Ultimate block? Theory.

[V12|V12]

Well FFFFFFFFFFF UK! Damn this logging in! I had a great LONG reply, but even though I was logged in, it made me do it again and there went everything!

SO a summary.... The thicker the base-plate the more heat will be asorbed from the core= better temp buffer, BUT flowrate should be slower...all that heat has to be absorbed by the incoming water and that takes some time (not long in human standards).

Thin base-plates suffer from faster "heat overload" where the thinness prevents more heat from being absorbed, so a faster flow rate with the most amount of cooled water is need to keep the heat from being transfered back to the core.

A graph of Thermal-spread-resistance VS Thinkness would be nice! But I don't have the time

[murray13]

I guess what I was trying to get at is that the thicker the copper plate is the better it will spread out the heat so that there is more surface area to create a higher delta T with the water.

There will be a upper limit for the thickness but I have never seen a waterblock with a half inch of copper at the bottom just to spread the heat out. If someone knows of one I'd like to see the results.

The radiator example was just for comparison.

[YellaT]

Im full on noob to this whole watercooling idea but....
it seems that your just placeing the water on top of the processer, not having any water hit it at all. what if you actually put the processing part on water ( i know you don't use water, but there is some ethonal type thing i saw on the screen savers). you could have the same flowing idea, but rather have one channel of water rushing over the processer, so it is constantly flowing in cold water (ethonal, whatever). it just seems like it would cool better if there were ACTUAL water flowing over the processer itself.

whateves, just an idea!

[bigben2k]

That's direct die cooling.

Unfortunately, it doesn't work better than a waterblock. The heat load is just too great for the water to absorb, plus, the kind of flow rates required would make mounting such a block near impossible, or at least certainly unsafe, in terms of leaks.

So we need some kind of baseplate.

"A graph of Thermal-spread-resistance VS Thickness would be nice" I agree. I think BillA and Joe should look into this, since they're the ones with a die simulator.

[Epion2985]

>Well FFFFFFFFFFF UK! Damn this logging in! I had a great LONG reply, but even though I was logged in, it made me do it again and there went everything!


heh, yeah, what you do is Ctrl+C your post before you post it ^_^ just in case, saves my ass all the time

[Auraka]

Sorry for using paint, no CAD stuff here.

Has this design been discussed, direct impingement with flow control. Jet could be positioned closer for lower flow/velocity pump situations.

EDIT: Sorry, having problems getting the image to show and it's getting late

[BillA]

Les posted this link
and this OCAU thread is now into nozzles

I have done a bit of nozzle testing but always it returns to the same issue - pump head capability

what's the point of a fancy wb used with a cheap pump ?
just match that pump with a 'good enough' wb and save the money for a window

[bigben2k]

I agree there BillA. The cheap solution I found was to use more than one cheap pump in series.

I'll read over the nozzle issue in a bit.

Auraka, Morphling is working on something similar. IMO, there is a better way, but we'll see.

[bigben2k]

Originally posted by unregistered
Les posted this link
and this OCAU thread is now into nozzles

I have done a bit of nozzle testing but always it returns to the same issue - pump head capability

Morphling tried a nozzle, but it turns out that it was too restrictive. Here's his thread

I suggested using standard ASME nozzle designs, on page 2 (never mind the MS Paint hack job...) (I realize by now that these may have been meant to shoot water in air, not water in water).

Aren't Les' numbers for GAS flow, instead of liquid flow? Seems to me that this V shape nozzle isn't optimal for a liquid.

[morphling1]

Well I haven't post any new block design in a long time, and that's because I'm busy making few new designs.
Yes my design is roughly similar to Auraka, but I wanted to control velocity of the water to be as constant as posible across the block, and not like that design where inside is open and volume is a lot bigger then mine.
I'll be getting two comercial blocks in two days (Swiftech and Magnum) and I'll close myself in the room for few days and make one roundup with four of my designs, I need good data to see where I'm heading with my design, I'll also try two exactly the same block just base thickness will be different (6 & 4mm) so in a week or so expect results.

[Cathar]

Originally posted by unregistered

what's the point of a fancy wb used with a cheap pump ?
just match that pump with a 'good enough' wb and save the money for a window

If said wb costs $55 USD and the performance is still better than anything else commercial with a $60 pump, and better again with a high-performance pump, then this makes it a very valid point to do so.

If you want to sell blocks you don't want to be asking people to buy $500 pumps, or attempt to scrounge a cheap one off Ebay, just to make the block perform. If a design is optimised for high pressure pumps but sucks with low pressure pumps, no-one will buy it. If a design straddles the two extremes and offers near optimal performance but with a low pressure pump, but is still able to scale its performance up with a high pressure pump and still perform just 0.5C worse than a high-pressure focussed block, then that's a fair compromise.

More people buy BMW M3's than Ferrari Spiders, 'cos people are happy to pay 1/4 the price for a car that performs 90% as well but is far more practical in the real world.

At least, that's how I view it.

[Volenti]

I tend to agree with cathar on that, I've been re-focusing my water block designs away from "absolute" performance at any cost toward ones that will still perform well at low flow rates with small (600L/H .5-1m head) pumps.

[BillA]

read back Cathar

that comment was a facetious addendum to a post about nozzles,
which to optimize WILL require the matching of the nozzle to the pump/system P-Q characteristics

do try to bear in mind that at this stage of your wb's development you have not yet quantified so much,
your claims of ~3°C improvement notwithstanding

3°C is worth 'only' 25MHz OC ? (perhaps I'm misquoting you ?)
nooo, I don't think so, but I'm not an OCer either
3°C lower CPU temp is a REALLY BIG DEAL over a fairly top-flight wb like the Maze3

and I do not believe it
get your temp measurement capability in order and come up with some directly comparative results
I will, and do, predict that you will find the difference far lower

?
you going to generate a curve ?
or you intend on equalizing the flow how ?
or perhaps it will be a 'in system' test whose results are predicated on the asstd components used ?

testing is a can of worms
but you are the one making the claim, so its your burden

EDIT: I am not arguing the practicality aspect of utilizing aquarium pumps,
but this thread did say "Ultimate", no ?

