Do you think I could get better results if I take out the core, and mangle it in my garbage disposal for a few seconds:D


BTW, I've edited my first post with the best MAZE2 measurement that I could manage. Delta is 9.4C, water 21.3C, ambient 18.9C

All measurements +/- 1C

probably not, but if you aimed a sandblaster at it you might be able to drop the temps

To tell the truth, I don't understnd why it performs so well. There is a lot of copper between the core and the water. I would think this would cause a large temp gradient to develop.

ROFL, no I don't think either the block or your garbage disposal would like that experiment Schoolie :) (I bet the block would 'win') BTW, wasn't saying it is a bad block at all, just that it could be made better. With most blocks, while a lot of attention is paid to surface area and/or flow rates, not enough attention is paid to the type of flow.

From the pics it looks like the water makes contact with the copper slug all the way down to the base of the slug, so while it's a long way to the top of the slug, it isn't that far to the lower perimeter of it. And the thermal resistance from the lower rim of the base to the core is pretty low.

The key to the equation is what the temp is at equilibrium. I would suspect that the top of that slug stays right around water temp, so yes, you do have a fair temp gradient. At the same time, having that much thermal mass at a low temp is part of why the CPU doesn't get too hot.

Another way to look at it. Let's say the C/W of the copper slug is 8 (from top to bottom). If the top of the block is 1 degree above water temp, then the bottom of it stays at 9 degrees above water temp...which is what is touching the CPU die. Since the C/W of a proper layer of ASII(I) is very low, that means the die will stay close to the temp at the bottom of the slug.

I would be curious what the temps would be if you left Prime95 running overnight.

Yeah, I think you're correct about the block winning that fight:) The block was basically at equilibrium when I did the test. I'd been running Genome@home for about 12 hours before the test, then ran prime95 for about 30 minutes before I made my measurements. The cylinder is about 3.5cm in diameter, and the distance from the edge of the cylinder to the core is about 15mm, which is large compared to the 4mm base thickness (I'm guessing) of the MAZE2. Anyway, I think I could seat it better because the collars of the hose barbs were hitting the mounting gear. I'm going to switch one more time and test it.

Thanks,

Rob

well, why doesn't someone modify their block to having the inlet and outlet much lower on the block so that the coolest water is at the bottom of the cylinder

it looks like a stacked plate heatsink enclosed in aluminum. That is probably a great design. I would imagine that the lower you vould get your inlets to the core the better since cooler water would be getting the the core first.

--Matt

Odd that you have having problems with the 1/2"OD silicone; I am using 1/2" ID thick silicone from DD with no problems. I don't have the threads bolted in terribly far though; a couple turns and then I used goop to seal everything :)

As far as modifying it to get the water at the base more let me think about how that could easily be done. Most of the simple ways would either kill flow or else result in air getting possibly trapped.

Schoolie - don't forget on the DD2 that 4mm spacing is only at the center surface and increases as you go outward in the channel (distance from the center added to the thickness).

pHaetus - he's using nylon barbs as I remember. They tend to be slightly larger at the shoulder for strength. As far as adjusting the height on the barbs, would probably be best to leave the outlet as far up (not very far into the block) as possible. Ignoring ease of the mod - you could take the inlet barb and remove half of the body for the portion that would actually be in the water, put an index mark on it, then put it all the way in so the removed half of the body was facing the center. (cut the threaded portion down the centerline of the opening, then make a 90 degree cut to that a number of threads up from the bottom - see the quicky drawing attached).

I have a question about that inovatech block. It appears to be a copper plate heatsink encased in a block of aluminum(anodized?). Is corrosion a big risk when using something like that in a system or does the anodization of the metal prevent that from being a problem. Or, if it is, will something simple like water wetter fix that problem?

--Matt

Anodized aluminum is pretty corrosion resistant and is electrically non-conductive at the surface (which prevents the problems of galvanic corrosion). So as long as there are no gaps in the anodization, there should be no problem.

From the pics the Innovatek w/b looks to be anodized, but one can't be certain just based on a pic.

Yes, the block is anodized, both inside and out...Of course I managed to scratch the outside pretty well:) After I was done with my Innovacool, it looked less like a feat of German engineering, and more like a high school science project.

if you cut the channels deeper on the core you would increase surface area but you would simultaneously reduce water turbulance (it could go more or less straight through, instead of having to curve around and be buffeted by the sides of the channel) and inhibit how well heat travels to the top of the core. they could probaly do with being a bit deeped though, it needs to strike a balance between the two i guess.

slop

Is it aluminium shell? Isn´t plastic?

