View Full Version : we need better stuff, thermal paste doesn't cut it anymore.
dimmreaper
08-21-01, 02:27 AM
I'm wondering if anyone has any ideas on a better compound to place between the CPU and heatsink.
You can have the coolest cooling system in town, but if you can't get the heat away from the CPU it doesn't do any good. So naturally we need to adress this problem. Up until recently increasing the temperature difference between the CPU and Heatsink(ala Peltier, or just a better heatsink), and lapping, was good enough. But I think we are definately seeing the point of diminishing returns.
The only thing I've been able to dream up is possibly using JBweld (the stuff does transfer heat well I'm told) rather than thermal compound. But I havn't had the guts to try it yet.
Anyone else have some Ideas?
FerrariF50
08-21-01, 06:35 AM
why the hell do they call it lapping? your useing very fine sandpaper right?
The Overclocker
08-21-01, 08:18 AM
i am still waiting for artic diamond paste, you could solder it on, i suppose
chromebone
08-21-01, 08:47 AM
:) hers one to check out.
Prematex (brand) aviation form-a-gasket
ive used it on one heat sink seem to be workin fine.
I originally bought it to bond my volkswagon block back to gether
(they sepereate in two) and it seems to be a great product. when at a lose for products when changing processors i dug it out and used it on my heat sink . seems to work fine but not
sure if right for you.
available at most automotive parts stores around $5.00
let me know what you find out!!!!
Maybe they should just produce less power-hungry CPUs....
Originally posted by cjtune
Maybe they should just produce less power-hungry CPUs....
They need to come up with optical processors. High speeds, low heat. I remember hearing rumors a while ago about optical processors, but haven't heard anything about them in about a year or so.
FerrariF50
08-21-01, 09:50 AM
Well if they don't do anything soon they are going to have to find some other fourm of cooling because right now air cooling is just about at it's peak......
Originally posted by Wa11y
They need to come up with optical processors. High speeds, low heat. I remember hearing rumors a while ago about optical processors, but haven't heard anything about them in about a year or so.
Heh heh, I heard about organic CPUs as well. Right now I think they're having trouble just getting the tissue to form correct neural connections. God forbid if your Orga CPU catches cold during a critical app :D. I think the overclockers forums then would have a section on 'healthy living' as well.
Originally posted by FerrariF50
Well if they don't do anything soon they are going to have to find some other fourm of cooling because right now air cooling is just about at it's peak......
Stock waterblocks on each boxed CPU? All casings then might cost at least $99.95 more as they would have to come with radiator, pump, tubing and all the neccesary chems. Just add water and stir...
What? No mention of Arctic Silver? It has the highest thermal conductivity of any availble thermal grease.
JB Weld is epoxy with steel particles. Steel does not conduct heat well. Arctic Silver Epoxy is similar with silver instead of steel for better thermal conductivity. Even so, the thermal grease is a better solution unless you want the HS to stick to the object you intend to cool.
The combination of high power consumption and small die size make it difficult to wick the heat away from the cpu. Air cooling of AMD's has been pushed to the max as the average person who just goes to the store and buy's a pre built system will not tolerate the screaming fans that we will put up with. I think in the next major revision AMD makes they will have to do something to address this issue.
fuzzba11
08-21-01, 01:30 PM
My question would then be why do they make the core so small? They should extend the core over the entire chip, and maybe make it out of gold as well as the pins. We have heatsinks that have at least 10 times the 2D dimentions of the core...
Originally posted by fuzzba11
My question would then be why do they make the core so small? They should extend the core over the entire chip, and maybe make it out of gold as well as the pins. We have heatsinks that have at least 10 times the 2D dimentions of the core...
well just looking at my P4 and the core does span almost the entire chip :)
BobbyEwing
08-21-01, 03:13 PM
Even if there where a solution to making the silicone chips cooler by decreesing line thickness and thus voltage, the silicone chips will soon have reached it's physical limits. The minimum theoretic line size is about 7 micron. Anything smaller then this and there will not be enough insulation between currents to prevent crosstalk. T-birds use 18 micron technology, and Intel recently released pIIIs with 13 micron tech. Something else, be it optical or biological chips has got to be created.
