direct die cool a heatspreader

[MadHacker]

another option is to remove the heat spreader.
I have removed them from my P4 2.6Ghz & P4 2.53 Ghz cpu.
get a much better overclock with it removed.

[Butcher]

Removing the heatspreader and using a proper block is likely to give better results than anything you do with the heatspreader on.
People seem to think AS is a good conductor of heat. News flash - it sucks. It's better than pretty much any other TIM, but TIMs as a group are about a factor of 100 worse than copper for thermal conductivity.

[j813]

So what is an effecient TIM replacement, is there any?

[jaydee]

Originally Posted by j813

So what is an effecient TIM replacement, is there any?

Not really.

[Jabo]

Arctic Ceramique applied extremely thin is good Actually, the thinner the better bt then you goota have perfectly flat mating surfaces to start with

[Butcher]

Ceramique sucks more than AS5...
From articsilver.com I got the following numbers for thermal conductivity:

Ceramique:   5.08W/m-K
AS5:         8.89W/m-K
Copper:    390   W/m-K

There's really no substitute for proper metal to cpu contact.

[snowwie]

thermal conductivity, though important, plays only a part of the effectiveness of a TIM

ceramique and AS5 are about the best you can get, but other greases are almost as effective.

[Jabo]

Well, there's also sth called spreadability and filling ratio (how fine the thing is) and AC is supposedly better.
AS5 may be better with physical properties (thermal conductivity) but application is a different ball game alltogether.

If it was better to ommit TIM in your cooling assy then nobody would be using it by now!

Direct metal-to-cilicone contact without TIM is laughable in terms of thermal transfer compared to the same thing used with TIM.

You may want to use mercury but I doubt you'd survive t tell the story

[gruntledweasel]

Originally Posted by Jabo

Well, there's also sth called spreadability and filling ratio (how fine the thing is) and AC is supposedly better.
AS5 may be better with physical properties (thermal conductivity) but application is a different ball game alltogether.

If it was better to ommit TIM in your cooling assy then nobody would be using it by now!

Direct metal-to-cilicone contact without TIM is laughable in terms of thermal transfer compared to the same thing used with TIM.

You may want to use mercury but I doubt you'd survive t tell the story

Not to put words in another's mouth, but I suspect he (Butcher) was referring to having a TIM joint versus eliminating said TIM joint (context being the earlier discussion of removal of the heatspreader).

[Jabo]

If so sorry for misunderstanding

[Butcher]

Originally Posted by Jabo

Well, there's also sth called spreadability and filling ratio (how fine the thing is) and AC is supposedly better.
AS5 may be better with physical properties (thermal conductivity) but application is a different ball game alltogether.

If it was better to ommit TIM in your cooling assy then nobody would be using it by now!

Direct metal-to-cilicone contact without TIM is laughable in terms of thermal transfer compared to the same thing used with TIM.

You may want to use mercury but I doubt you'd survive t tell the story

You're not comparing TIM vs direct contact with cpu cores though. You're comparing TIM vs air. And air is about as bad as it gets for thermal conductivty outside of say a vacuum.

If you could somehow acheive direct metal to silicon contact over the core the thermal transfer rates would be very very high compared to a normal TIM joint. Getting surfaces to match up on the microscopic level like that is nearly impossible though, and certainly far to expensive to seriously consider.

AS5 and ceramique are the best TIMs, and I'm certainly not advocating not using them. But to suggest putting a layer of AS5/ceramique in as a substitute for a real metal bond is flawed.

[snowwie]

Originally Posted by Butcher

But to suggest putting a layer of AS5/ceramique in as a substitute for a real metal bond is flawed.

how would you arrive at this "real metal bond"?

i don't think i understand

[|kbn|]

What about soldering it on? Yes youd need a metal melts below 150c, well 130c for safty of the chip. Should work better than current TIMs? probebly not by much though. A very small amount of mercury would be interesting as a test, but it could never be used on a more than tempory scale beucase of health issues - unless maybe it was cooled way below freezing point using cascade phasechange, which would then have to be on 24/7 etc.. does (normalish/diy) phasechange even go that low?
There are ways to get things VERY flat, perfectly flat is sorta impossible though, I think, because of the structures of atoms... Lapping helps (which should be done anyway or whats the point in this thread?) but overall its best to just have lapped + lots of pressure there to sqaush out all the extra paste, giving a much better tim layer..
It would seem possible that the core can be plated safely with silver or copper (maybe using electroless plating?) and then joined using different method to another metal...

