Discussion in 'Modding' started by Drysofar, 5 May 2008.
but they wont know
Maybe making a mb backplate and arm clamp like Drysofar made in sketchup would work, and use some O-rings to keep the water in the core? Might put too much pressure on the processor though, also it wouldn't be easy to get the block off without water pouring out everywhere, but I felt like I was onto something with that idea. I do like the idea of brazing the IHS to the block though, but if you do that you might as well just put a normal block on a exposed core and forget about the brazing hassle, and you have the efficiency of the water channels in a conventional block.
That heat spreader is glued to the PCB, and they are !@#$ tough to get off. Anyone know what they use in the manufacturing process? It also survives the thermal cycling of the CPU so if anyone has any idea what it is, it may be a good idea.
This remind me a little of a direct impingement water cooling experiment I saw a few years ago (but I can't find the link ).
I'm curious how the surface area for a more direct thermal interface trade off will work out.
I just saw this now, awesome!!! I've read a couple of ideas on this sort of thing years ago, but i haven't seen any success stories yet...
It really needs to be done as it eliminates the resistance of thermal compound and the base of a waterblock. The heat absorption of the water from the heatspreader can be increased by roughening up the middle/center area of the surface, with like 80 grit sandpaper - these pics must be posted everywhere, so you can watch people go "you noob, you need to lap that not roughen it up!!! you ruined it"
You could mess around with automotive silicon to form a custom o-ring and try to clamp that down, but even when cured, you can't clamp that down very tightly without breaking the silicon.
How about using a round waterblock with a nice little o-ring gutter, so that you can get a nice thick round o-ring for it, and clamp it down nicely! This would work great i think, as long as you don't roughen up too much of that heatspreader
just wait, you know Asus is sticking HS with built in waterblocks on there mobo's, well i wouldn't be surprised if Intel or Amd slap a simular idea on there extreme CPU's.
As far as the expansion and contraction of the joint and JB weld with heat, I dont think it should be a problem it mentions these uses and more on the web page.
cracked engine blocks
transmission & rear-end casings
cylinder heads & sleeves
generators, starters, water & fuel pumps
axles & hubs
press fit bearings
I think a climb from 25 to 35C or so on startup and maybe a rise to 50C on load should be a problem. But I always use non conductive Fesser one coolant now just in case
body shop repairs
Is it possible to remove that stuff later on? I don't think it's a good idea to irreversably bond the cpu to a waterblock.
Who cares? This is just like voltmodding - you aren't trying to maximise resale value, you are trying to maximise performance.
You shouldn't trust what JBWeld says about it's product. It's one of the oldest epoxies on the market, and the claims are notoriously grand. This is a Bubba product, with no technical specs or chemical resistance information.
All we ask is you make a clamp to hold it in place, and switch to a automotive silicone sealant.
This isn't like a voltmod. A voltmod can be undone for starters. Second, a voltmod doesn't prevent the component from being used in other systems.
If he attaches the cpu to the waterblock permanently then it can't be used in another system unless it's also watercooled, and the chances at selling it off secondhand at a decent price are slim.
if the whole point of this is to maximise performance then he's better off with a d-tek fuzion. It'll come with less risk, probably perform better and he'll be able to sell both the cpu and the waterblock when he decides to upgrade. Jaydee from the procooling boards already proved that direct IHS cooling is less effective than a commercial waterblock.
You can't just pump water over a cpu and expect to get better cooling than an apogee GTX for example. You have to experiment alot and try different designs if you want to even hope to beat a high-end commercial block with a design that's been researched and tested. So far this is Drysofar's first design and you're suggesting that it's ok if he goes ahead and permanently bonds the cpu to an untested cooler that may potentialy not perform so great.
It is like voltmodding as in they are both experimenting at the risk of damaging hardware.
There is the potential here to better any waterblock imaginable, as there is no thermal compound resistance at all. There is also no resistance from a waterblock either. Copper is a great conductor of heat, but it does not have zero resistance to heat, and does not absorb heat anywhere as fast as water does.
This sort of experimentation is courageous, and should only be given support, not comments like "i don't think that's a good idea"
I've personally been told that many times when attempting many different things, not just limited to pc's or electronics either, and every time i hear it, i tell that person to just watch and learn.
Jaydee from procooling.com undertook this project less than a month ago. Please take a look at his work: http://forums.procooling.com/vbb/showthread.php?t=14751
It's a simple design. As a cheap alternative to a commercial block it's pretty cool, but if you want to aim for better performance than a high-end copper block you will need to research and experiment with it a bit more than Jaydee has done.
Well that sucks Ill keep going till i get the one i have running just to see, maybe core2 cpu's have a better connection between actual silicone and the IHS, Im working up a design for a clamp on version which might be more versatile.
Jaydee did it well It's interesting to see the results, which aren't as good as i expected, but that definitely does not mean the experiment is not worthwhile or anything like that!
Better temps will definitely be had if the heatspreader is roughened up, not where the o-ring makes contact of course, but a modern cpu doesn't have to be abused to find out how much better it will be. It would be great to perform this experiment on an unwanted hot prescott system!
Ok here is what I came up with for a clamp-on. Read the bit about impingement and smaller jets so incorporated that.
How it sits on the board, could be mounted in any direction.
With the x-ray specs on
The main cavity and the outlet, showing the inlet holes
The inlet block, bolted and sealed on the end.
I havn't bothered to show the seal rings, but they would be there.
Good to see that you are gonna use an o-ring, which is the method i was supporting O-ring also removes much of the hassle with the small amounts of expansion & contraction & flexing, etc.
Maybe someone has tried this with something like a prescott, and actually roughened up the ihs. Without roughening up the ihs you can't hope to get significantly better results than jaydee did. I dunno why he didn't roughen up that ihs? It's such an old and cheap cpu, he had nothing to loose really, all the hard work was done! He said there was no point going any further, which is crazy talk! That's like saying it doesn't matter if the inside of a waterblock, has any fins or pins or anything to increase the surface area & interference or not.
You know, if you don't have a prescott or something that you can do the ihs modding on, then it's pretty cheap to grab one for butchering
Excellent design. Nicely done!
Made some revisions, added more holes to spread the cooling
done with cross mills, front and back.
waiting on a 2mm slot cutter so I can knock up a non virtual one.
Any ideas on the hold down force needed to compress o-rings the required ammount?
Awsome idea! I like that you've revised it with jet holes and the like but I think that the idea is to have jets directly over the die (even in a commercial water block).
Just an idea but some thing like so:
Ashamed to say it, but this was actually done in photoshop, but with very tacky paint techniques.
I'm too tired to have photoshoped it properly but I think that you get the Idea.
Separate names with a comma.