Last year a group of swedish scientists managed to actually teleport one electrone across a small distance. The term teleportation may be off, but what happened was that the electrone dissapeared from location 1 and an identical electrone appeared at location 2 within a timespan that equals 4 times the speed of light. This form of technology may be one possible way to go with future computers...
I believe molecular (SP) computing is in the lead right now to replace the chip. Molecular computing in when they take polymers that have electrical conductive properties. And uses the polymer molecules at transistors and other composites.
FerrariF50
08-21-01, 05:10 PM
yeah I wish that they would make it bigger because when I put on a Hsf cooler I don't want to crack it because of it's small size.
Originally posted by fuzzba11
My question would then be why do they make the core so small? They should extend the core over the entire chip, and maybe make it out of gold as well as the pins. We have heatsinks that have at least 10 times the 2D dimentions of the core...
Originally posted by FerrariF50
yeah I wish that they would make it bigger because when I put on a Hsf cooler I don't want to crack it because of it's small size.
The solution to cracking is an integrated heat spreader like the K6-2 line had. This will be especially important when AMD goes to .13 form .18 I wonder why they stopped putting the heatspreaders on their cpu's....maybe to save a few cents in production?
FerrariF50
08-21-01, 05:57 PM
That metal cover onver top of the chip? That said Amd K6 or whatever?
Originally posted by ken257
The solution to cracking is an integrated heat spreader like the K6-2 line had. This will be especially important when AMD goes to .13 form .18 I wonder why they stopped putting the heatspreaders on their cpu's....maybe to save a few cents in production?
FerrariF50
08-21-01, 05:58 PM
Hopeing that Amd is smart they better do something fast or they might put themselfs in a corner untill they can find new stuff that makes the Cpu go faster without less heat.
Originally posted by cjtune
Stock waterblocks on each boxed CPU? All casings then might cost at least $99.95 more as they would have to come with radiator, pump, tubing and all the neccesary chems. Just add water and stir...
William
08-21-01, 05:58 PM
what happened to the Gold Leaf idea?
CrystalMethod
08-21-01, 08:40 PM
Not sure, but last I read on it was that the layer of foil would just transfer the heat to the next conductor (presumably aluminum, or copper) resulting in negligible results. I'm thinking the next step for us is total system liquid immersion.
FerrariF50
08-21-01, 09:47 PM
Well Amd Cpu cores are strong....I took the over off my Amd 486 and put a tourch to the underside of the core. And the whole chip got so hot I just cucked it out of my hand.
Originally posted by touser
well just looking at my P4 and the core does span almost the entire chip :)
I don't believe it's the core that spans the whole chip. I belive that's just the heat spreader thats attached to the core. The core is most likely much, much smaller. I could be wrong but that's what I was lead to believe.
Originally posted by fuzzba11
My question would then be why do they make the core so small? They should extend the core over the entire chip, and maybe make it out of gold as well as the pins. We have heatsinks that have at least 10 times the 2D dimentions of the core...
More yield per silicone wafer. Saves $$. HSFs ARE MEANT to have 10x the size (and the surface area) for better heat dispersal. If the cores were large enough, we wouldn't even need HSFs. Remember the old 386 and 486 CPUs? Yeah, and for the best possible contact between CPU and HSF, they might have to be built as ONE. Next time, the die, the core, and the HSF might be fabricated together and that'll be the twilight of third-party HSFs manufacturers... :( Only choice we may have then would be the type of fan/cooler to go with the integrated CPU/HSF.
fuzzba11
08-21-01, 10:31 PM
Originally posted by cjtune
More yield per silicone wafer. Saves $$. HSFs ARE MEANT to have 10x the size (and the surface area) for better heat dispersal. If the cores were large enough, we wouldn't even need HSFs. Remember the old 386 and 486 CPUs? Yeah, and for the best possible contact between CPU and HSF, they might have to be built as ONE. Next time, the die, the core, and the HSF might be fabricated together and that'll be the twilight of third-party HSFs manufacturers... :( Only choice we may have then would be the type of fan/cooler to go with the integrated CPU/HSF.
