Micro-channel concept water block

[Cathar]

I've just picked up my block from the machiners. It's built around the micro-channel concept. The exact design and proportions were decided on a mixture of theory, simulation with a thermal simulator I wrote myself, some challenges to think a bit harder by BillA, Aesik and Les, and the practicalities of getting such a block machined and made, and still flow well with regular hobby style pumps like the Eheims.

The block itself in total (without barbs) measures 76mm x 47mm x 12.5mm high. The channels and walls are 1mm wide, 30mm long and 5mm high. The base of the block that does the actual cooling is 6.25mm high, and hence the base-plate thickness is 1.25mm. Water goes in via the central barb, and exits via the two side barbs. Despite the small proportions, the block flows very well, achieving 7.5lpm flow rates with a heater-core and an Eheim 1250 pump attached, which is about the same as every other block I have (Cyclone1, Cyclone 5, Maze 3), with the limitation here really being the inlet barb inner diameter. The block is made out of (almost) pure copper, and the internals have been bead blasted.

Performance to date exceeds every block I've ever used. After some initial teething problems with fixing leaks due to poorly applied teflon tape on the barbs (fault of the machining shop), and mounting irregularities due to mismatched springs and stripped nylon posts (my fault), the block is now shining strongly. I've been able to achieve a higher stable overclock (Prime 95 Torture Test) on this XP1700+ CPU than any other block I own. (1980MHz with Maze 3, 2024MHz with Cyclone 5, 2035MHz with micro-channel block).

I've been having some difficulties obtaining a reliable indication of temperature with this motherboard (Abit KR7A), however an extremely crude estimate is that the block is cooling my 2.2v/2.0GHz AthlonXP by around 3C better than my best block to date being the Cyclone 5, which is in turn itself better than the Maze 3 by a few degrees. I'll be attempting to get a clearer idea of temperatures other than this vague indication, but for now this is as good as I can manage.

I know you all want to see it, so here's the piccies.

http://www.employees.org/~slf/concept/cb1.jpg

http://www.employees.org/~slf/concept/cb2.jpg

http://www.employees.org/~slf/concept/cb3.jpg

http://www.employees.org/~slf/concept/cb4.jpg

http://www.employees.org/~slf/concept/cb5.jpg

http://www.employees.org/~slf/concept/cb6.jpg

[kev82]

Looks nice, be sure to tell us how it performs :)

How much did you pay the machinist to make one for you?

[Diggrr]

That looks pretty similar to mine, though I had only an inlet and outlet in the ends of that channel.
I would add that your's looks much better having been done by a machine shop instead of my dremel :eek:

Mine was tie-ing my Gemini block, but alas it's sold to my brother before BillA had an opportunity to test it.

How thick is the copper underneath the channel area? Mine was 1/4", but your's is obviously much thinner...looking good!
*NM...reread your post ;)

[Cathar]

Originally posted by Diggrr
That looks pretty similar to mine, though I had only an inlet and outlet in the ends of that channel.
I would add that your's looks much better having been done by a machine shop instead of my dremel :eek:


You were able to achieve 1mm wide channels with a dremel?

I tried the side-to-side configuration by blocking off the center barb. Performance is worse. The turbulent water flow coming in right over the CPU core really does help.

[redPEPPER]

Wow that's a fine looking waterblock!

I have some questions though. These are not critics, as I have zero experience with waterblock building. I'm just wondering how this fits with all I learned on this forum, and I wish to learn more :) Here goes.

- The intake barb has a wider screw, but it seems the inner diameter of the barbed part is exactly the same. So why the wider screw?

- This leads to another question: the outtake is twice as wide as the intake. Isn't the waterflow restricted by the intake?

- Based on your pictures and the given dimensions, I estimated the inner diameter of the barbs to be about 12mm. A rapid calculation indicates that the surface of the flow is around 113mm² in these barbs. Now if your dimensions are correct, you've got 8 channels that are 1mm wide and 5mm deep, which gives a flow surface of 40mm² on each side (80mm² total). That's lower than the intake (let alone the outtake). So the WB is probably going to be the most restricting part of your setup. Is that done on purpose? I thought a faster flow was better for cooling.

[Cathar]

Originally posted by redPEPPER
- The intake barb has a wider screw, but it seems the inner diameter of the barbed part is exactly the same. So why the wider screw?


The block is built for 1/2" ID tubing. All barbs are 1/2".

