Water Block Roundup Part 1 - Discussion

[Brad]

I'd actually like to know how well it does cool. like THG is hardly reliable, and the german testing doesn't compare too many blocks we know about. Remember they compare the old maze2, and they use a low flow rate as well

[webmedic]

Is it too late to get a block added to the testing. Alpha tech sent me a pal 8045 for modifaction into a water cooling bock. If I get it done in time can it be added to the testing.

[Joe]

Sure!

I am moving so there is going to be about a month delay on getting the roundup posted.

So anyone who hasnt sent stuff to me yet, hold on till I find out where I am going to be moving to, so I can give you my new address and such.

[Brad]

hey web, how are you going to arrange it? You should have a centre inlet, with an aluminium U shaped bracket. So half the water turns left, and half right, and goes around to an outlet just behind the inlet

[webmedic]

Actually I was looking at incasing most of the fins into an enclosure and then simply pumping water through all that surface area. The idea being that if it cools so well with air what will it do with water being shot over the fins. It will have an inlet over the cpu. The outlets have not been decided on yet but it will evacuate through another set of barbs that I will probably mount on the sides.
Some of the fins on the sides will have to be removed to make way for the enclosure. It will still mount using the factory mounting holes so they will need to be outside of the enclosure.

[Brad]

yeah, but if you have the inlet in the centre, no channeling, and have the outlet on the side somewhere, it won't work too well I think.

Anyway, look at the pic, thats my idea

[Joe]

Brad thats basicly the design that OCWC used on that IceFloe and CuSil block

[Brad]

I know that, but what else can you do?
what about this idea?

[Bollocks]

heh, im waiting for brads design to look like tic tac toe

[mkosem]

how about this one

[Brad]

remember this is an alpha pal8045, not a free design

[Pinkster]

wonder if the socet a will be delayed enuff to let the new aussie designs (the cyclone 3,4,5) compete.. they come out in about 2-3 weeks.... considering the cyclone 3 a 55$ buget design is supposed to outperform the cyclone 2 (according to silverprop)... the $125 cyclone 5 should be fantastic. (note they are also releasing some northbridge and gforce 3 blocks for 55$)

[Elemental]

Originally posted by icel0rd
Hey Elemental, where did you get the mounting hardware for that K4.1 block? That is exactly what I need!

Thanks in advance!

www.cooling-solutions.de

You can find the mounting under "shop/zubehör" .

cu
Elemental

[Brad]

yeah, Silverprop shows a lot of promise, their only weakness being they are in Aussie without an US distro.

But for someone in NZ, it is great....

[WebMasta33]

This is my design. But I'm going to eliminate the outer channels. (Had a friend render it from my autocad deisgn)

I'm going to alter the deisgn, and stagger drill (because I'm going this with a drill press) the center channel. By that, I mean make every hole that I drill about .025 inches of center of the channel line. This will improve heat transfer by causing some micro-swirling (hopefully) in every hole. But, because they are only a small but off the center line, there won't be that much resistance (hopefully)....

[ksw]

umm ist it smarter to reverse engineer an alpha pal 6045 and drill 1/2" hole in middle, through the fins, and stop at bsa, and have 4 outlest, connected to a tank?

[scythe]

Originally posted by WebMasta33


This is my design. But I'm going to eliminate the outer channels. (Had a friend render it from my autocad deisgn)

I'm going to alter the deisgn, and stagger drill (because I'm going this with a drill press) the center channel. By that, I mean make every hole that I drill about .025 inches of center of the channel line. This will improve heat transfer by causing some micro-swirling (hopefully) in every hole. But, because they are only a small but off the center line, there won't be that much resistance (hopefully)....

wonder what it would be like if you cut the middle channel and just use the out channels with the swirlys

[Mr Evil]

That is indeed a very thorough review, but I would like to make a couple of comments and suggestions:

I agree with newbie that temperature-delta is the most important figure, but it should really be the difference between core temp. and water intake temp, as this eliminates every part of the cooling system except the waterblock itself. Personally I would then prefer it if this where then divided by heat output and displayed as thermal resistance in C/W as this is a standard measurement of cooler performance.

I also agree with newbie that maximum temperatures are the important ones, as it is maximum temperature that determines the maximum stable core speed. It doesn't how fast or slow the temperature changes.