Of course they're supposed to be 10X the size in the way you're thinking, but I'm saying that why does the BASE of the heatsink need to be that big? Shouldn't it only be as big as the top of the core? (like the new Zalman heatsinks). If it's possible, you might as well have a big ol core, then there's more contact area which leads to better heat transfer and therefor dispersal.
William
08-21-01, 10:51 PM
Originally posted by CrystalMethod
Not sure, but last I read on it was that the layer of foil would just transfer the heat to the next conductor (presumably aluminum, or copper) resulting in negligible results. I'm thinking the next step for us is total system liquid immersion.
that makes no sense at all. The point would be a more efficient heat transfer, which it would do, probably better than AS2.
The next step is to design cooler CPUs. A company can do that, often they alternate generations. The P3 is designed to run cool while an Athlon is designed to run hot.
Crash893
08-21-01, 11:11 PM
The core is very small much smaller than the chip its self
that’s what makes it go faster (less distance between transistors means less time wasted)
I think the next advanced in cooling will be
perhaps make the chip out something other than ceramic maybe like metal or have some sort of integrated micro tube in the actual cpu but only the chip manufactures can do that and I think it will take them just a little while longer
partial or total liquid emersion
it seems the best way to lose cpu water block bottle neck is to actually lose the part that’s causing it.
My own personal idea
find some way to cool the bottom of the cpu as well
I envision some sort of socket extender that lets you raise the cpu up a little and stick some sort of Pelzer under there basically make a cpu sandwich
k
dimmreaper
08-22-01, 01:18 AM
Originally posted by ken257
The solution to cracking is an integrated heat spreader like the K6-2 line had. This will be especially important when AMD goes to .13 form .18 I wonder why they stopped putting the heatspreaders on their cpu's....maybe to save a few cents in production? AMD stopped usind the "heat spreader" plate because the thermal epoxy between the CPU die and the spreader plate just added more thermal resistance. The "heat spreader" was used on the K6-IIs to allow the use of flip chip technology without redesigning the the heatsink clip. But once Intel started going to flip chip technology, heatsinks with the proper clamp became widely available, so AMD followed the leader and went topless.
I'm wondering how the P4/Tulatin cores are bonded to there "heat spreader" though. It would be interesting to find that out.
dimmreaper
08-22-01, 01:20 AM
Originally posted by Colin
JB Weld is epoxy with steel particles. Steel does not conduct heat well. Arctic Silver Epoxy is similar with silver instead of steel for better thermal conductivity. Even so, the thermal grease is a better solution unless you want the HS to stick to the object you intend to cool. So that's what the put in JB weld. Thanx for the input Colin!
Originally posted by fuzzba11
Of course they're supposed to be 10X the size in the way you're thinking, but I'm saying that why does the BASE of the heatsink need to be that big? Shouldn't it only be as big as the top of the core? (like the new Zalman heatsinks). If it's possible, you might as well have a big ol core, then there's more contact area which leads to better heat transfer and therefor dispersal.
The CPU can be thought of as a heat source that conducts heat radially -so ideally we would like to cover as much 'angle' about the CPU hence the large bases for most HSFs nowadays. Now imagine the (possible) future models of Zalman, an 'iron maiden' kinda HSF that looks like a spiked ball -the core is integrated in the centre of the entire metallic mass and a long plastic column extends down to fit in a Slot to transfer data. Old heatsinks of this kind can be fitted at the end of a stick and used for self-defense... :D
BTW, gold, copper, or silver foils at the base is next to useless because the heat needs to travel sideways as well as upwards and because of the thickness of the foil, the heat doesn't really get spread out. A gold PLATE would do better.
dimmreaper
08-22-01, 02:43 AM
Originally posted by cjtune
A gold PLATE would do better. But a Silver one would be better yet . . . .
I liked the sandwich idea. But instead of loosing precious pin space, the MB should have a hole in center of the socket, and the CPU core's backside should be exposed. Thisways there could be double the surface and two HSFs on both sides. Or mayby the slot design was good for this?