[nexxo]

Either way this design has nothing to do with performance nore is of any use in reality being it hasn't even been proved to work yet. Just a bunch of marketing BS so far. Also that design in the IHS above would not work for obvious reasons. It would still need a Air cooler on top to make up for the lack of water coverage. Seems to me these jokers don't have any idea what they are doing and trying to make it look like they do for marketing fluff. Looks are not always as they seem though.

Oh ye of little faith... With CPUs becoming more powerful and, well, hotter, there is serious money to be made in the area of effective cooling solutions. Perhaps this idea is a bit rough and ready (got to secure the patent fast, after all) but it may well work. I've seen crazier ideas work, and work well.

Bigger and more is not always better --some small diameter/slow pump setups shift some serious heat just by the fact that they are very well designed. Look at the engine cooling in modern compact cars --you won't believe how efficient they can be.

[jaydee]

Originally Posted by nexxo

Oh ye of little faith... With CPUs becoming more powerful and, well, hotter, there is serious money to be made in the area of effective cooling solutions. Perhaps this idea is a bit rough and ready (got to secure the patent fast, after all) but it may well work. I've seen crazier ideas work, and work well.

Bigger and more is not always better --some small diameter/slow pump setups shift some serious heat just by the fact that they are very well designed. Look at the engine cooling in modern compact cars --you won't believe how efficient they can be.

CPU will not continue to become "hotter". You logic fails there. OEM's will not tolerate it. Simple as that.

[Jabo]

With current technology CPUs ARE going hotter and use more energy with eachnew release, facts of life here.

There is a trend to optimize thermal design of CPUs (mobile technology) but this is just tinkering not changing simple physics (more transistors on smaller area = more thermal output).

Check out Israeli Intel R&D Dept's results with optoelectronics, very ecouraging I must say

[jaydee]

Originally Posted by Jabo

With current technology CPUs ARE going hotter and use more energy with eachnew release, facts of life here.

There is a trend to optimize thermal design of CPUs (mobile technology) but this is just tinkering not changing simple physics (more transistors on smaller area = more thermal output).

Check out Israeli Intel R&D Dept's results with optoelectronics, very ecouraging I must say

I am not talking about tomorrows tech like next year, I am talking about the 5+ years from now. CPU will not get hot enough to warrant anything resembling one of our current water cooling setups. The CPU's will have to get over 200watts before air cooling isn't a viable option anymore. That isn't going to happen. Someone said I didn't have FAITH which is JUST THE OPPOSITE. I have faith in tomorrows tech to make cooler faster running CPU's. Water cooling just isn't needed now nor will it ever be. I had this very conversation 4 years ago. They said by now water cooling would be main stream. Is it? No.

[MadHacker]

Originally Posted by jaydee116

...I had this very conversation 4 years ago. They said by now water cooling would be main stream. Is it? No.

but I'm a shure to guess that systems that are watercooled has increased by a factor of 10... mebe more? anybody have an idea as to what that number would be...

and what ever the number is... it will keep going up...

I have been bulding machines for 7 years now... and only now have one that is water cooled... and love it... except of course for the leaks because i didn't tighten the hose clamps enough...

[|kbn|]

I sorta agree. What type of USERs will need watercooled pc's? Only where low niose is imporant, and then watercooling doesnt work that much better, most people still have the rad in the same room/CASE, makes w/c a bit pointless may aswell just use a bigger air heatsink and bigger fan (which might not exist yet, but could be DIY'ed just ike much of the w/b have been).

One problem is its not just cpu heat now, everything else is starting to make just as much heat as old cpu's were (amd k6-2 500 for example puts out about the same as a nforce2 nb, then theres also the voltage regulators - which on nf7-s put out quite a bit). Graphics cards put out a lot of heat, current ones just below 75w the next gen ones rumoured to be upto 2x that (nv40). It would have helped if pci-express cards were flipped theright way up, to let the hot air rise, instead of getting trapped under the cards...