The wider screw on the middle barbs is so the opening straddles across all the channels. Its inner diameter is slightly wider than all the channels so when water flows in from the top, it widens out and feeds water into all the channels beneath it.


- This leads to another question: the outtake is twice as wide as the intake. Isn't the waterflow restricted by the intake?


The two outlets get joined externally by a Y connector back into a single 1/2" tube. The waterflow rates are exactly as one would expect for a full 1/2" system.

http://www.employees.org/~slf/concept/y.jpg


- Based on your pictures and the given dimensions, I estimated the inner diameter of the barbs to be about 12mm. A rapid calculation indicates that the surface of the flow is around 113mm² in these barbs. Now if your dimensions are correct, you've got 8 channels that are 1mm wide and 5mm deep, which gives a flow surface of 40mm² on each side (80mm² total). That's lower than the intake (let alone the outtake). So the WB is probably going to be the most restricting part of your setup. Is that done on purpose? I thought a faster flow was better for cooling.

The ID of the barbs is currently 9mm. I intend to drill them out to 10mm on my drill press. 9mm ID = 64mm^2 of flow area. At 10mm ID, it will be 78.5mm^2, or a pretty good match for the 80mm^2 of the channels.

Actually, 80mm^2 is not the full story. The ID of the larger central barb "screw" is 15mm. (Actually I've widened it slightly with a file at the edges so the "entrance" into the very edge channels is more open - this was done after the pictures were taken). This gives us 176mm^2 of opening area, of which half is walls and half channels, so really just 88mm^2 of the tops of the channels is open for the water to flow into, which happens to just nicely match with the 80mm^2 of the sideways channel area, and the 78mm^2 of the 10mm inlet barb after I drill it out. It's not really possible to mill out 1/2" barbs much wider than 10mm ID as the walls start to get too thin.

Good questions all but as you can see, everything is designed around balancing all the various openings to ensure good flow through the block.

[JFettig]

WOW that is very professional! you should try using the 2 outlets, plug the middle one and have it go in one outlet then out the other, see if there is a difference, that looks like a very professional block!!! i like it!

[SemiCycle]

Any "Home" block that lest you push your XP to 2Ghz must be nice. Great job.

[Cathar]

Originally posted by maskedgeek
you should try using the 2 outlets, plug the middle one and have it go in one outlet then out the other, see if there is a difference

Have already tried that. It performs a little bit worse. However, at this level its hard to track this down to mounting irregularities or actual difference in the way the water flows. Needs more testing, but initial results was that it was worse.

[redPEPPER]

Originally posted by Cathar
Good questions all but as you can see, everything is designed around balancing all the various openings to ensure good flow through the block.

Yep. It seems to be nice-looking AND well thought-out :)

One last question: the intake diameter will be 10mm you say. The outtakes won't be a problem until you combine them with the Y connector. What's the ID of the common part of that connector? :p it should be 10mm too if you don't want to restrict flow, right? It seems to be a plastic connector. Can you mill it too? Or is it already wide enough?

The question stands for all the other barb connectors on the 1/2" circuit.

[safemode]

i'd be wondering if the better performance (lower temps) are due to the design or the fact that you have less distance for the heat to travel from the cpu to the water and out of the block than normal blocks do. Normal blocks do the bigger base in order to get the heat uniform once it reaches the block's water cavity. I believe this is a hold-over from air heatsinks that need that in order to make use of all it's surface area in the fins and not just the middle ones. I dont think it's necessary in water block designs and i'm pretty sure your block is proving that. The design (micro-channel) may actually be very minor to your performance gains.

[Diggrr]

Originally posted by Cathar

You were able to achieve 1mm wide channels with a dremel?


No, not micro channels, I openned mine up to give more flow by cutting the channels into triangles, yet have more surface area than just an empty chamber.
Where your fins are, mine was like this horrible ascii depiction /\/\/\/\/\. There were a dozen peaks in 1 1/2" channel width.

[fletch]

thats so ingenious. great idea, it looks cool, and seems to perform the same ;)

[Cathar]

Originally posted by safemode
i'd be wondering if the better performance (lower temps) are due to the design or the fact that you have less distance for the heat to travel from the cpu to the water and out of the block than normal blocks do. Normal blocks do the bigger base in order to get the heat uniform once it reaches the block's water cavity. I believe this is a hold-over from air heatsinks that need that in order to make use of all it's surface area in the fins and not just the middle ones. I dont think it's necessary in water block designs and i'm pretty sure your block is proving that. The design (micro-channel) may actually be very minor to your performance gains.