Also if water cooling got somehow standardised, I wouldnt dismiss the idea of boxed waterblocks. This would just need the cases to come with standardised water cooling equipment. Imagine the water flowing right ontop, just 1mm away from the core.
I think the heat spreader can do good, they just need someting else than epoxy to attach it. They could have the core more flatter and the spreader should have a "hole" size of the core on it with some AS2 to gather the heat from the top and sides of the core. Then epoxy only on the spreader corners to keep it still.
Originally posted by dimmreaper
But a Silver one would be better yet . . . .
Agreed, but the key issue here is the thickness of the heat spreader layer. Foils just aren't cut out for it.
Kingslayer
08-22-01, 09:57 AM
Well if ASII isn't cutting it anymore, then you guys need to scope this out. Looks like it out performs AS!
http://www.overclockersclub.com/nanotherm.shtml
Finally, an alternative to AS2. Maybe that'll cause some disturbance in the 'force'.....or at least the prices.
William
08-22-01, 02:25 PM
Originally posted by cjtune
Agreed, but the key issue here is the thickness of the heat spreader layer. Foils just aren't cut out for it.
the idea of foil is not to spread the heat it is to transfer the heat to the heatsink as a good replacement. Because it is thin and readily sticks, it should make for a good substituted. I wish someone with a spare flip chip could do this(or send me a duron to try it on :D)
Originally posted by cjtune
BTW, gold, copper, or silver foils at the base is next to useless because the heat needs to travel sideways as well as upwards and because of the thickness of the foil, the heat doesn't really get spread out. A gold PLATE would do better.
The heat wouldn't get spread out as good as with a gold plate, but you also notice that most HS (at least the ones I've seen...) have a thickness between the base and the bottom of the pins. The gold foil would transfer it to the HS (better than AS), and then the HS itself would spread out the heat. AS dosen't spread out heat, the thickness of the base is supposed to do that. The gold only needs to get the heat to the HS, so it can radiate. The "thick" base of the HS allows all the pins to get heat.
Just my opinion....
Too bad they can't increase the core size ($$ restrictions someone said)! That would decrease the heat quite a bit!
JigPu
Originally posted by FerrariF50
Well if they don't do anything soon they are going to have to find some other fourm of cooling because right now air cooling is just about at it's peak......
But water & peltiers still do the job :) You aint seen nothing yet...
Time to forget those screaming CPU-fans??
FerrariF50
08-22-01, 08:27 PM
Heheh hey Fjeld well yeah but the Amd 1.4 Ghz uses like a Pelt at 175 Watts or so.....Do they make them any higher??
Originally posted by Fjeld
But water & peltiers still do the job :) You aint seen nothing yet...
Time to forget those screaming CPU-fans??
Originally posted by cjtune
Heh heh, I heard about organic CPUs as well. Right now I think they're having trouble just getting the tissue to form correct neural connections. God forbid if your Orga CPU catches cold during a critical app :D. I think the overclockers forums then would have a section on 'healthy living' as well.
Didn't they do something like that on Star Trek: Voyager (not that I ever watched it!), where the ships computer runs of "bio-memetic gel packs" that once got sick, so it screwed everything up? But then, keeping your processor cold would be bad, because it could kill your processor! Literally, kill!
SteenkyBastage
08-22-01, 08:38 PM
i'm not sure if i'm following the gold foil/gold plate idea right or not, but if i am...i think that it would still require the use of artic silver. the purpose of the foil or plate would be completely different then the purpose of the AS2.
the purpose of the AS2 is to fill in the microscopic gaps where the air would otherwise go and insulate rather than conduct heat. the plate idea would help to spread in exactly the same way that the thicker base of the HSF already does. so if you did add a gold or silver plate, you then need to add two layers of AS2 in order to fill the microscopic gaps on both the hsf/plate side and the core/plate side.
once again, i may be getting the wrong idea from what people have posted about the plates, so plz dont get upset if i'm way off base.
