Increasing pressure on waterblock to lower temps

[FishaOfMen]

Yeah, so the springs that came with my Danger Den Maze 2 stink. The only time I get enough pressure on the block is when I compress the springs all the way down, effectively making them a set of spacers. I'm looking at getting a new set of springs that will exert more pressure on the block and thus, lower my temps 2-4C.

I have a couple questions regarding this:

1. What is the maximum sustainable pressure that AMD has rated their cores at? I think it's 24-30 lbs.

2. How much pressure can these cores really stand up under?

3. How much pressure can a video or northbridge chipset handle?

I look forward to experimenting with this and posting results.

[Boli]

I allways thought it is bet NOT to put them under preasure... about 1-2 bar I allways thought was best

[pHaestus]

To a point more mounting pressure should lower temps. As far as a max pressure for the core goes, I can't imagine one crushing it as long as the block is flat on the core before pressure is applied. I CAN imagine that the motherboard traces might start having problems if you put a ton of pressure on them. If you are going to really crank the block down, then I would make some sort of spreader for the backside of the motherboard to distribute the pressure away from a small area near the socket holes.

[Cathar]

AMD state 15-25lbs of static mounting pressure.

They also state a maximum of 100lbs of dynamic pressure, being shocks incurred by bumping the case, such as driving to a LAN or something.

If you never move your case, that 100lb rated maximum is probably close to where you can take the pressure to. However, I'm no materials engineer so I could also be totally wrong on that, but I do know people who have cranked down the White Water springs all the way on AMD T'Bred CPU's, which equates to around 70lbs of pressure with no ill effect.

Then I've heard of people who drop the springs altogether and tighten until the motherboard flexes slightly. I've even done this myself, and this probably equates to around 100lbs of pressure.

As long as the pressure applied is flat and even, you can put fairly dramatic amounts of pressure on the core without apparant harm.

[pakman]

Originally posted by Cathar
Then I've heard of people who drop the springs altogether and tighten until the motherboard flexes slightly. I've even done this myself, and this probably equates to around 100lbs of pressure.

Now the question is, did you see any improvement temps when using that much mounting pressure?

[jaydee]

Originally posted by Cathar


Then I've heard of people who drop the springs altogether and tighten until the motherboard flexes slightly. I've even done this myself, and this probably equates to around 100lbs of pressure.

Thats the ONLY way I ever mount my blocks. I never use springs. Yes the temps will drop some. The more pressure the easier the heat transfers as there is less resistance. I tighten mine down to were I can pick the board up by the block and it will not move.

There has been a bunch of crap spammed around back when the first socket A CPU's started to roll out as the heat sinks were not designed well and newbs (I was a newb then to) kept cracking cores because of it. Soon after the newbs (including myself) came out of the wood work complaining about it and there lies the over rated problem of cracked dies. I only ever cracked one die and it was with a cheap Orb (not thermaltakes) that didn't touch the 4 pads on the CPU and had a miserble clip. I quickly destroyed a CPU with that POS. Now it should be pretty difficult to crack a die as the die's themselfs are stronger and the HS's have much better clips and designs. You still need to be carefull but common since "should" prevail. If your stupid enough to use one hand ontop of the block/HS and apply all your might and then latch the clip/tighten the screws you deserve a cracked die! people are just to worried about it, but that is easy for me to say....

[#Rotor]

the only time that much pressure is required.... is when you have that pink goo (that comes on std HSF's) between your block and the CPU. the core is hard as glass, the copper... well not quite as hard, but hard enough not to deform, under any reasonable pressure, 2 non budging flat hard surfaces getting squeezed together, I know it will not make the slightest difference whether you had 3Lb or 30Lb on that block. the TIM has been formed the moment all the access thermal paste has been squeezed out, and that's it, nothing more to gain by more pressure...other than maybe a crack or two...

[pakman]

Interesting...

For those who maybe familiar with the springs that come with Dangerden kits (I have a maze3 w/ the steel bolts), would you recommend skipping the springs alltogether (they appear to compress easily). I currently use the springs and they are compressed 100% and then more... With the springs fully compressed, the block doesn't budge when I try to lift it or move it around.

[Cathar]

Originally posted by pakman
Now the question is, did you see any improvement temps when using that much mounting pressure?

It's hard to say because it's quite difficult to get a correctly even pressure across the die which is so important for a good thermal junction.

Over standard AMD mounting pressure I suppose I could say that I might pick up close to 1C by getting rid of the springs, but I find that I need to remount a few times to get it right.

[pakman]

Originally posted by Cathar
It's hard to say because it's quite difficult to get a correctly even pressure across the die which is so important for a good thermal junction.

