Uh, one of the benefits of going to 90nm or 60nm fab processes is that the chips run cooler while allowing higher colock speeds. So it doesn't seem to make a lot of sense to worry about water cooling for cjips made with 90nm or 60nm fab processes... right?
eric67
Oct 7, 2003, 08:51 AM
Originally posted by paulc
Uh, one of the benefits of going to 90nm or 60nm fab processes is that the chips run cooler while allowing higher colock speeds. So it doesn't seem to make a lot of sense to worry about water cooling for cjips made with 90nm or 60nm fab processes... right?
actually it makes a lot of sense for PowerBook G5.....
how to cool it down efficiently without draining too much battery life.
hmmfe
Oct 7, 2003, 08:55 AM
Originally posted by paulc
Uh, one of the benefits of going to 90nm or 60nm fab processes is that the chips run cooler while allowing higher colock speeds. So it doesn't seem to make a lot of sense to worry about water cooling for cjips made with 90nm or 60nm fab processes... right?
90/60nm chips are cooler. Since they are smaller, though, they tend to hold the heat more than their 130nm counterparts. That is the purpose of the new coolers.
Xerocs
Oct 7, 2003, 08:56 AM
This would reduce this huge passiv cooling system...
Maybe for a new small powermac? ;) - something smaller
AidenShaw
Oct 7, 2003, 09:04 AM
http://www.reukeboom.com/art/takes/wit_tandy1989.jpg
KentuckyApple
Oct 7, 2003, 09:16 AM
This sounds good. I could use it for the super loud fan on my radeon 9800 pro!
sososowhat
Oct 7, 2003, 09:22 AM
A water-cooled chip allows for more heat to be created in a smaller chip area. My (intuitive only) understanding is that by going to smaller processes, even if each transistor generates less heat, each square millimeter generates more heat - but more transistors are packed into the area, therefore better performance.
So: the cooler chip innards allow for hotter machines.
So: this won't mean fewer fans in the unit. The heat that's now not in the chip, is still in the box, and there's more of it than ever before.
Unless, of course, you run the water right out of the case.
manu chao
Oct 7, 2003, 09:23 AM
Uh, one of the benefits of going to 90nm or 60nm fab processes is that the chips run cooler while allowing higher colock speeds. So it doesn't seem to make a lot of sense to worry about water cooling for cjips made with 90nm or 60nm fab processes... right?
90 and 60 nm chips produce less heat and use less power (good for battery life) - assuming the frequency isn't ramped up at the same time cancelling out all efficiency improvements - but produce more heat per chip surface area.
If you hold your hand into the sun in the summer, it might get warm a little bit. Use a magnifying glass to concentrate all that energy on one spot on your hand and your going to feel the difference, even if you're reducing the sunlights intensity a bit.
Chaszmyr
Oct 7, 2003, 09:41 AM
Originally posted by sososowhat
So: the cooler chip innards allow for hotter machines.
So: this won't mean fewer fans in the unit. The heat that's now not in the chip, is still in the box, and there's more of it than ever before.
Unless, of course, you run the water right out of the case.
The radiator can ofcourse vent outside of the machine (Like it does on most water-cooled PCs), or you can make use of fans. Once the heat is off the chip its easier to get rid of
magikminotaur
Oct 7, 2003, 10:06 AM
Several years ago radio shack had speakers they called ferofluid cooled, which enabled them to handle supposedly higher wattages. Although the idea of cooling with a liquid, preferably oil verses water which would more kin to leakage(viscosity of oil vs water) may be a solution.
However, these marketting ploys are no more practical than playing with the liquid nitrogen cooled superconductors of the 80's. Making a computer more mechanically complex is neither an advantage or innovation. Let Dell, the innovative comglomerate that it is, fair the liquid cooling seas. I do think that Intel will face those challenges much before any G5 or 6 will ever need such complexity.
cc bcc
Oct 7, 2003, 10:49 AM
Originally posted by mac15
sounds 'cool' haha. and the pathetic joke of the week winner is (drum roll) mac15 for his craptapular 'cool' joke
Sounds "cool", but never trust "leaked" information! :o
DharvaBinky
Oct 7, 2003, 11:51 AM
Originally posted by paulc
Uh, one of the benefits of going to 90nm or 60nm fab processes is that the chips run cooler while allowing higher colock speeds. So it doesn't seem to make a lot of sense to worry about water cooling for cjips made with 90nm or 60nm fab processes... right?
Not exactly...