The base-plate thickness is critical to the performance, but it's not as straight-forward as you're making out.

Blocks like the Swiftech MCW-462 derive their performance from using a thicker base-plate. This allows the heat to spread more evenly over the "pool" area, effectively giving a greater surface area to cool with. With such designs 6-9mm thickness is essential, unless you can start pushing flow rates right up, where at >3gpm a 4mm thick copper base-plate starts to become optimal. As you push the flow rates higher you can get away with thinner base-plates, but we're going to start to reach pretty extreme pressure levels to get such flow rates in 1/2" or even 5/8" systems. A flat 1mm base-plate block would perform horribly with very high heat loads (>50W real). There's a very good reason why Swiftech make their waterblock base-plates as thick as they do. It's very good balance for a range of CPU heat loads, the size of the water pool area, and the typical flow rates/pumps that people will push through it.

The real problem though is the copper itself. What's better at moving heat away from a spot where it's hot to somewhere else? More copper? Or turbulent rapidly flowing water via convection? However, convection is also proportional to the surface area that the water is flowing over, so we have to strike a balance of maximising surface area for the water, while minimising the amount of copper that the heat has to make its way through, yet also having enough copper in the right places to carry the heat away efficiently and far enough to help maximise the convectional surface area.

Through my calculations/simulations, the ideal ratio of copper:water is between 3:2 and 1:1, and for hot CPU heat sources, the base-plate should only be as thick as the water channels are wide for channels <3mm in width. Physical constraints and structural strength and integrity, coupled with ease of manufacture place some pretty hard lower bounds on how small we can make our channels. I chose a 1:1 copper/water ratio because I also had to balance flow rates and maximise heat copper over the very small die sizes of the CPU, and 1mm wide channels is as small as can be affordably machined without going to EDM (Electro-Displacement Machining). Ideally the channels should've been just 0.7mm wide.

The micro-channels maximise the convective surface area right above the hot CPU core. There's a surprising amount of convective surface area concentrated in a very small distance from the CPU. The micro-channels also make the use of the thin base-plate possible because the base-plate doesn't need to be thick to spread the heat far to make it up into the channel walls. A thinner base-plate means less thermal resistance due to the heat having to make its way through the copper.

[Daemonfly]

I'm glad to see home-brew blocks that are great successes :D Especially different designs from the comercial blocks.

[gone_fishin]

Very nicely done.:D Do you anticipate a slightly different cavity design if it were to be made for the P4?

[Cathar]

Originally posted by gone_fishin
Very nicely done.:D Do you anticipate a slightly different cavity design if it were to be made for the P4?

Actually, and maybe somewhat surprisingly to some, no.

The Northwood 0.13u P4 die size is 11.27 x 11.27 mm in size, or really about the same size as an AthlonXP.

The themal heat spreader which is primarily made of copper (nickel plated) is just 1.5mm thick. The bulk of the heat at this stage is still going straight up, perhaps spreading out over a slightly larger area than the core CPU die size. The IHS is 30mm x 30mm, but really its job is to spread heat in the event of a total heatsink failure (heatsink falls off), and when a heatsink is on top, the heatspreader doesn't really act to spread heat that much.

Due to this, we can model the behavior of the waterblock as having a base-plate thickness of 2.75mm (1.25mm for its own base, and 1.5mm for the IHS), and with a CPU die of 11.27 x 11.27mm. In reality, the channel area does not need to be any wider even for a P4, despite the IHS giving the impression that it does.

[Since87]

Nice design.

Do you have any concern about the thin baseplate bending under the Block to CPU pressure?

Obviously there is plenty of stiffening along the long axis of the block. Would it make sense to leave a ridge of copper perpendicular to the microchannels. With the water flow splitting at the center, a perpendicular ridge at the center might provide additional stiffening without much loss of cooling effectiveness.

I'm no mechanical engineer, so I'm just speculating here, but if the base deforms even a small amount, there could be serious problems when reseating the block.

[Cathar]

Originally posted by Since87
Nice design.

Do you have any concern about the thin baseplate bending under the Block to CPU pressure?

Obviously there is plenty of stiffening along the long axis of the block. Would it make sense to leave a ridge of copper perpendicular to the microchannels. With the water flow splitting at the center, a perpendicular ridge at the center might provide additional stiffening without much loss of cooling effectiveness.

