Pro/Forums - View Single Post

Cathar · 12-14-2004, 08:06 PM

Volenti is perhaps the most successful person I've seen with a direct die setup using a set of high-speed jets. Using a fairly meaty pump (MD-20RZ) he was able to get it to roughly perform on par with his old White Water clone that he made up himself, but this was also on a P4 die (~147mm^2).

I have a feeling that an extremely well setup direct die setup can come close to a top-end enclosed water-block, but it requires the use of fairly strong pumps to generate the jet velocities required to over-come the surface area disadvantage.

Really this is what it comes down to. Enclosed waterblocks have the TIM + conduction cost but make up for it with increased surface area. Direct-die doesn't have the TIM + conduction cost, but will typically have 1/2 the convectional surface area available.

It's a case of 6 of one and half-dozen of the other. Jet impingement on such a small scale is highly dependent upon jet velocity, and hence the pumping power. This is an issue that I tried hard to solve with the Storm blocks, but none of the performance enhancing features of the Storm blocks would be available in a bare die.

My best answer is that "maybe direct die could be as good as an enclosed waterblock, but my feeling is that it will fall somewhat short". Direct-die waterblocks come with the added hassle of being more difficult to setup and maintain without dripping water over the inside of the computer. Since CPU's are not water-proof either it becomes even more of an undesirable cause to attempt to prove because the CPU will die anyway.

Given the drawbacks and the evidence which suggests that at best direct die cooling is still somewhat behind the current top-end waterblocks, and at worst a long way behind, this hardly instils me with a lot of faith that it's something worth pursuing. This makes the answer just pure theoretical conjecture in the mean-time until someone proves otherwise, and right now the limited experiments and theory available tell us that enclosed waterblocks are superior for performance, as well as for practicality (and also for lack of CPU death).

12-14-2004, 08:06 PM	#50
Cathar Thermophile Join Date: Sep 2002 Location: Melbourne, Australia Posts: 2,538	Volenti is perhaps the most successful person I've seen with a direct die setup using a set of high-speed jets. Using a fairly meaty pump (MD-20RZ) he was able to get it to roughly perform on par with his old White Water clone that he made up himself, but this was also on a P4 die (~147mm^2). I have a feeling that an extremely well setup direct die setup can come close to a top-end enclosed water-block, but it requires the use of fairly strong pumps to generate the jet velocities required to over-come the surface area disadvantage. Really this is what it comes down to. Enclosed waterblocks have the TIM + conduction cost but make up for it with increased surface area. Direct-die doesn't have the TIM + conduction cost, but will typically have 1/2 the convectional surface area available. It's a case of 6 of one and half-dozen of the other. Jet impingement on such a small scale is highly dependent upon jet velocity, and hence the pumping power. This is an issue that I tried hard to solve with the Storm blocks, but none of the performance enhancing features of the Storm blocks would be available in a bare die. My best answer is that "maybe direct die could be as good as an enclosed waterblock, but my feeling is that it will fall somewhat short". Direct-die waterblocks come with the added hassle of being more difficult to setup and maintain without dripping water over the inside of the computer. Since CPU's are not water-proof either it becomes even more of an undesirable cause to attempt to prove because the CPU will die anyway. Given the drawbacks and the evidence which suggests that at best direct die cooling is still somewhat behind the current top-end waterblocks, and at worst a long way behind, this hardly instils me with a lot of faith that it's something worth pursuing. This makes the answer just pure theoretical conjecture in the mean-time until someone proves otherwise, and right now the limited experiments and theory available tell us that enclosed waterblocks are superior for performance, as well as for practicality (and also for lack of CPU death).