Moving to a smaller process allows a chip to run faster OR cooler. That is... if you moved the G5 to a .90 micron process, and kept the clock speed the same (1.6-2.0GHz) it would run considerably cooler. However, this same shrink would allow you to run higher clock speeds at the same heat production.
Also...
someone brought up the possiblity of this reducing battery life significantly, however, I don't see it. Unlike air cooled systems, the liquid cooled system would be closed, so you the coolant would develop inertia, thus reducing the power requirements to keep it circulating. Air doesn't. It has to be constantly accellerated past the sink.
:)
Dharvabinky
reyesmac
Oct 7, 2003, 11:54 AM
I can't remember Apple ever using a brand new technology that they didnt invent for any of their products. The technology has to be around for quite a while and work for them to start using it.
All I know is that Apple better be working on how to cool these chips without resorting to making holes on all sides of their cases. They cant do that with the imac or notebooks like they can with the Powermac.
srobert
Oct 7, 2003, 12:03 PM
Excuse my ignorance but aren't water cooling systems for computers way more expensive than fans? I remember reading this article on PM:
Chip Cooling Technology (Popular Mechanics) (http://popularmechanics.com/technology/tech_tuesday/2003/8/tech_tuesday_08_05_03/index5.phtml)
frostythesnowma
Oct 7, 2003, 12:06 PM
oh i can picture a powerbook user with a water cooled machine
PB=powerbook user
AC=Apple Care Person
PB: picks up phone calls apple care
AC: thank you for calling apple tech support how may i help you?
PB:Ya hi i picked up my powerbook this morning and noticed a huge problem.
AC: and what problem is that?
PB: ITS LEAKING!!!!!!
AC:let me transfer you to the department that handles this.
apple care:You have been transfered to apple care power book tech support.
Apple Care:If you called because your powerbook is leaking and you need the
iPatch press #1
Apple Care: if your calling because your cooling system froze and you used an oven to thaw your PB out,press #2 for a list of lies on how it was a manufactureing defect.
srobert
Oct 7, 2003, 12:17 PM
Somehow this thread reminds me of MacQuariums:
http://www.applefritter.com/hacks/macquarium/showoff/jem-dowse/macquarium.jpg
Credits: ? ? ? Not sure, but it's british. If you're the auther of this marvel, just wack me until I give proper credits.
I knew this technology was coming soon. I thought of it, undoubtably after someone else did.
FlamDrag
Oct 7, 2003, 12:33 PM
I'm not an expert on the subject, but my understanding is that liquid cooling still needs airflow to cool the liquid.
This I know: Heat doesn't just magically disappear.
So, logically, the liquid has to go SOMEWHERE away from the heat to cool and come back. This by no means eliminates the need for fans. You probably need even more powerful fans to push more cool air past the water pipes to cool the water before it makes its way back to the processor.
Either way, it won't be a terribly quiet machine. Imagine a small fishtank inside of your machine, pumping water. It might be somehwat similar. Flowing water can get pretty noisy.
srobert
Oct 7, 2003, 12:38 PM
Interesting!
This french guy had an old PC working for a whole week entirely submerged in 25 liters of oil (i.e.: non conductive).
It's a different approach than pipes, but interesting nonetheless
Submerged PC (http://www.ginun.com/index.php3?page=Mods/huiloboy01.htm)
aarond12
Oct 7, 2003, 01:26 PM
Arrrgh. Someone beat me to post a smart-a$$ remark about the old-style Microchannel interface slots.
I fondly remember that Tandy 5000HD system. It was SOOO fast for its time.
For the time being, I will stick with PCI slots, thank you.
-Aaron-
sososowhat
Oct 7, 2003, 01:35 PM
A lot of people seem to be reading this as "water cooled computer". It's not.
It's basically a way to get the heat INSIDE the chip evenly distributed away from hot-spots, and then out to the edge of the chip where it can disipate in the normal way. I've no idea how much water this would take, but I'd guess it's not gallons - maybe not even ounces.
It has nothing to do with fans - except in as much as it allows chips to generate MORE total heat for the fans to get rid of because it's better distributed within the chip.
What will reduce power consumption eventually is lowering the voltage at which these things run. There are still technical barriers to doing this.
lynnpye
Oct 7, 2003, 01:50 PM
Originally posted by Jerry Spoon
Don't worry about that lame joke. Consider it water under the bridge.;)
Oh, come on. If you thought that was funny at all, you're all wet.
cubist
Oct 7, 2003, 02:29 PM
Any of you guys listen real close to the BLOWERS (not fans!) which cool the Xserve? The machine sounds like a jet engine. Working near one is enough to drive one to drink. Maybe the computer wants to tip back a cold one too?