I'm no mechanical engineer, so I'm just speculating here, but if the base deforms even a small amount, there could be serious problems when reseating the block.

No, not concerned at all. The top plate is lapped and sits flush with the tops of the channel walls (which is also flat). Any "bowing" of the centre would only happen if the top plate (6.25mm thick copper) was also "bowing".

It may be a problem if there was no top-plate pushing down onto the tops of the channel walls, but even then you'd probably crush the CPU core first.

[Nick C]

mind if I... imatate this design for meh next block? you've got a winner here!

[Since87]

Originally posted by Cathar
No, not concerned at all. The top plate is lapped and sits flush with the tops of the channel walls (which is also flat). Any "bowing" of the centre would only happen if the top plate (6.25mm thick copper) was also "bowing".

Ahh, I didn't look closely enough. In the larger picture you posted, it looked to me like the fins were lower than the sidewalls. Optical illusion I guess.

Good work. Looks like you've done a good job of balancing the tradeoffs between a lot of complicated issues here. I do the same type of stuff professionally, (in electronics not mechanical/thermo) and I can appreciate the effort that went into this.

[Penguin4x4]

Hmm, maybe you could zip up some CAD files for those across the Pacific Rim, :D:D

[Cathar]

The biggest problem with making this block is that it is very hard to mill pure copper down 5mm with 1mm wide channels.

I've already given a fair amount of information. I didn't use any fancy CAD stuff, I used MS-Paint to draw up the block plans and handed them to a professional machining shop, who then took a fair while to decide just how it could be made. Milling 3mm wide channels isn't that hard. Milling 2mm wide causes people to pause. Milling 1mm wide is dramatically harder. If you want to imitate it, you're welcome. There's no copyright/patent on this stuff, it's all been done before (see Lytron, etc).

I'm trying to chase up a way to get them made affordably, and it's not easy.

[Freeloader]

I wonder if those channels could be cut with a circular type blade? If they could, it would speed up the process a great deal. Have you looked into that at all?

[safemode]

cast it that way. make a cast of of the block in steel or something similar ....and pour your block. Doesn't everyone have a kiln/oven that gets that hot? Gotta love working with molten metals.

[Nick C]

hot enugh to melt copper? at 1035 deg <b>C</b>??

[EgeWorks]

Nice design. I did one very similar to that but homemade with aluminium. Good to see another Aussie putting out some quality stuff.

[JFettig]

heres what i suggest, get it all milled like you want without the micro channels, then take a dremel to it, it can be done neatly and all, i made a block with a dremel and a drill, use fiber reinforced wheels so they dont explode on ya, the other ones work better i guess but they will explode on ya if yoru not careful, like break then fly into peices...

[r0ckstarbob]

incredible work. well done. :burn:

[TheGhengisKhan]

Originally posted by maskedgeek
heres what i suggest, get it all milled like you want without the micro channels, then take a dremel to it, it can be done neatly and all, i made a block with a dremel and a drill, use fiber reinforced wheels so they dont explode on ya, the other ones work better i guess but they will explode on ya if yoru not careful, like break then fly into peices...
LoL, I almost lost an eye to those hard disks on my first attempt at a W/B

[Cathar]

Picked myself up an Asus A7V333 mobo which reads the CPU on-die diode. This allows much better accuracy for taking temperatures than using an in-socket diode like my Abit KR7A has.

I have results for the Silverprop Cyclone range, the Danger Den Maze 3, and the micro-channel concept block. I also modified the concept block to introduce a 5mm wide x 15mm long rectangular nozzle that straddles the channels under the inlet barb to boost water velocity and get water-jet flow impingement down into the critical hot-spot area.

I used an Eheim 1250 pump, and a dual fan heater-core (aka the Big Arse).

Using an AthlonXP @ 1925MHz, 2.15v. Using BurnK7 (CPUBurn) to generate load. Tests were run until temperatures had stabilised for 15 minutes (typically 30 minutes to an hour per test).

The Maze 3 drew in some pretty poor results initially, even getting some system instability. I lapped the base of the Maze 3 and found that it was slightly concave. In fact, I could even see the oval shape outlined in the base early on during the lapping. After lapping, the results for the Maze 3 improved by around 2C, as well as removing stability issues.

I mounted each block 3 times to gather data and reduce the effect of mounting irregularities and took the best results.