Originally posted by JigPu
The heat wouldn't get spread out as good as with a gold plate, but you also notice that most HS (at least the ones I've seen...) have a thickness between the base and the bottom of the pins. The gold foil would transfer it to the HS (better than AS), and then the HS itself would spread out the heat. AS dosen't spread out heat, the thickness of the base is supposed to do that. The gold only needs to get the heat to the HS, so it can radiate. The "thick" base of the HS allows all the pins to get heat.
Just my opinion....
Too bad they can't increase the core size ($$ restrictions someone said)! That would decrease the heat quite a bit!
JigPu
Hey, aren't you the fella that start the super-lengthy thread "Copper and Aluminium" some time back ago? I spent a lot of time there y'know....
Recap:
A thick base is undesireable because it is actually a thermal resistance to the vertical flow of heat. But then a thin one would constrict heat flow sideways so that the outer fins (that are exposed to the most airflow) work at their best. Material wise, a metal with as high a conductivity as possible should be used for the heatsink -but if costs are prohibitive, maybe then just make the base out of it. Compromises are needed during design.
Thermal interface materials ease the passage of heat between contact surfaces by being both as conductive as possible and as 'runny' as possible (being able to fill in micro gaps). Thermal tape by itself actually has a higher thermal conductivity than normal zinc oxide paste but since it doesn't fill in gaps too well, it's a pushover for most O/Cing purposes.
Sorry William, I didn't catch on to the gold foil thingy initially. Was it that the a thin goil foil was supposed to be sandwiched under high pressure between the CPU core and the HSF? If so, the cores might have to be constructed stronger in order to stand up to the pressures needed to squeeze gold molecules in between micro gaps. Extra ceramic substrate thickness might then be needed and it'll ****** heat dissipation -a plus and a minus. Maybe the best soln might still have to be a kinda paste...or solder, anyone?
cool_hand_joe13
08-29-01, 09:27 PM
:D I thought about using anti-sieze compound,beacuse it has copper and grease mixed as one,but scared to try it.
Warlord2
08-29-01, 10:31 PM
I donno if anyone here is into guns but they make this epoxy that is used for stock bedding
its supposed to have alot of metal in it and is stronger thin JBweld
its called Micro-Bed
AZZKICKER
08-29-01, 10:47 PM
gold isnt a very good conducter of heat
electic current - yes one of the best, but not heat
i been thinking and i think i got a idea on a heatsink that will work
never seen anyone try it
but i aint telling...hopeing a heatsink company will work with me and mybe sell the right to them
as for the core on a amd........i think a bigger core would help
look at the Pali. amds the core is bigger
they run cooler.......so......it they made the core half the size of the chip........more heat could tranfur.....but that costs $$$.....they wont do that
so why not a plate made of high thermal compound of multi metals
thats in my mind too and wont tell how to do it
but i think it would work SWEET
need something to suck heat off the core as fast as it can....spread it over the surface of the plate and transfur it to the heatsink over a wider area of the heatsink
this takes hotspots out of the heatsink and will cool more eficaint
yes thermal compund worries me.........got that under construction on the rough draft
wish some big azz heatsink manufactur would PAY me for a idea......i think it would work better than any air cooling out
also fans......need to use bigger / quieter fans...i got a 7 inch , 235 CFM fan i wish i could figure how to mount on this heatsink in my head
and make the air flow MAX with proper turbulance
anyway im rambling on..........
speedy4500
08-30-01, 04:00 PM
The best option is to completely eliminate the compound (and any other obstacle) entirely, using direct core cooling. I am not sure whether water would cause problems if it were squirted directly on the core, but if it wouldn't cause problems, you could create a water proof seal around the core on the ceramic so that you have water to core contact. The more "layers" of heat transferring material you have, the less efficiently you will transfer heat (this doesn't mean that you won't reach lower temperatures, just that it will be less efficient). Even better would be to use a phase change cooling system. directly on the core.
