direct die, partial success...

[Volenti]

The small copper jets are 1/16''(1.6mm) id (as best I can measure) the overall inlet/outlet dia is 3/8''

Flowrate in system is ~53gph (200L/H) (with the 2 186gph (700L/H) pumps in series)

nozzle-die clearance is a fraction over 1/8'' (3mm)

The 3 stage pump mod is still in progress, should get some decent progress done on it this weekend, ETA is about a week.(fingers crossed)

 

[utabintarbo]

As I recall, Volenti's block is epoxied to the chip (right?). I can't see a way to use an o-ring here - no way to clamp the block to the chip with sufficient force to seal. I suppose we could clamp to the socket or the mobo, but that almost guarantees a mess when removing the block for mods or whatever. Is there something I am missing here?

Bob

 

[Albigger]

That's great. i'll be watching the pump mod thread to see how it turns out.


what kind of head loss do you get due to your block? All this talk is making me want to go direct die even more now.....

 

[JFettig]

Volenti, you have inspired me to do this, before this i just wanted to.. didnt really think about it more than that....... yes i think you could use an O-Ring on the block and use the 4holes to hold it down... kind of how we already have it, but instead of silicone or epoxy, just use an Oring
Yes that would be a mess, but I think there is a way around that...

 

[JFettig]

here is something I have noticed.. my temp equilibrium seems lower, I did not measure the ambient temp so i cant be sure of this, but its normally around 26-27c, now its down to like 23-24c

 

[Tokayla]

nice?

 

[Tokayla]

coundenty you expoxy a shim to the processor and then waterproof it then connect the block to the shim?

 

[Frodo Baggins]

wow, cool masked, but a bit dangerous for me

 

[JFettig]

coundenty you expoxy a shim to the processor and then waterproof it then connect the block to the shim?

that would be mostly pointless when you can just coat the bridges and such with epoxy

and frodo, its dangerous if you do it rong

 

[800XL]

A shim with threaded studs sticking out the top would be convenient way to attach the block. You epoxy the shim to the CPU permanently, and then design the block to use the studs for retention. That gives you a way to remove the block for modifications without having to take it apart installed on a board.

 

[Tecumseh]

I don't think direct die cooling can match a convential waterblock simply because the heat has no where to spread out except into the water, and waters does not conduct heat like copper.

This is why it is so important to get the jets right. Jet impingement
done right CAN beat copper...by a big amount. The key here is
"done right." The direct die WBs that have had poor performance
in the past almost certainly have been poor because of poor
jet design.

Anyone who wants to blow away conventional WBs with a
direct die design should focus efforts on optimizing the jet
design. This is the key.

 

[Iron Hawk]

That would be cool to see the core while it's running (something noone ever has, for long IMHO).

When I direct Die cooled my block you could see the core while it was running. unfortunatly that project only lasted a few days, I might try It again sometime though.
If i can find the pics I have of it Ill show you.

 

[Liss]

wut shows on the core????


and wouldn't the water pressure destroy the core over time?

 

[Volenti]

As I recall, Volenti's block is epoxied to the chip (right?). I can't see a way to use an o-ring here - no way to clamp the block to the chip with sufficient force to seal. I suppose we could clamp to the socket or the mobo, but that almost guarantees a mess when removing the block for mods or whatever. Is there something I am missing here?

Bob

Yea my block is epoxied to the cpu package, I had no capability at the time to make a channel in the block for the o-ring.

I feel that, in it's current mounting setup, the block should seal ok with an o-ring.

With the potential mess from removing the block (with an o-ring seal) in this block design I can give the intake hose a firm blow and force out the majority of the water inside the block (which doesn't hold a lot of water anyway)

In a normal vertical mount suitation you could position the exit hole at teh bottom of the block so you could get almost all the water out of the block before you take it off.

That's great. i'll be watching the pump mod thread to see how it turns out.


what kind of head loss do you get due to your block? All this talk is making me want to go direct die even more now.....

Head loss is substancial, it reduces the no head flow of just a single pump to 1/7 , but that's the price you pay for more pressure...

A shim with threaded studs sticking out the top would be convenient way to attach the block. You epoxy the shim to the CPU permanently, and then design the block to use the studs for retention. That gives you a way to remove the block for modifications without having to take it apart installed on a board.

yes, I looked at that approach for a removable block, I can't see any reason why it wouldn't work well.

This is why it is so important to get the jets right. Jet impingement
done right CAN beat copper...by a big amount. The key here is
"done right." The direct die WBs that have had poor performance
in the past almost certainly have been poor because of poor
jet design.

Anyone who wants to blow away conventional WBs with a
direct die design should focus efforts on optimizing the jet
design. This is the key.

exactly, my direct die block beats (temp wise) all my copper blocks bar one (which it matches) and gives a more stable (far more stable) overclock than that one. When I finish my 3 stage pump mod and (hopefully) gain signifigantly more pressure, we'll see if that can improve temps a bit more.

wut shows on the core????


and wouldn't the water pressure destroy the core over time?

I think he means he can see the stepping code stamped into the die.

oh sure, water is pretty erosive, especially under pressure, and will , in time, wear through the core, but I'll most likley be dead by the time that happens, so it's not something that concerns me

 

[TonyMc]

If you have a steady hand or can get your mill to rotate real slow you could use a bit like this to get the holes tapered. Look up #7134 Diamond Taper Point.

 

[JFettig]

oh yeah, forgot to add, my cpu died after a short while, everythign was good and sealed and waterproofed, death cause, unknown

 

[SysCrusher]

good advice, from what i know (not a whole lot)

has anyone attempted to adapt a pressure washer to do direct die? It would be insane, but man i bet it would cool good....

How much velocity do you think the core can take, from the water impact? before higher velocities start to get dangerous??

I wouldn't try it with a pressure washer but try it with a garden hose nozzle. I tried this with a normal TC4 water block but not in a direct die situation. I can tell you it works great but looked way to ghetto for me and I ruined a TC4 top. I just soldered a straight garden hose sprayer to where the barb would fit. Then you could adjust the spray to what you wanted. It created the fast pressure velocity you needed plus it held the water in the rad longer to be cooled a few degrees below ambient. My only problem was that it looked to ghetto for my taste plus wasn't perfectly leak proof - I didn't trust it. You also need a pump with a high head or capable of creating high pressure. The normal pumps we use won't cut it with their one foot or less head. It worked well with atleast 5C better than most blocks out there, with a high head pump it should be better. Just for fun I hooked it up to my sink faucet and it kept my temps at 20 - 22C. So I'm sure you could do that with a high pressure pump.