This might have been tried before but maybe this is different enough for a new post. I am transfering this from another forum so its all in one go instead of multi posts. Looking at the core2 cpu and the way the 775 hold down works I thought it might be possible t bond a waterblock directly to the top of the cpu Here is my sketchup model with transparent on. I have milled an aluminium top made to fit on a the CPU which will fit through the Socket 775 hold down bracket. I am going to use magic metal adhesive to attach them together it is millable and drillable so should be tough enough and it can be used to repair boats and fuel tanks so coolant should not be a problem either. Here are some pictures. It took about 4 hours work milling. Block Top Block underside With Fittings press fit on Core2 CPU. I need a flat end tap to get the fittings bottomed down I am sourcing a cheap cpu for first tests, not going to try on my £160 E8400 I got the aluminium bar in a 1 meter length so plenty to go at. Mk 2 Waiting for a plug tap to make the Mk1 usable, so I have done a Mk2 version with the holes closer together, pics below Mk1 and Mk2 together. closer holes Underside with Fittings Got a £15 core2 cpu from ebay to use as pressure test rig, but im gonna wait for the plug tap and use the Mk1 so if it works the Mk2 goes on my Q6600. Mobo has arrived Foxconn946GZ7MA gonna be my Home server after MS fix the problems. So here are pics of block in position not glued yet, and no presure on the hold down. fitted through the back of the hold down fine. Just need to ease the back corners slightly to give some wriggle room. So, just need to deepen the thread on my MK1 and im good to go for preasure test. Might just drill the hole down slightly see if i can leave some edge, I only need 2 or 3 mm more. Well, its stuck on. Think I used a bit much hardner cos it didnt half set fast but seems to be on and air/water tight. Got the fittings on and used an old pump in a small loop just to test for preasure leaks. Plumbed in block no leaks while sitting the loop, should do to test for leaks. Need to dig out my variable PSU then I will power the sucker up. Fingers crossed. Its running no problems so far gonna leave it going over night in the shed psu drawing some juice. I squeezed the exit pipe a few times to increase the presure and no probs. ** Well that's 5 hours and no leaks, gonna check again in the morning ** Was looking good was fine after 14hrs on bench then fitted in mobo. water, water everywhere. Think the hold down flexed the block away from the CPU prised it off with a penknife, the adhesive is more like a clay paste than a metal so I think it was the excessive amount of hardner. going to try and get some new J-B weld and try again.
How friggin' awesome is that?! That looks like very sexy engineering, and eminently feasible. Sorry about your setback, but it looks like if you can get it to work it will be an outstanding achievement.
I got some J-B Weld result seems much better, it was more fluid and took a few hrs to set. I am now testing the block mounted in the Mobo, where it has been circulating coolant for 4 hrs now. with no losses. Setup the board in the shed (just in case) Closeup of CPU and another from other side If it goes ok for the rest of the day I will try powering it up. once I get a rad into the loop.
Awesome! given right materials, this would have a lot of potential for cooling, as more suface of cpu is covered. Great idea
This is a great project, but like most ideas on the internet it's far from new. Search for "direct die cooling" on google. It was done quite a few years ago. I don't remember the details but I believe that the distance from cpu surface to intake barb was crucial, placing the intake directly over the center is a must and jet impingement would give even better results. Well, this isn't direct die but more like direct IHS cooling. Not to mention it's not single core either. I guess the same principles would apply. Good luck with your project. If you can beat the commercial blocks it'll be a great accomplishment.
It's a great edea and the execution is flawless, but you will never be able to get a good hold on the die with conventional epoxies. the smooth surface makes it difficult fo the glue to grab. The proof is in how the first glue came free of the chip cleanly. The constant heating and cooling will eventually cause it to fail. -I'm trying to work around a similar problem right now.
Apparently there isn't any mounting mechanism? It's all held together by the epoxy. Maybe a mounting plate+backplate and silicone sealant instead of epoxy would work better.
i would remove the IHS and braze the block onto to it, then refit the IHS. maybe stick a little heatsink to the IHS to improve the transference... great work mate!
thinking about it, i would just remove the IHS and put a water block direct to core, maybe use a shim to protect against crushing. back many moons ago i made my own water block for my A64 setup, the base was 1.5mm thick so there was less material causing a resistance to temperture transference. i might make a new one soon for my Q6600.
Maybe using an o-ring to seal the join between the cpu and the block would be safer than just relying on the adhesive. And also, maybe rough up the two surfaces you are bonding. However, having the coolant in direct contact with the cpu strikes me as inherently a bit dodgy . Im not sure I'd be happy with it from anything other than a short term, experimental pov. I have vision of water suddenly pouring into the cores of my proccessor However, having said that rock on
Brilliant idea - shame that S775 is fundamentally tricky to get working with this method. Any other socket you could knock up a hold-down clamp type thing, seal it with an o-ring, and voila, all is solved. Now the stupid mobo socket thingy gets in the way...
I did think about a sealed version but my milling probably isnt up to that yet, sketchuped it though. I had it with coolant flowing for a good few hours installed in Mobo, but cant kick it into life atm, either all my playing arround has screwed the board or it was iffy from the shop,its rma'd so should get another shortly. The JB weld hopefully should be up for it, it is rated up to 500F temps, and I sanded the top of cpu for greater bondage .
Yeah, but it's the repeated expansion/contraction that Cheapskate is worried about --in particular that the IHS will expand and contract at slightly different ratios than the block... how flexible is JB Weld in that regard?
Not to detract from your hard work but since it's pretty obvious you've abused this mb (getting water all over it while on counts as abuse) then you really shouldn't be rmaing it.