In the quest for a ultra high-end system with xtreme cooling, I was wondering how does the ultimate waterblock look like?!, provided the design is feasable and reasonable.

We all agree on the use of copper, since it is the best material our budget can buy. But I was wondering if the block should go up? Should it be wider? Is bigger always better? Would a ThermalRight SP94 make a good base for the ultimate waterblock?


Your thoughts on this one please, and a good chance you'll see one of your ideas installed on my CPU soon

Welcome to SPCR!

Much as I hate to say it, this is probably the wrong place to ask, much as I hate to say it. Somewhere like ProCooling would be better, as this sort of thing gets discussed there frequently. Another thing to look up would be the work of Cathar in Australia. He came up with the WhiteWater block that everyone is currently busy copying, as well as the Cascade which is probably the best block currently out there.

This forum on the other hand is concentrated on making watercooling quieter, rather than higher performance.

Currently the absolute ULTIMATE WC block is the *Sterling Silver* Cascade, by Cathar. Next best is the Copper Cascade, also by Cathar. See ProCooling for tons of discusson on both.

Gooserider

As goose mentioned, the SS cascade is currently the best and is custom built by a fellow named Cathar in Australia. They aren't cheap and the lead time is several months (or more).

I think your question really is "what is the best performing block that I can build without a CNC machine shop?" For that I would recommend a design by 3rotor sometimes called #rotor. It is a pin-fin block made only using a drill press and dremel tool. Search at procooling.com or ocforums.com for discussion of the design.

you can't get the cascade (ss) anymore. and in fact, it isn't the best waterblock anymore, since cathar has now been pipped by... cathar himself he's almost finished a new design called the cascade xs which has an unbelieveable 149 jets/cups (compared to the cascade's 52).

it'll be an expensive beasty though, although he intends to scale it down a little to make it more affordable.

For instance, the best waterblock, in terrms of best CPU-to-water heat transfer may not be the best block if you want to use a lower flow/pressure pump. For instance, from this list, I recently selected the Swiftech 5002 http://www.cooltechnica.com/Merchant2/merchant.mvc?Screen=PROD&Product_Code=SMCW5002-AMD&Category_Code=WB as it was, IMHO the best cooling-for-pressure-drop trade-off.
You're asking this question on a silentPCreview forum, so I'm making the assumption that you care about pump noise.
Now, if you're watercooling a CPU only you really don't need more than a Eheim 1046 (very quiet) if you pick a low restriction waterblock.
Have a look at http://www.overclockers.com/articles373/wbsum.asp. A lower C/W is better - and (assuming a very modest but also very quiet pump like the 1046), then lower pressure drop is better, too.
FWIW, if you want to go yet quieter, then IMHO, get the new Eheim HPPS 12V pump http://www.highspeedpc.com/Merchant2/merchant.mv?Screen=PROD&Store_Code=HSPC&Product_Code=HPPSDC&Category_Code=Pump. I've just one installed one and it's as close to a silent pump as I've ever heard (not heard?).
The final part of your watercooling system is the radiator/fan. I'm currently getting decent results with a BlackIcePro (which is lower air restriction than the BlackIceXtrreme) and a 120mm Panaflo 'L' series undervolted to 5V. The Panaflo starts fine at 5V and is very quiet.
I'm about to try a chevette heater core http://www.aquastealth.com/index.asp?PageAction=VIEWPROD&ProdID=25 with a pre-made shroud http://www.aquastealth.com/index.asp?PageAction=VIEWPROD&ProdID=107 and the same fan - still undervolted to 5V. I expect to get somewhat better cooling for the same very low noise level, but "don't know 'till I try it...".
Bob

bobkoure: at low flow rates, the caascade is still the best waterblock. have a look at this graph:



and you can see that the cascade at 0.5GPM still betters the mcw5002 at 2.5GPM. (and obviously there wouldn't be that much pressure drop).

a note about the eheim 1046 - innovatek in germany sell a kit which converts the AC version of it to 12v DC and a modified impeller, which boosts its performance up to above that of the 1048. worth looking at http://www.innovatek.de

edit: correction, the graph shows the mcw5000. but even the mcw5002 isn't so much better than the mcw5000 to change what i said above

and just for the sake of it - i'm currently using an mcw5002

chylld-
hmmm... got a link to where-ever that graph came from?
I'm guessing that "Delta T(CPU-wb in)" is the temp above coolant that a CPU might show.
I'm also assuming that the coolant temp is unvarying and that it's not actually a CPU, but a CPU-equivalent resistor pack (and how much wattage is being shed?), and likewise, something more accurate than a CPU diode for temp readings.
I'd like to read the whole thing, please...?