AZZKICKER
08-31-01, 08:33 PM
bump
Crash893
09-02-01, 11:18 PM
i know this sound werid but do you think there is a way to dirll a hole directly threw the cpu ( not in the core but in the cermaic next to it i know this would be very very risky but if you could put a couple holes in it you could make it inline with the watercooling tubeing.
other than that the only thing i can think of is try to find ways to space the cpu from the mother board and try to cool it on both sides.
funnyperson1
09-02-01, 11:44 PM
i have this stuff i just got called solder-it and what it is is a liquid solder that you heat up (a cigarette lighter is good enough and it will solder stuff together..it is silver based...the heat of the cpu should be enough to solder..my only worry is that it said dont use it on aluminum.....if i get enough courage(or 150 dollars) i might just try it on my gf2pro...thatthing runs hot as heck
Crash893
09-05-01, 12:34 AM
what about some sort of diamond chip or soemthing like that
line the bottom of your heat sink with a diamond dust
just a thought
how far can you sand down the core before it breaks?
*walks up behind 7 of 9 as she is working on a conduit and takes a molecular bonder from her repair kit after recieving a questionable look*
Hehe
Here ya go guys, NOTHING will get your hs and cpu thermaly bonded better... However its perminent :)
One at a time.. one at a time..
*starts taking in flock of cpus with hs* :D
-Trek
gold? gold has a ****tier thermal properties than lead? why would you use it? use some space shuttle tiles too.
~e
Well Gold would be out because it has less thermal conuctivity than silver.
Solid silver would be OK but would break the bank.
How about some diamond powder in a bonding agent?
How about some of the newer graphite compounds for heat sinks?
drunkmonkey
11-08-02, 06:42 PM
We could make processor like they did when computers first came out. Every transistor a full size one:D It would be F***ing huge but think of the cooling options and can you say 10-20Vcore? lol
(This post is sarcasm, I know that would never work)
Ugmore Baggage
11-08-02, 09:57 PM
barring a heatspreader to increase the contact patch, the only way to get more transfer is either more/cooler air and efficient use of mass/area (assuming air).
It water is an option, CPU dies could come with a built-in waterblock, with fluid making direct CPU contact. No need for clips then, except to ensure you don't accidentally rip the CPU out of the zif.
An aftermarket version shouldn't be too difficult to design, I'd make the hole only in the center of the unit, with concentric ribs suitable for a gasket material, and wire clips to hold it on firmly. Dunno if you need a special type of water; I mean, can it be more conductive than copper? Put the pump on the outflow line to ensure suction.
Or... my first thought was billions of nanohsf units... trillions for extreme OC heh
Chips with biggercores, more surface area, better insulated chips, so that some kind of water block could be placed over it allowing the water to actually make contact with the core itself ?
Gold? Why gold? Only feature it got is like heavy and it doesn't oxidize at fast as the others.
Isn't silver and copper more efficient ?
If we can't come up with colder cpu's, we have to focus on meaner materials in the heatsinks.
Copper is good... silver is better... pyrolytic graphite is 4-5 times better than silver.. diamond is like twice as good as pyrolytical graphite.
I for one have a slab of pyrolytic graphite at home...
I haven't however machined it into anything useful yet....
Rumor is that PG doesn't conduct heat in all three directions though. But PG is cheaper than diamond :)
/Paxmax
Volenti
11-09-02, 08:37 AM
Originally posted by speedy4500
The best option is to completely eliminate the compound (and any other obstacle) entirely, using direct core cooling. I am not sure whether water would cause problems if it were squirted directly on the core, but if it wouldn't cause problems, you could create a water proof seal around the core on the ceramic so that you have water to core contact. The more "layers" of heat transferring material you have, the less efficiently you will transfer heat (this doesn't mean that you won't reach lower temperatures, just that it will be less efficient). Even better would be to use a phase change cooling system. directly on the core.
Discussion on this (and an example :D) can be found here (http://forum.oc-forums.com/vb/showthread.php?s=&threadid=132447&highlight=direct)
Toysrme
11-09-02, 08:40 AM
JB weld wouldn't work very well. I don't advise using it. I mean look at it... It's nothing but a two part 15 or 30 min EPOXY with some METAL FLOATING IN IT.
That'd be TONS worse than just plain old white compound.
-Toysrme
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.