Pro/Forums - View Single Post - More Teory

Cathar · 11-19-2003, 05:00 PM

What you're describing is a true micro-channel block. Pretty much no good for doing a central-in impingement zone that the White Water uses. The White Water is sized the way it is because it allows the primary impingement region to be wide enough to cover the size of most CPU cores, basically around 2.5-3x of the nozzle width. The walls don't need to be more than 5x the channel width high. Anything more is bordering on pointless, besides machining 1mm deep x 0.1mm wide channels is pretty difficult to do.

So we get to what you're proposing, which is in essence a 10x shrink of the White Water channel design, with 10x more channels, and without an impingement region.

In terms of the cost of conducting through the copper, you're winning because the heat doesn't have to move as far, but losing because you've lost the effect of the impingement that soaks up a large proportion of the heat before it even needs to start moving up a White Water's fins. You may be gaining 1C or so here.

Basically you're asking the water to travel though a net orifice area of around 3.8mm^2. Les would be a good person to help out here, since as you restrict water to flow though very tiny channels the flow resistance goes more more than linearly to what one would expect.

I don't know an exact figure, but let's assume a basic 20x increase in the flow resistance over a White Water. With something like an Eheim 1048 you're probably looking at something less than a 1LPM flow rate. The actual water velocity in the channels probably won't be a great deal higher than what occurs in a WW's jet, maybe 2x at best - so we gain a bit here. Then we lose again because the water will be warming up significantly on a hot CPU. At 1LPM you can expect the water to be warming up by around 1C, whereas for a WW it's heating up by around 0.2C, so you've lost again.

Now the above is a simplistic assessment of my thoughts on true micro-channels. For very low flow pumps/systems, they totally rule. If all you're going to do is push 1-1.5LPM through your system, then not much else will match it. The Cascade design could be tweaked for better ultra-low-flow performance than what it presently offers and match and possibly beat micro-channels at such low flows, but this is a case of designing for a setup that is not in common use.

In general there is greater freedom when targetting >2LPM flow rates.

A number of companies are pushing the benefits of true micro-channels, and while I do agree with them that for super-low-flow rates (<0.5LPM) that they are indeed about the best that can be done, I do take issue with the claim that such a design is the best than anyone can do, especially if the peak flow-rate limit is lifted.

11-19-2003, 05:00 PM	#2
Cathar Thermophile Join Date: Sep 2002 Location: Melbourne, Australia Posts: 2,538	What you're describing is a true micro-channel block. Pretty much no good for doing a central-in impingement zone that the White Water uses. The White Water is sized the way it is because it allows the primary impingement region to be wide enough to cover the size of most CPU cores, basically around 2.5-3x of the nozzle width. The walls don't need to be more than 5x the channel width high. Anything more is bordering on pointless, besides machining 1mm deep x 0.1mm wide channels is pretty difficult to do. So we get to what you're proposing, which is in essence a 10x shrink of the White Water channel design, with 10x more channels, and without an impingement region. In terms of the cost of conducting through the copper, you're winning because the heat doesn't have to move as far, but losing because you've lost the effect of the impingement that soaks up a large proportion of the heat before it even needs to start moving up a White Water's fins. You may be gaining 1C or so here. Basically you're asking the water to travel though a net orifice area of around 3.8mm^2. Les would be a good person to help out here, since as you restrict water to flow though very tiny channels the flow resistance goes more more than linearly to what one would expect. I don't know an exact figure, but let's assume a basic 20x increase in the flow resistance over a White Water. With something like an Eheim 1048 you're probably looking at something less than a 1LPM flow rate. The actual water velocity in the channels probably won't be a great deal higher than what occurs in a WW's jet, maybe 2x at best - so we gain a bit here. Then we lose again because the water will be warming up significantly on a hot CPU. At 1LPM you can expect the water to be warming up by around 1C, whereas for a WW it's heating up by around 0.2C, so you've lost again. Now the above is a simplistic assessment of my thoughts on true micro-channels. For very low flow pumps/systems, they totally rule. If all you're going to do is push 1-1.5LPM through your system, then not much else will match it. The Cascade design could be tweaked for better ultra-low-flow performance than what it presently offers and match and possibly beat micro-channels at such low flows, but this is a case of designing for a setup that is not in common use. In general there is greater freedom when targetting >2LPM flow rates. A number of companies are pushing the benefits of true micro-channels, and while I do agree with them that for super-low-flow rates (<0.5LPM) that they are indeed about the best that can be done, I do take issue with the claim that such a design is the best than anyone can do, especially if the peak flow-rate limit is lifted.