Anyway, I don't dispute the cascade being better than the 5002 in many/most situations - was looking at the far low end of the graphs in Joe Citarella's article, as quiet pumps like th 1046 are also at the low end. BTW, the 5002/5000rev2 is supposed to be about .02 C/W better than the 5000rev1 - which'd be about 2C cooler on the chart you posted if the resistor pack was 100W. Don't have enough info to be able to speculate beyond this.

I know about the 12V Innovatek/Eheim 1046 - I'm using one of the "new - 12V throughout" models - no high voltage in it (don't know if the adapter kits actually inverted 12V up to 220 as we never saw them here in the land to 60Hz, but get the impression that they did based on the blurbs about the new models saying that they dont do this anymore)
Anyway, yes, they're quieter than a 1046 and have more pressure as well. You haven't seen a flow/head performance curve for this pump anywhere, have you?

Of course, IMHO, all of this is marginal compared to the C/W of whatever's being used to dump heat into room air (or wherever). I suspect that there is a lot more C/W variability between, say a Black Ice radiator and a reasonably sized heater core - even if both have the same amount of air being pulled through 'em.

the graph came from here. might have to fish around a bit to find the exact testing info you want.

i haven't seen any graphs of the mdoded 1046 nope all i know is that the pressure head is higher (according to the innovatek website) than the 1048's.

(getting slightly OT here) yep a really good heatercore is fundamental to watercooling success - an entry-level one (BIP, BIX) you're not getting much performance out of the waterblocks. when i switched from an internally-mounted BIX to an externally mounted BA, i got a 20c drop in load temperatures, at a higher overclock.

Thanks for the link, I'll check it out tomorrow (bedtime on this side of the planet - yawn).
20C!! I'm running 32-43 idle-100%load, ambient is 20. I'm not expecting to drop to 12-23
Oh - yeah I'm overclocked, too - Athalon 2600+ XP-mobile at 12.5x203MHz 1.75V.

This has gotten interesting - but we're pretty far off from the original topic of How does the ultimate waterblock look like?!
I'm wondering if it makes sense to move a lot of this conversation into a new thread.
My own personal bias would be to call it something like "Living on the edge of semi-adequate cooling for the sale of low noise - silent PC watercooling considerations" - oh wait - that'd run off most everyone's screen - but you get the idea.

As a "for instance", I run 120mm Panaflo L1A fans. They start out OK noise-wise and get pretty quiet once I undervolt 'em to 5V
If you go to Panasonic FBA12G performance curves you can see that the rated performance of the FBA12G12L1A (the fan I'm talking about) is at 13.8V - so at 12V it's already undervolted.
Anyway, if you look at the curves for this fan, the 7V performance is less than half the 13.8V performance - so it's a pretty good bet that 5V performance (even though it's "off the chart") is less than 5/14th the rated performance.
So... 13.8V rating is 68.9CFM and 3.30mmH2O - and I know I've got less than 5/14th of that or 24CFM and 1.18mmH2O (.046inH2O)

So now, when I go to Radiator Heat Dissipation Testing The author (Bill Adams) says "...reasonable low and high speed/thick bodied axial fan applications (0.015 and 0.05 in. H2O) resulting typically in 75% or so of “rated” output..." (emphasis mine), so, hmmm reduce my WAG by 25% to 18CFM and .0345inH2O.
The amazing thing (to me) here is that I'm not "off the bottom of the scale" - which is what I expected when I started this exercise in guesswork.
Could be I did something obviously wrong, though (feel free to correct me...)

good point. here is what an early prototype of the ultimate waterblock looks like:



a close up of the nozzles in the middle layer:



before cathar went into production with these blocks, he made some changes, i.e. converted it into a 1in-1out arrangement (as the 2nd outlet made no performance difference) and he increased the number of jets and cups by 20%.

there are 52 jets/holes in the production version (which is out of production now). he's currently testing a version with 149 jets/holes. that's right - one hundred and forty nine. all within the area of a cpu die.

Wow...
Are those actually 6-sided protrusions glued on (or milled away between)?
I totally understand why these are 3X the price of waterblocks designed with an eye to manufacturing costs (surprised they aren't more).
EDIT - nevermind the questions - found 'em all answered http://forums.overclockers.com.au/showt ... did=169230

amazing isn't it

the width of the jet walls is appoaching the width of human hair.

the beauty of this design is that it isn't easy to copy, even if you have access to high-end cnc machines. there's an extraordinary amount of maths behind it and so many variables that if you get it even slightly wrong (e.g. 0.1mm) you're going to get crap results.