Flow Rates

Danger Den Maze 3: 7.0lpm
Silverprop Cyclone 5: 7.5lpm
Silverprop Cyclone 4: 6.7lpm
Silverprop Cyclone 3: 8.0lpm
Micro-channel Concept: 7.5lpm
Micro-channel Concept w/nozzle: 7.0lpm

Results

Radiator Air Intake: 23.0C
Water temperature: 25.5C (radiator fans at 7V)

The CPU on-die temperatures observed were:

Concept Block (w/nozzle): 43.5C
Concept Block: 44.0C
Maze 3: 46.0C
Cyclone 5 (3mm): 46.5C
Cyclone 4 (3mm): 46.5C
Cyclone 3 (6mm): 46.0C

All blocks have been mounted with a fairly extreme amount of pressure. The motherboard begins to flex slightly around the socket with the mounting pressure. I apply WAY more pressure than AMD recommends as kosher, but these organic package CPU's can handle it just fine. The difference between mounting these blocks with some of the flimsy springs that are provided can be more than 1C warmer.

Currently I'm using springs meant for the Pentium 4 and wind them right down to even heights. I'd estimate CPU block-mount pressure would be around the 50lbs (23kgs) mark (AMD recommend 15-25lbs).

Edit: (corrected base-plate thickness on Cyclone 3)

[Colin]

Cathar – A hearty welcome to Overclockers from Mr. Thompson. Your thoughtful expertise in pelts and watercooling should be a nice addition to the forums. Spiffy waterblock too! :D

[JML]

Do you have any plans on testing some other blocks against your own? For example, the Spiral series, TC-4, or (and what I would consider most interesting) MaskedGeek's best block? Of course, being in Australia would you be able to get those blocks if you wanted too?

[r0ckstarbob]

Originally posted by Colin
Cathar – A hearty welcome to Overclockers from Mr. Thompson. Your thoughtful expertise in pelts and watercooling should be a nice addition to the forums. Spiffy waterblock too! :D

cathar cathar cathar... hm, what other forum do i know that name from?

[Freeloader]

It's nice to see some decent block testing finally. I appreciate your effort. Can you post a picture of your custom nozzle. It sounds interesting.

[Cathar]

Originally posted by r0ckstarbob


cathar cathar cathar... hm, what other forum do i know that name from?

Most commonly:

Homeworld (the on-line space combat game)
MadOnion
Overclockers Australia

Less commonly:

pro-cooling
hardocp
amdmb

Anywhere else and it probably isn't me.
BillA has dropped my name a few times here as well.

[wildfrogman]

That micro-channel block looks awesome, what would the performance be if the fins were say like thin copper plate soldered on. Say 1/16th inch wide but like 12 or so in the channel? Maybe find a cheap skived copper fin heatsink, cut it to shape, and grind down the base to thickness needed. Would be pretty neat to have a skived fin waterblock.

[Nico3k]

thats a very nice looking block u got there. i am impressed. performance also looking good. i heard nothing but good things from micro channel blocks.
how much would you sell one for?

[Cathar]

Actually they are one week from being ready.

If you want one, please contact me via email here:

http://forum.oc-forums.com/vb/member.php?s=&action=mailform&userid=18274

and we'll continue correspondence via email. I can get an optional P4 mounting plate made up for you if you want.

I'll supply all mounting bolts and an external 1/2" Y-piece connector as well.

The final blocks looks like the following:

http://www.employees.org/~slf/concept/prod2.jpg

http://www.employees.org/~slf/concept/prod3.jpg

http://www.employees.org/~slf/concept/prod1.jpg

[a1cnolan]

you know ive been looking at this for quite a little while trying to figure out how to make 1mm channels. not much of a mathematician but i came up with that at .040 of an inch. which makes it thinner than almost anywelding rod we have in my shop. i figure i could get some of our .040 rod and file em flat and solder them into place. but that would suck. we don't have any .040 cutters smallest size is 3/16 or rougly .187 of an inch. which is too big. the only thing i can come up with is that they use a slitting wheel chucked up on a milling machine. which also are very expensive to get that thin.. thank god i have a surface grinder at work. so ill just take one of our thicker ones and make it thinner. that idea about the grinding wheel is nice but grinding wheels do two things that are not helpfull to machine such small areas. they flex and they wear.. i always here people say they blow up on them.. i ve never had that but the other two are enough for me to throw that out cause if im gonna copy someones block im gonna do it accurately. and dremels and accurate do not go together... anyway... i like your design cathar. i think im gonna work up a mix between yours and that S masked block thing that other guy has. only issue ive come up with now is if i should have the water flow down with gravity or up with the flow of the heat. seems like dragging the heat down would be antiproducitve but then pushing the water upwards would also be counterproductive to flow rates. im not an engineer i just make the stuff you draw for me...lol anyway way to go..