Head Wound
Oct 7, 2003, 02:36 PM
Originally posted by FlamDrag
I'm not an expert on the subject, but my understanding is that liquid cooling still needs airflow to cool the liquid.
This I know: Heat doesn't just magically disappear.
So, logically, the liquid has to go SOMEWHERE away from the heat to cool and come back. This by no means eliminates the need for fans. You probably need even more powerful fans to push more cool air past the water pipes to cool the water before it makes its way back to the processor.
Either way, it won't be a terribly quiet machine. Imagine a small fishtank inside of your machine, pumping water. It might be somehwat similar. Flowing water can get pretty noisy.
Unfortunatly, you got it backwards. The idea behind water cooling is to move the heat from the CPU, which has a very small surface area, to a heat exchanger, which has, compared to the CPU, a gigantic surface area. This alows you to use FAR quieter and less powerful fans to get the same cooling power, and have a system that is nearly silent. If the heat exchanger is large enough, you could theoretically even eliminate fans completely, though most people don't go that far. Too costly and to much space required.
This microchannel technology is even better then current tech, though, because it pulls heat from the CPU to the water FAR more efficently. Perfect for a powerbook!
Speaking of water cooling in powerbooks, there is a VERY good reason to do it that I don't think anyone has mentioned yet. The form factor of a powerbook means that there would not be a whole lot of additional cooling power gained from water cooling, but don't over look the huge improvment in design flexability! With water cooling, you can place the CPU nearly anywhere in the case and just move the heat to the fans at the edge. Water cooling could be the key to smaller, thinner, and even lighter powerbooks.
AidenShaw
Oct 7, 2003, 03:05 PM
Originally posted by Head Wound
With water cooling, you can place the CPU nearly anywhere in the case and just move the heat to the fans at the edge.
My three year old Compaq Armada M700 uses boiling water^H^H^H^H^H liquid in heat pipes to cool the 700 MHz Mobile Pentium III. Pictures of the underside of the PowerMac G5 motherboard also show heat pipes cooling some of the controller chips.
Design flexibility is the key - M700 has the fan and heat exchanger *beside* the CPU, not on top.
Sounds like the new "microchannel" is an improvement, but not a radically new technology.
(edit: IIRC Compaq definitely was saying that the heat pipes on the Proliant servers used low pressure water)
Originally posted by AidenShaw
My three year old Compaq Armada M700 uses boiling water in heat pipes to cool the 700 MHz Mobile Pentium III. Pictures of the underside of the PowerMac G5 motherboard also show heat pipes cooling some of the controller chips.
Design flexibility is the key - M700 has the fan and heat exchanger *beside* the CPU, not on top.
Sounds like the new "microchannel" is an improvement, but not a radically new technology.
Usually they don't use water.
Originally posted by acj
Usually they don't use water.
Replace water with any non corrosive, low density, non flammable, non electronicly charged liquid ;)
MrMacMan
Oct 7, 2003, 04:16 PM
Ah finally really really quiet technology!
:p
Bring on those G6 WaterCooled xServer's w00t!
:D
yamabushi
Oct 7, 2003, 04:29 PM
Heat pipes are great for getting the heat away from where it can do the most harm. You can gain additional cooling performance by combining them with fans and/or increasing the surface area of the pipes in areas that are relatively cooler than the heat producing areas. Actually, while a pump for the fluid can increase performance, it isn't required.
I would love to see a Powerbook with heat pipes along the inside and outside of the case. I think it would look very cool - pinstripe aluminum. No power would be required for the heat pipes. The Nexus heat pipe laptop cooler helped inspire this idea.
http://www.xpcgear.com/notebookcooler.html
I hope Apple is considering a design similar to this. If Nexus can do it this cheap, cost should not be prohibitive.
AidenShaw
Oct 7, 2003, 04:29 PM
Originally posted by acj
Usually they don't use water.
http://www.tomshardware.com/howto/20030131/shuttle-07.html
"A heat pipe uses a hollow receptacle (metal pipe) to transport heat directly from one point to another.
The metal pipe is filled with fluid, 90 percent of which is distilled water; the remainder consists of special ingredients added to optimize the liquid's thermal transfer properties.