Over standard AMD mounting pressure I suppose I could say that I might pick up close to 1C by getting rid of the springs, but I find that I need to remount a few times to get it right.

ah, the things we do to get that 1c drop in temps...

[jaydee]

Originally posted by Cathar
but I find that I need to remount a few times to get it right.

Yes, agreed.

Don't agree abit with #Rotor on this one though. it is the same logic used in TEC cooling. More pressure equels a better running TEC!

[Cathar]

Originally posted by #Rotor
the only time that much pressure is required.... is when you have that pink goo (that comes on std HSF's) between your block and the CPU. the core is hard as glass, the copper... well not quite as hard, but hard enough not to deform, under any reasonable pressure, 2 non budging flat hard surfaces getting squeezed together, I know it will not make the slightest difference whether you had 3Lb or 30Lb on that block. the TIM has been formed the moment all the access thermal paste has been squeezed out, and that's it, nothing more to gain by more pressure...other than maybe a crack or two...

Suggest you spend a little time on Google looking for the effects of pressure on thermal interface junction efficiency.

There is a wealth of well conducted research that supports the opposite to what you are saying, and supports what Jaydee and myself have witnessed.

[FishaOfMen]

Wow! Awesome responses guys! I was looking at buying some new springs, but I'll have to try the no spring method instead. I just hope I can keep the waterblock still so that my perfect TIM application won't get messed up .

[#Rotor]

disagreement is good

but now, could it not be concluded, that by running the liquid side of this thermal interface in a high pressure state... I.E pressurizing the water-loop. It would have the effect of squeezing the water molecules closer to the copper, right? water does not compress much, or at all... so would it work on the water side????

I'm looking at this on the molecular level. you can only get the two surfaces to touch that much, if nothing is bending, and nothing is deforming, then you will gain no more conduction than what you had at the point of optimal touch. I do not mean that you only need a feather, for down-force.... heat transfer is not dependant on pressure, it is a factor of surface and Delta T and material character.

the reason pressure seems to have an advantageous effect on heat transfer, comes from that goo they put on the heatsinks. it needs to get squeezed out, from between the core and the sink, now obviously the more "gooey" the goo, the more pressure you would need to get rid of it. Proper thermal paste, (white, stuff) is a paste, it is not elastic and not gooey, thus it will squeeze into perfect conformity and it will allow for crests in the two surfaces to actually make contact, for this you do not need 30Lb on the block with a minute little core.

A tec appears to need so much more, because of its surface area, being so much larger, pressure is given as down-force per surface area integer, correct? thus, to acquire the same amount of pressure, over a larger area, you would need more down-force, if I remember correctly ( been a long time )

as for googling on it, I prefer to reason it out with people that actually know what they are talking about...
Google is a last minute, emergancy solution to a problem that is bigger and stronger than me.

[bigben2k]

Ok, I gotta chip in, because there's a need to clarify a few things.

First, glass actually increases in structural strength, under compression, until it reaches a breaking point. Glass however makes a very poor structural material, when it's in tension, and will shear off relatively easily.

Copper acts like a normal material/element. Googling will reveal the force, in psi, to deform it, and it's surprisingly low: copper is a relatively "soft" metal, unlike steel.

The biggest problem with increasing pressure on the CPU die, is that we don't know what the actual surface geometry is, and we only have a poor idea about the direction of the forces applied.

If ya'll just think about it for a second... if you tighten up the 4 surrounding bolts, and given that the block is essentially supported by a center point (the pads are just there for stability), if the baseplate deflects in any way (and it does), what you have is a concave baseplate over a CPU die, and the mounting pressure points are now on the corners of the die.

Now if we're lucky, the baseplate was convex a little bit to start off with, and clamping it down forms a near perfect flat contact, under the clamping pressure.

So in the first scenario, you actually made the TIM joint worse. In the second, you made it perfect. Now ya'll apply Murphy's law here for a second, and realize that the only way to get it right, is by trial and error.

So we remove the springs? The problem with that, is that it becomes infinitely difficult to gauge the applied pressure: a fraction of a turn will make a significant difference.

In designing Radius, I opted to go with springs. Unfortunately, I lacked a little foresight when I ordered them. I've got a set of 4 * 6 lb springs. In retrospect, I would have ordered 9 lb springs, and really, after going over this thread, I would probably have ordered something even stronger.

Note that compressing the springs all the way down, puts them past their limit, and actually damages them, to where they will no longer respond as per specs.

[Boli]

So in conclusion:

Don't tighten the bolts too much... you may crack the core.
If you don't tighten them enough you won't get the best contact and/or your block may fall off.
Everychip/waterblock is different
Use trial and error to get the best fit for your block/core.