[utabintarbo]

Originally posted by a1cnolan
[B]you know ive been looking at this for quite a little while trying to figure out how to make 1mm channels.

...B]

CNC'd with 1mm carbide cutter, I assume.

Bob

[JFettig]

I do say you do a diff design than my type-s, because for this theres tons of better designs..

[a1cnolan]

CNC'd with 1mm carbide cutter, I assume.

well the thing is most shops either a) don't carry cutters that small or b)wouldn't stick it on a cnc because carbide cutters that small are very easily breakable and flexible. so i thought that would probably use a different method. either that or they just really must have taken thier time on those cuts. Im not saying they couldn't do it im just saying it would a be a pain im not willing to undertake...

I do say you do a diff design than my type-s, because for this theres tons of better designs..

well i looked at your block and i thought. it looks good. and water would flow very well through it. only thing i didn't like was the relatively small amount of surface area around the core area. i was thinking of either putting some channels in over the core areas or or using a drill and making those drill marks a lot of people make when they just chain drill out their block. im still deciding. just messing with ideas in featurecam my shops letting me do my blocks for training... the number one thing i took away from cathars design is the o-ring seal. i just don't know why that never popped in my head before

[Overbrazil]

very nice block.. I have read at procooling. put in the gallery. Just a question. Did you tested in the KR7 ? i know that that board dont read the xp diode, but if you have this information....
[]´s

[Cathar]

Originally posted by Overbrazil
very nice block.. I have read at procooling. put in the gallery. Just a question. Did you tested in the KR7 ? i know that that board dont read the xp diode, but if you have this information....
[]´s

Yes, I have tested on the KR7A. The results there were slightly better than on the Asus A7V333, then again, the KR7A allowed me to overclock the CPU further with a higher voltage, so the KR7A would've meant that the CPU was dumping even more heat. I was seeing around a 6C improvement over the properly lapped and mounted Maze 3/Cyclone 5, with the CPU at 2024MHz/2.15v, and was able to get the CPU stable at 2046MHz/2.15v with the micro-channel block, whereas before that was not possible.

[TonyMc]

Hey Cathar. I'm a little confused here. I was under the impression from comments you made here (http://xtreme.routehero.com/forums/showthread.php?s=&threadid=4869), and here (http://forum.oc-forums.com/vb/showthread.php?s=&threadid=142825). That reading the temp off a motherboard probe was not a proper way to read temps and a "hobbie horse I like to ride", but yet here (http://xtreme.routehero.com/forums/showthread.php?s=&threadid=4964), and in this thread you clearly state you are getting your temp info off your motherboard. I'm not trying to be an a** or anything I would just like some clarification as to the proper way to read and report temps.

[Cathar]

FRAGN'STEIN, my comments about motherboard temps were always about reporting absolutes. I don't believe in the in-socket thermal probes to report anything resembling accuracy for an absolute temperature. I also don't believe in the on-die thermal probes to report an accurate absolute value. I do believe in the in-socket probes to give a rough indication that something is cooling better than something else but prone to a large margin of error (+/- 3C), and the on-die thermal probes to give a fairly good indication that something is cooling better than something else within a tighter margin of error (+/- 1C).

I don't believe in stating absolute temperatures and using those absolutes as guidelines for which many people can judge the performance of their systems relative to each other. I do believe that a singular system when very used under very carefully controlled circumstances can be used to make comparisons between multiple blocks and arrive at a fairly good level of confidence within a +/- 1C margin of error, and that's the only way that a motherboard probe can be used to make comparisons. That is to say I have some level of confidence that a particular system can be used to make decent relative measurements of performance, not absolutes.

I don't report my actual motherboard temperatures, meaning the actual value, because I believe it's misleading, unless someone specifically asks me for them and I've had a few whiskey's and reason temporarily slips past me. ;)

It's all relative...

[TonyMc]

Originally posted by Cathar
unless someone specifically asks me for them and I've had a few whiskey's and reason temporarily slips past me. ;)

It's all relative...