Here's how it works: the liquid is subjected to a very low pressure, reducing the evaporation point to approximately 30 degrees Celsius. When cold, the pipe contains very little water. However, when the heat pipe contacts the CPU directly on one end, the water evaporates and transports the thermal energy to the cold end of the pipe."
ibidanon
Oct 7, 2003, 04:40 PM
I was interested in tracking down the research since that would be public information, whereas no doubt Cooligy as a company is now keeping very sercretive.
I am a Stanford EE grad student and was interested to discover that this work comes out of some Stanford reasearch. Check out the following link:
http://me.stanford.edu/globals/goodson2.html
yamabushi
Oct 7, 2003, 04:41 PM
I have heard of some systems that use methanol or alcohol.
I used a machine coolant for CNC Milling several years ago that is pretty neat. You added this oil to a bunch of water (about1:9 if memory serves) and mixed it together. You wound up with this milky green liquid that got sprayed directly on the cutting tool. No, it didn't ever seperate like salad dressing. It was a water soluable oil. I never realized there was such a thing before that. Perhaps this is what they use in some heat pipes, or maybe not.
iHack
Oct 7, 2003, 05:07 PM
Originally posted by AidenShaw
http://www.reukeboom.com/art/takes/wit_tandy1989.jpg
Does it say MS OS/2 in the text? I thought OS/2 was IBM's brainchild?
Very cool, tho, at a mere $8999,-!
BTW does CA. have a governator yet?
M.
Sheebahawk
Oct 7, 2003, 05:41 PM
no governator for half a dozen more hours, I'm so torn, I hate everyone on the ballot, and arnold, well have you seen any of his speeches? hes utterly insane. It kida bugs me that this is hot topic news to people who live in the netherlands. I guess california isn't that interesting once I've been numbed to it.
and gray davis is such a pansy, I think I'll vote no on the recall though... still undecided, No on recall, yes on uhm legalization. the only candidate I thought I would vote for dropped out. this whole things pretty damn ludacris, at least it will be over soon
www.recallbush.org
Originally posted by yamabushi
I have heard of some systems that use methanol or alcohol.
Hey Great, Apple can just combine this with those rumoured methanol fuel-cell powered laptops they have been working on, just pump the methanol fuel over the red hot processor and …
yamabushi
Oct 7, 2003, 06:13 PM
Originally posted by iHack
Does it say MS OS/2 in the text? I thought OS/2 was IBM's brainchild?
Very cool, tho, at a mere $8999,-!
BTW does CA. have a governator yet?
M.
IBM and MS worked together for a while on a next generation OS project, then split over differences in opinion as to the best design. IBM wanted a clean break while MS wanted more continuity with legacy OS design. The result was OS/2 and Windows NT. Issues about ownership and marketing probably caused some confusion at the time.
robodweeb
Oct 7, 2003, 06:57 PM
Originally posted by FlamDrag
I'm not an expert on the subject, but my understanding is that liquid cooling still needs airflow to cool the liquid.
This I know: Heat doesn't just magically disappear.
So, logically, the liquid has to go SOMEWHERE away from the heat to cool and come back. This by no means eliminates the need for fans.
It doesn't necessarily require fans either. The microchannels of warm fluid could be thermally coupled to the structure of the computer/laptop. This increases the radiative surface area and could make use of ambient airflow to dissipate the excess watts. It doesn't matter where you move the heat to ... just get it away from the electronics and smear it out. The universe will continue onward towards its inevitable heat death ... speeded along by Intel Pentiums and less so by G5s ...
robodweeb
Oct 7, 2003, 07:02 PM
Originally posted by ibidanon
I am a Stanford EE grad student and was interested to discover that this work comes out of some Stanford reasearch. Check out the following link:
http://me.stanford.edu/globals/goodson2.html
Excellent link! Folks have been wondering how the fluid would be impelled through the microchannels and the electro-osmotic pump described in this link sounds like an extremely innovative solution.
tbdavis
Oct 7, 2003, 10:13 PM
Originally posted by reyesmac
I can't remember Apple ever using a brand new technology that they didnt invent for any of their products. The technology has to be around for quite a while and work for them to start using it.
According to this PDF http://web.mit.edu/hmtl/www/papers/GOODSONpres.pdf , Apple demoed a version of the ElectroOsmotic Microchannel Cooler on a dual processor system in 2002.