And I right?

I put just over 1 atomosperic preasre on mine i.e. it is as stable on its side as it would be upright. I don't think I need any more preasure both surfaces are in contact.

[#Rotor]

here is what I can muster, with 2 fingers. Yes only 2 fingers worth of downforce, and I'm sitting on my ass, while doing it





yes only a small round dab, smack in the middle.... I am crazy indeed .....



so what you guys are saying, is that once I get to a perfect footprint, evenly over the entire core, with no room left between the core and the sink, having more pressure on it, will improve things, I can honnestly not see how......
I mean I used 2 fingers to get the perfect seat, and I was not even going at it all the way.... I'd guess about 7 to 9 PSI on the actual core, max....(considering the core is much smaller the 1 square inch. my fingers are puny little stumps, maybe capable of about 3 LB of pressure... (I'm a wimp! I know)

[bigben2k]

Well, if you remember that a contact without any paste results in an actual contact area around 5%, then it's not so hard to see that adding pressure will result in more direct copper-to-die contact.

Cathar ran the numbers for his Cascade block, and the TIM joint, for an 80W source leave an 8 degree Celsius spread, between the block and die, through the TIM joint.. So anything you can really do to minimize that, surely helps.

edit: of course if the CPU is idle, that temp difference would be proportionally lower.

[#Rotor]

I here you on the more pressure, in order to squeeze more goo out, so more actual core/sink can touch. and that I fully understand and completely agree with..... BUT more pressure does not result in better conductivity..... It is the better contact that give results.


What I'm saying, is that you do not need to put more down-force on the chip, than what is required to get the perfect seat. once that seat is established any additional pressure, is not going to do you any good. I always "TUNE" my blocks, when I mount them. and this is in-fact what I find when doing so......

Tuning a block, very similar in procedure to torqueing a cylinder head onto an engine block.

I find that I can apply pressure to a point, where my temps are as low as they seem to want to be, I play with the leveling of the block, to get the perfect seat, then I can go tighter and tighter to my hearts content, it does not do a single thing for heat transfer, nothing, other than getting me all jittery and nervous. the optimal tention I found easily obtainable with tightening the nuts with my fingers.

[bigben2k]

Originally posted by #Rotor
...BUT more pressure does not result in better conductivity...

You're right, it doesn't improve conductivity. What it does is reduce the temperature gradient, between the heat source (inside the die) and the ambient air. By reducing the gradient, it's possible to reduce the die temp.

Personally, I'd rather stick to AMD specs. If I had Bill's surface gauging tool, and I knew the deflection that I can expect from my block, and if the surface geometry could be bend to be flat with a reasonable amount of pressure, then I'd give it a shot. IMO, clamping the block near 100 lbs, puts the core in a position where if you even bump into the PC, it could crush the core.

[#Rotor]

is that gradient not just an easily understandable way of looking at the heat transfer coefficient of a system?

AMD spec is based on us using that pink goo on our blocks, it give an indication of what the core will be able to take, yes, but for me that is all it is. Using a far superior thermal past on waterblocks requires looking as the down-force requirements. It gives us the ability to get perfect performance with much less than what would be required, if We had to use the STD heatsink goo that comes with AMD HSF's

[bigben2k]

I'm not sure what you mean, but you can look at it from an opposing perspective: what if there was no mounting pressure whatsoever?

The paste would have to fill more than 95% of the gap, and the temperature range between the die and ambient air would be greater. The standard (spec) mounting pressure gives us a 5% pure copper to die contact.

I'm still working on getting a thermagon sample: we'll see!

[#Rotor]

"what if there was no mounting pressure whatsoever?"

aaaa I like the way you think about things.... taking them to extremes are very often a very good way to illustrate something that otherwise, would be rather difficult to grasp.

your extreme to that side does hold true, but going to the opposite extreme will not result in the junction loosing all its thermal resistance, as infinite pressure applied to the junction is approached, that is my argument. There is a limit to the amount of gain that can be had by applying more and more pressure to the junction. This perceived gain is as a result of the diminished length of the thermal compound barrier( or layer), between the two surfaces.This length-reduction as result of added pressure, is however finite, once the two surfaces are touching (remember, they are not going to deform each other) the compound will not be able to get squeezed any more, right. That is where the optimal seat is achieved, more downforce from this point on, will not help, as there is nothing more to give way.

a graph.... I hope it makes sence.



note that anything from about 25N and more, gains are virtually flat-lining.

[bigben2k]

Nice one! I wish I had remembered it myself, because I posted a link to something almost exactly the same, a couple of months back!

[#Rotor]

I have some old EDO ram here, if you need some ya know... to get them memory banks up to scratch again.......[jk]