Yeah I've had that syndrome myself a couple times;) :D Thanks for clearing that up.:)

[Cathar]

Originally posted by utabintarbo


CNC'd with 1mm carbide cutter, I assume.

Bob

Correct:

http://www.employees.org/~slf/concept/machine1.jpg

http://www.employees.org/~slf/concept/machine2.jpg

[Cathar]

...and some more recent piccies of the blocks as they appear before being packaged and leaving the door:

http://www.employees.org/~slf/concept/shiny1.jpg

http://www.employees.org/~slf/concept/shiny2.jpg

http://www.employees.org/~slf/concept/shiny3.jpg

http://www.employees.org/~slf/concept/shiny4.jpg

[Liss]

Originally posted by Cathar
Actually they are one week from being ready.

If you want one, please contact me via email here:

http://forum.oc-forums.com/vb/member.php?s=&action=mailform&userid=18274

and we'll continue correspondence via email. I can get an optional P4 mounting plate made up for you if you want.

I'll supply all mounting bolts and an external 1/2" Y-piece connector as well.

The final blocks looks like the following:

http://www.employees.org/~slf/concept/prod2.jpg

http://www.employees.org/~slf/concept/prod3.jpg

http://www.employees.org/~slf/concept/prod1.jpg

i do'nt get wut the middle plate is for?????????

[redPEPPER]

The key part of the middle plate is the slit in the middle. Its size is calculated to spread water evenly to all the channels below.

[Cathar]

Originally posted by Liss


i do'nt get wut the middle plate is for?????????

It's a nozzle plate. It restricts the opening from the top plate and forces the water through a narrow opening. As such, the water is accelerated to about 3x the velocity of when it enters the barbs. The location of the nozzle directs that rapidly moving jet of water down pretty much directly over where most of heat is concentrated and it impinges the base causing a very high (compared to most blocks) region of thermal convection. The accelerated water is also flowing rapidly down the sides of the channel walls, creating a very intense region of high thermal convection and turbulence right where it needs it the most. Without the middle plate, the block performs around 2-3C worse.

[Liss]

how wide is the slit?
just by looking at it. it looks like it restricts flow... ????

[Cathar]

Originally posted by Liss
how wide is the slit?
just by looking at it. it looks like it restricts flow... ????

Is it always bad to restrict flow?

The block is performing between 4-6C better on AMD systems than any other block, and from 3-5C better on P4's for people who are using it, depending on the overclock. The higher the overclock, the better the results being reported.

http://forum.oc-forums.com/vb/showthread.php?s=&threadid=138604

and

http://forums.overclockers.com.au/showthread.php?s=&threadid=109948

For an Eheim 1250, and a heatercore, a Maze 3 saw 8.0lpm through it, while the White Water (the name of my block) saw 6.7lpm through it.

Yes, it restricts flow, but does so where it does the most good, and doesn't really restrict flow that much.

[JasonKosi]

Is that milk that you are using to cool the bit/project or is it just water with bits of slurry in it?

[Cathar]

Originally posted by JasonKosi
Is that milk that you are using to cool the bit/project or is it just water with bits of slurry in it?

It's a machining lubricant/coolant. It just happens to look milky.

[JML]

Cathar, do you think there are other WBs out there that would benefit from jet impingment (sp)

[Cathar]

Originally posted by JML
Cathar, do you think there are other WBs out there that would benefit from jet impingment (sp)

Every block can benefit from it to varying degrees. It's the reason why the 3/8" Swiftech's outperform the 1/2" models. The narrower inlet of the barb makes the water flow faster before striking the base right above the CPU.

Basically stick an inlet above where the CPU centre is, and restrict that inlet in some way to get the water moving faster before it hits the base of the waterblock.

The problem with jetting is that to really do it well requires higher pressure pumps than what most people use. I optimised my block's nozzle plate around the 5'-10' pressure head pumps that most people use. You can't just keep on making the hole smaller and hoping for the water to keep getting faster. Eventually you reach a point where the pump simply cannot supply the required pressure and the water velocity goes up much slower than the volumetric flow rates drop off, and so cooling performance also starts to drop off again. This is what I mean by optimising for certain pressured pumps. With a higher pressure pump I could've made the slit even smaller and seen even better cooling performance. The same principles can be applied to any block with an inlet over the CPU.

One day I'll grab an Iwaki MD-30 and develop an optimised nozzle plate for it, and then I'll be able to offer the block in regular, and high-pressure performance variations.