Hedgren: The shell is Aluminum.

Since the water only contacts the side of the copper cylinder, and the top, I don't think forcing the water to the bottom would improve performance. It would be better to have a shorter inlet. I think cutting the channels deeper in the core would improve suface area and increase turbulence because the water would be forced over a more jagged surface. However, I'm sure the folks at Innovatek experimented with this and arrived at the best shape for the cylinder surface.

True, after all it is a feat of fine german engineering :D

Since there are a lot of questions about optimizing the design of this block, does anyone know it it is possible to purchase aluminum bolts of ~ this diameter and constructed of high purity aluminum? It might be easier to just construct an all aluminum knockoff and then play with the design parameters rather than destroy a bunch of these expensive (and pretty) blocks...

You can get 1145 and 1100 round or square bar in close to that size. (1100 is 99% pure, 223 W/m-K, 1145 is 99.45% pure, 227 W/m-k). If there isn't a metal supply house nearby, you could contact Metal Men. Not sure how they are with small quantities though.

Innvatek has phat stuff but highspeedpc dont sell none of it check the dull rad they got this thing looks nuts also does anybody know were tom got that waterblock for that p4

http://www.innovatek.de/sites/wkuhler.html

Last I checked AMD's spec changed on thermal readings. They now recomended using a probe under a core or the thermal diode. I dont have the quote but I found it very interesting abotu 2 months ago to be reading that they ditched the lil hole in the HSF for something that measured the core temp from underneath.

I am guessing this has changed again since then cause reading from under the core on the new XP/MP cpu's is pretty pointless (and from all measurements so is the thermal diode)

If I get some time tonight I will have to go and read around AMD's site to see what they recomend now. Also I personally dont buy reading temps with the lil hole in the HSF or WB... never have and never will. Just one of those things...

But it all doesn't bother me. I just dont see testing on Cores as being too practical in the near future. Just as Joe at OC'rs has done with his "simulator" I am also in the process of designing one. Just need the funds to get it fabricated. One thing that amazes me is how cheap precision heating elements are (even the very high thermal density ones), I thought they would be far more expensive... the PSU is the expense.


As far as how they get the slug into the aluminum, I would bet the Heat/Chill shrink fit along with some sealant material for extra measure.

it drilled ok, interesting wb
that hole serves not to measure the die temp, but the bp
(not talking about AMD's use of it)

got a source on "cheap" heating elements ?
what voltage they run at ?

indeed, a 400W regulated DC PSU is pricey

Bill I am sure you know of a few Thermal element providers yourself but Minco is one I have been dealing with for the past month on some ideas. I talked with a company about 7 months ago that could make a 200 watt test core the same size as an athlon core... but it cost in the ball park of 2500 - 3000$ for the core plate, the PSU, and the computer cables and software to regulate and monitor it. After they told me that I sorta just forgot about going that route.

Going with a bigger heating element (lower density.. in the range of 350 watt/Sq Inch)and a well insulated heat "concentrator" design to give a solid contact patch = the size of a CPU Die ( mutliple concentrators to use to simulate multiple CPU types and pelt types). I will prolly post something with the final design and part #'s and all the materials used when I get it built.

Its actually sorta funny how I priced out a Wesgo metals CuSil plate to mill the concentrator out of, then like 1 week later Wesgo comes out and says that CuSil really aint that great, and Silver is better than it hehehe. I used to know someone who would sing the praises of Cusil :) hehehe ( just poke'n fun):dome:

joe... why dont you put a modified cold plate on it?
maybe something that is 1/2" thick that covers the whole element.... then ontop of that something the size of the die?

easily done with 2 pieces of copper, but you could have one milled for better conductivity.

Cause a 1 peice milled concentrator thats designed to focus the heat in one small square will do much better than just a cold plate with another plate soldered on top. Also these will be silver more than likely. ( depending on price, silver is somewhat "cheap" compared to some other materials)

as I see it, eh ?:

the nut of the difficulty is KNOWING the actual die 'face' temp
(without allso creating a 'thermal shadow', a known problem)

I sure don't know of a 1sq.cm ceramic heater with an embedded TC (though more likely an RTD/PRD)
that will even fit on the same page as the word cheap

in case you missed it, some details here

and some equip description here

in particular the links in the last post to the 'industry' norms

kind of fortunate that the writer did not recommend Cusil, as it played out
(ck the text, Cu, Cu, and more Cu - something about cost, even Ag)
ehem, your friend Chip gave me some Cusil, need it ? < G >

EDIT: the 2 piece bit won't do it, for sure must be monolithic
(making a mess out of this post, trying to clean it up)