There are plenty of servers that are watercooled, one of the companys I recently applyed for a job with had an opertron cluster with about 100 really niosy big fans - they are building a soundproofed room for them so it doesnt annoy the workers, watercooling would be perfet here.

Do you think air cooling would be sufficent with mobos comsuming 100w + in total, cpus at 200w, graphics probably upto 170w, + the rest of the heat a pc makes? Would need lots of case fans ( = niose)


Madhacker is right, now there are lots of companys making watercooling stuff, esp the cheap crappy kits, marketing hype will make them sell, more people will be watercooling there systems. 4 years ago companys wernt making watercooled graphics card kits (gainward).

My point is, noone really needs it, but its better (mostly) and costs more, people that dont know better will buy it, becuase they think its good.

[MadHacker]

even the crappy kits that you can buy are better the the stock heatsink option.

I myself am ammazed what a difrence watercooling makes in temperature compared to air,

in my system all i have is a BIX rev2 and a Bixmicro2. and my CPU never goes above 33C with full load (onboard temp probe for P4)
my other system that i will be watercooling next runs at 60c steady.
and I have a lot of fans on it... spec here and with at least half the noise volume.

[Jabo]

Speaking of noise... I got my hands on second Hand Volcano7+. Never run anything else than Zalman contraptions on air before, when I hooked it up and gave it power I thought that sth went badly wrong (was making such a racket)
I cannot imagine how people can work/play having such a thing inside their cases???

I am still of opinion that watercooling is going to become mainstream quite soon.
4 years ago only a handful of enthusists were water cooling their machines.
Now everyboy does it (enthusiast). Water cooling technology is not yet at corporate level and not being endorsed by any OEMs yet has no chance of entering mainstream computing for now....
It just needs a bit more work on it and time to mature, thats all...

[Butcher]

Originally Posted by nexxo

Bigger and more is not always better --some small diameter/slow pump setups shift some serious heat just by the fact that they are very well designed. Look at the engine cooling in modern compact cars --you won't believe how efficient they can be.

It's much easier to get high efficiency in a car though - you want the engine to run at around 95-100C a massive difference over ambient, with computers you want to run as close to ambient as you can which is much difficult.

kbn, I think you'd have difficulty soldering to a cpu - silicon doesn't bond to solder well and there would likely be issues with shorting the chip.

[MadHacker]

Originally Posted by Jabo

Speaking of noise... I got my hands on second Hand Volcano7+. Never run anything else than Zalman contraptions on air before, when I hooked it up and gave it power I thought that sth went badly wrong (was making such a racket)
I cannot imagine how people can work/play having such a thing inside their cases???

I use to run one of them Vantec Tornado for a fan on my CPU...
first day i hooked it up and turned it on.... I was stunned as to how loud it was...
got right on the net and odrerd a baybuss... 3 days later i was able to leave my machine on...
My wife made shure it was off when i wan't home...

[jaydee]

Originally Posted by MadHacker

but I'm a shure to guess that systems that are watercooled has increased by a factor of 10... mebe more? anybody have an idea as to what that number would be...

and what ever the number is... it will keep going up...

I have been bulding machines for 7 years now... and only now have one that is water cooled... and love it... except of course for the leaks because i didn't tighten the hose clamps enough...

Of course they have. Computer users in general have increased 10 fold aswell. Doesn't mean that the percentage of WC users against non have gone up though. In fact the majority of the sites I browse are not to into water cooling nore have even heard of it. Most people are not hardware computer enthusiests and have no desire to be. Most people just want a Dell or equel that works and that is why WC will not go mainstream.

[LPorc]

Originally Posted by jaydee116

Of course they have. Computer users in general have increased 10 fold aswell. Doesn't mean that the percentage of WC users against non have gone up though. In fact the majority of the sites I browse are not to into water cooling nore have even heard of it. Most people are not hardware computer enthusiests and have no desire to be. Most people just want a Dell or equel that works and that is why WC will not go mainstream.

Water cooling is entering the mainstream. I'm seeing local builders advertising water cooled systems.