An additional calculation that could be done (not that you would want to do any more, having seen how much work you put into this!) is to work out how much heat, in Watts, is being transfered through the waterblock. This can be calculated from the temperature difference between water entering and leaving the block and the flow rate, and knowing the specific heat capacity and density of water. This would be useful because it would show how much heat is not going through the block (i.e. is lost through the underside of the CPU).

[Joe]

Simply, I think C/W is not accurate enough to be trusted.

The simple reason for this is: Can you prove to me how many watts of heat is going to the block?? Can you prove to me that "Radiate" is accurate? If you can give me 100% proof that I am putting EXACTLY that many watts of heat to a block, then I will start to use C/W. Till then I think its an irresponsible measure to use on CPU based tests since no one can promise you exactly how much heat is being put through it.

[pHaestus]

I don't think that the issue with C/W is that great; the maximum heat wattages (which is what radiate uses) are right from the AMD technical documents, and if you read up on some of the stress programs then they usually agree that ~ 89% of the theoretical max load is what they can generate. Here is a nice post about that:

http://www.benchtest.com/calc.html

Looking at the AMD technical document for the XP, then they also come up with 89% as the typical load temp. Anyway, the point of reporting C/W isn't to try to pull the wool over everyone's eyes or to add another fudge factor; it is designed to make results comparable between users more readily. If I see you have a TBird at 1600MHz and I have a 1600+ XP at stock, I can still estimate quite well what delta T that I would get from a waterblock by looking at your roundup. To do this though I have to normalize to the heat load and like it or not that means C/W. As long as I go get the heat wattage from the same place (AMD tech documents are good sources for this) then the comparison is perfectly valid.

The good news is that there aren't a lot of different programs out there that do this; Radiate is pretty widely accepted and seems to be pretty close to the 100% load numbers in the AMD pdfs. I personally use the 89% heat wattage for C/W as per the benchtest.com page, but the wattage I use in the equation is always right there in the review or article.

[Joe]

unfortunately I do see it as just another number to add to the mix. And I know where Radaite gets its numbers from, I just think there are too many variables that can affect the actual heat output of a CPU.

I mean you are trying to get something meant to do math to generate a constant heat. Well there is an OS running on that CPU, background services, temp reading services ( if you are using MBM), and many other things that will take CPU time away from your load app.

Windows is not that smooth with CPU time handover either, it will cause a lag before the load app gets control of the core after a process takes some time away. This will cause a drop in CPU load and a drop in overall heat output. The only place I think C/W should be used it on a Lab quality heat source that is known and tested to be true, and isnt affected by what ever background process may be running in an OS.

I may be wrong but I thought we were going for accuracy and not just nice looking numbers? Giving C/W may be the "in thing" now, but I dont buy it and dont care about those numbers when I see a review. Its that damn Fuzzy Math again...

[pHaestus]

I dunno about that totally. With my diode reader and 1600+ I get the same C/W every time I test the dynatron BH-635. There are of course slight variations in cpu load, but they are totally masked by the 1C resolution of the reader and the ambient temperature probe on my A7V-133.

Here is a useful aspect of C/W:

http://www.overclockers.com/articles518/index02.asp

See how the C/W values change with heat load applied? That is because there is something awry with his reader (probably the leads are too long). If the results were only reported as delta T, then I wouldn't be able to pick up on that nearly as easily.

[Jason711]

Originally posted by webmedic
Is it too late to get a block added to the testing. Alpha tech sent me a pal 8045 for modifaction into a water cooling bock. If I get it done in time can it be added to the testing.

let me know how you do that.. ive been kicking around that idea too.

[ksw]

yea acctualy. whoi hast. i first got the idea when i saw a zalman flowercooler. gold plater. thaught u could make some modifications and make one badass waterblock.

[UnloadeD]

Hey better late than never. Just read ya roundup. Still a good read, even though it's pretty dated by now. I just wanted to point out something about the graphs on the last page. The Peak Core Temp graph is missing the block from cool-computers. And something that would make the 3 graphs more reader-friendly. If you used the same color bar for same block across the 3 graphs it would be much easier to digest. I can't wait for the next installment 8)

peace.
unloaded