[JFettig]

wow I missed those extra pictures!, thoser are looking insanely nice!!! looking nicer and nicer everytime i see them! thats a lot of lubricant! holy crap!

I do see you set up the jet so its offset from the center like the core is, how far is it offset? do you know exactly? I was trying to measure.... didnt get too far, I measured 1/16inch

[JFettig]

how long does it take to machine those? and how deep do they go with the 1mm cutter per pass? I really doubt they take the entire thing all at once...
you should try multiple 1mm wide jets! how would that work? I think it would spread out the jet a little bit so it covers the entire core and would maybe give higher velocity?

[Cathar]

Originally posted by JFettig
how long does it take to machine those? and how deep do they go with the 1mm cutter per pass? I really doubt they take the entire thing all at once...
you should try multiple 1mm wide jets! how would that work? I think it would spread out the jet a little bit so it covers the entire core and would maybe give higher velocity?

The time it takes is reflected in the price.

By multiple 1mm wide jets, do you mean multiple slits? If so, I've tried that. Performance is worse. Because there's so much flow and not enough pressure, the multiple jets end up interfering with each other.

[JFettig]

ooh i see

yes i know how the time reflects the price, How much time does it take to machine one of those? It took 11minutes to do just the channel in the type s blocks I beleve.

[Cathar]

Originally posted by JFettig
ooh i see

yes i know how the time reflects the price, How much time does it take to machine one of those? It took 11minutes to do just the channel in the type s blocks I beleve.

I've never timed it as the machinists do it for me.

[LiquidSquirt]

JFettig...
By my measurements, the core is offset 1.5mm(0.059").

[JFettig]

thanx man! i was really close on my 1/16inch measurement, (.0625)

[JFettig]

cathar, are those pics taken with diff cameras or did you do something? one looks pink and one looks like it was made of gold:D

[Cathar]

Originally posted by JFettig
cathar, are those pics taken with diff cameras or did you do something? one looks pink and one looks like it was made of gold:D

Same camera. The initial pictures were taken of a block that was pretty much fresh from the machinists, and the only thing done to it was that the base was lapped. The plates do indeed have a pinkish sheen to them, which I suspect is a combination of the extrusion process by which the plates are made, and the machining coolant.

When they are sanded down to remove the surface pinkish layer and polished up the true copper color comes through. The blocks look copper color to me in the pictures, not gold.

[Frodo Baggins]

wow, real pretty blocks ya got there cathaar. I hope business is going good for you

[JFettig]

cathar! I gots idea!!! increase velocity! take a larger ball end mill to that slit, make it so its the width of the hole you want, but itll increase flow!, Ill do some autocad work real quickly..

[JFettig]

maximize flow and velocity
http://www.theforumisdown.com/uploadfiles/1102/JET.jpg

[Iron Hawk]

Yea, JFettig has a good Idea.

[Albigger]

JFettig -

first my disclaimer: i know little to nothing about fluid dynamics, but here's what my reasoning tells me (i.e. I may be totally wrong)

it looks like that design will create either

1) a lot of stagnant water around the middle of the jet

OR

2) a lot of water coming out of the jet at a more horizontal direction, instead of vertically toward the base. Would this not then decrease (some) impact velocity of the water on the base? I'm not sure that would be too great....

It would be less restrictive for flow however.

[Cathar]

JFettig, nozzles work best with totally the opposite curve to what you have pictured. The shape shown would result in a wide "spray" that is largely unfocussed and would actually restrict flow since it introduces section of near horizontal flows at the exit colliding with each other and into the stream that's trying to come out the middle of the exit hole, rather than forcing the water to all flow in one focussed direction.

Think of a classical "black hole" style vortex shape. That's the optimal inner shape for a nozzle chamber, the exact dimensions of which are defined by numerous engineering bodies, the most prominent of which is ASME in the USA. Take a bit of a poke around with google and you'll find plenty of information about.

While my nozzle plate arrangement doesn't really come close to being a proper AMSE style nozzle, it's a "good enough" appromixation that's easily machined using a chamfering end mill.

I experimented with different chamfer depths and the depth used provided the "straightest" flow directly down onto the base-plate, which was especially important given that the nozzle slit is off-centre.

[a1cnolan]

ive been a machinist for almost three years but you guys are freaking wow. i mean i need to start taking classes. i hope we finish with iraq soon otherwise ill never get around to school.

all im saying is you guys are really really really thinking about this and i was just thinking. whatever that means