And for all those worried about the additional mechanical complexity of such a system. It's not mechanical. It's electrokinetic. See http://www.stanford.edu/~dlaser/silicon_eo_pumps/silicon_eo_pumps.htm for more details on what that means.
Analog Kid
Oct 8, 2003, 12:05 AM
Originally posted by robodweeb
Excellent link! Folks have been wondering how the fluid would be impelled through the microchannels and the electro-osmotic pump described in this link sounds like an extremely innovative solution.
Ah, a Caterpillar Drive!
I was wondering what those symetric doors were at the bow of the case, and then again at the aft...
Analog Kid
Oct 8, 2003, 12:08 AM
Originally posted by yamabushi
IBM and MS worked together for a while on a next generation OS project, then split over differences in opinion as to the best design. IBM wanted a clean break while MS wanted more continuity with legacy OS design. The result was OS/2 and Windows NT. Issues about ownership and marketing probably caused some confusion at the time.
Basically, IBM got stiffed twice my MS. First on DOS, then (thank you sir! may I have another?!) on OS/2.
The first time, IBM arguably didn't know what they had with the whole PC thing and just got sloppy with the contracts. The second time it should have all been crystal clear and MS played them like suckers.
ClimbingTheLog
Oct 8, 2003, 01:43 PM
Originally posted by reyesmac
I can't remember Apple ever using a brand new technology that they didnt invent for any of their products. The technology has to be around for quite a while and work for them to start using it.
802.11b? 802.11g? USB? PCI-X? Pixo OS?
AidenShaw
Oct 8, 2003, 03:48 PM
Originally posted by ClimbingTheLog
802.11b? 802.11g? USB? PCI-X? Pixo OS?
Apple was first to include 802.11b internally, but cards and access points had been available long before Apple took that step.
802.11g - first, it's an extension of an existing standard, not something completely new. Second, Linksys and others had product on the shelves before Apple's announcement.
USB - tens of millions of PCs had it standard before Apple came out with the iMac. Apple jump-started the adoption by dropping all the other ports - but they weren't first with new hardware.
PCI-X - old technology, been on servers for years and on many Intel workstation chipsets.
Pixo OS - OK, there's one.
ClimbingTheLog
Oct 8, 2003, 04:44 PM
Originally posted by AidenShaw
Apple was first to include 802.11b internally, but cards and access points had been available long before Apple took that step.
But only as enterprise-class devices. I remember buying access points for $1600 and cards for over $200. Apple grabbed the nacent bull by the horns and invented home wireless networking. It wasn't proven at all that anyone would even want this.
802.11g - first, it's an extension of an existing standard, not something completely new.
Well, it is a new standard, even though it's backwards compatible with an older one.
Second, Linksys and others had product on the shelves before Apple's announcement.
Wait, first the argument is Apple doesn't embrace new technology, then it's if a vendor beats them to market it's not new anymore?
USB - tens of millions of PCs had it standard before Apple came out with the iMac. Apple jump-started the adoption by dropping all the other ports - but they weren't first with new hardware.
Do you remember the USB adoption curve shown at MacWorld SF '99? When the iMac came out there were a total of, I think 30 USB devices on the market, from all vendors. That's about as new a technology as you can get if you're not inventing it. Apple made a bet that vendors would adopt it, shown the benefits, and they were right.
PCI-X - old technology, been on servers for years and on many Intel workstation chipsets.
Yeah, you're right, though it's nice to see it in a personal computer.
Pixo OS - OK, there's one.
The only one you concede is the one that's not a standard. A standard, by definition, requires multiple vendors to be in a market. But if someone else is in the market, you just say Apple wasn't first, even if it's a very early adopter. And you already disqualified Apple-invented technology as a contender. You might be able to setup a straw man argument that can be won, but you'll have trouble convincing most folks that Apple doesn't embrace new technology.
yamabushi
Oct 12, 2003, 08:16 AM
Originally posted by Analog Kid
Ah, a Caterpillar Drive!
I was wondering what those symetric doors were at the bow of the case, and then again at the aft...
I laughed out loud at that one.:D
Originally posted by Analog Kid
Basically, IBM got stiffed twice my MS. First on DOS, then (thank you sir! may I have another?!) on OS/2.
Yes, they should have known better.
Wash!!
Oct 12, 2003, 01:20 PM
Originally posted by yamabushi
I laughed out loud at that one.:D
Yes, they should have known better.
Those doors are too big for torpedo launchers...
vBulletin® v3.8.6, Copyright ©2000-2012, Jelsoft Enterprises Ltd.