[Cathar]

but this thread did say "Ultimate", no ?

Before we go further, I wasn't the one who referenced my design in this thread. I made no claim to have "the Ultimate". I just noticed that my block was referenced here and was explaining the design decisions since they seemed to be in question. I saw this thread a while ago and for the exact reasons I explained above, I wasn't going to offer my design up for analysis.

As for temperature data, it's under-way.

I've lapped the Maze 3 base. It was indeed slightly concavely bowed in the middle effectively creating a small cavity above the core the CPU. After lapping its performance improved by 2C. It's now performing much better.

[BillA]

I see said the blind man as he picked up his cane and walked - peace bro

re the M3; kinda as expected, no ?
and they come closer together

am most interested in your results (will follow on OCAU)

[Cathar]

Originally posted by unregistered
re the M3; kinda as expected, no ?
and they come closer together

Well, since we've already broached the topic here...

Yes, the Maze 3 surprised me with how non-flat it was, despite being extremely smooth. In fact, early on during the lapping I could see the Maze 3's oval shape outline where the sandpaper had not reached yet on the base!

So yes, the Maze 3 picked up a few C which places it very slightly ahead of the Cyclone 5 (by 0.5C). The micro-channel block however is still a full 2C clear of the Maze 3. Interesting as on the Abit KR7A mobo with its dodgy on-board thermal probe and through the use ot measuring overclock stability by varying water temperatures I was able to estimate that the micro-channel block was doing around 2-3C better than the Cyclone 5, and now with being able to read the on-die diode I find that broad estimate was pretty much bang on.

However, since the design is pulling a clear lead over both the Cyclone 5 and the Maze 3, it may be worth debating its good points in the hunt for "the Ultimate", or at least find a way to discount it in place of a design that tackles the problem in a clearly superior way.

[bigben2k]

It is definitely a good design.

It's a shame that the channels had to be so wide. I'm sure it would have been interesting to test this with tighter fins.

Optionally, one could braze very thin strips of copper, in that design.

There is an ideal fin spacing and thickness, as well as height. It's calculable, but it's not obvious.

Good work.

BTW, (I asked in another thread), did you consider running the test from side to side while plugging the middle barb? I believe that the middle barb, if it wasn't completely plugged, could alter the results. If you could replace the middle barb with a plug, that would be ideal.

[Jessfm]

Freakly
I have been testign with a block similar to Cathars, but with wider fins. ( My R&D is for OCPC as company if anyone is unsure).
Beta2 uses 3mm fins with 2mm channels ( machine tooling limitations).
I would love to drop lower. But so far we have 2 x 5mm channel size.
Strange thing is this was R&D for a budget block. From your similar design you seem to be getting better results with thinner Fins.
Our limit would be 2mm fin 2mm Channel I may approach this for beta5, and this could result in a very compact design.
*Edit*
Our design is side-side , no plans on the Central injection/twin outlet.

[Cathar]

Originally posted by bigben2k
It is definitely a good design.

It's a shame that the channels had to be so wide. I'm sure it would have been interesting to test this with tighter fins.

Optionally, one could braze very thin strips of copper, in that design.

There is an ideal fin spacing and thickness, as well as height. It's calculable, but it's not obvious.

Good work.

BTW, (I asked in another thread), did you consider running the test from side to side while plugging the middle barb? I believe that the middle barb, if it wasn't completely plugged, could alter the results. If you could replace the middle barb with a plug, that would be ideal.

So wide?! 1mm? Some people are never satisfied! Micro-channel theory basically calls for 0.7mm-1.3mm wide channels. I'm smack bang in the middle 'cos unless I went to EDM, it's not really feasible to get the channels any thinner with conventional machining. Much more expensive to get into EDM territory.

Near as I can tell though with my simulator, 0.8-0.9mm wide fins with 0.6mm wide channels is where it's at (optimal). We can go smaller but get to the point of rapidly diminishing returns. Would expect about a 0.5C improvement in that config over my current 1.0/1.0 config. Optimal channel height is dependent upon the water velocity, which of course depends on the number of channels and flow rate. Have to juggle a few balls here to figure it out, but basically 4-6mm channel height is a good choice for a "general" solution, possibly dropping down to 3mm height and boosting pump pressure and flows to get the water whipping through, which will really pick up its performance.

Currently the block is a result of a set of trade-offs to work with conventional pumps and attempts to explore the limits of what's possible in most people's water-cooling setup when all we do is change the block. The design can be refined to more closely approach "the ultimate" using the above guidelines. Whack on a cross-sectional nozzle under the inlet barb to get good jet flows impinging down into the hot sections too, and we'd start to see something significantly better (another 1-2C again), but it all hinges on the pump.

That's my theory on "The Ultimate". Oh, I forgot, we'd have to work up an optimal base-plate thickness for each die size too. The actual target heat load will also impact various aspects of the design such as base-plate thickness and channel height. "The Ultimate" is a highly strung piece of work carefully crafted to do its job for only a specific application. Throw it at a different application and it will cease to be "The Ultimate".

Back to your other question:

I tried the side-to-side option and will do so again, and plug up the central barb properly with some blu-tak (putty).

I'll give it a burl tomorrow.

[bigben2k]

Thanks Cathar, and again, excellent work!

Did you calculate any improvement in increasing the fin height?

I recoded the sim. Turns out that a 1.5:1 ratio of copper:water width is ever so slightly better, so using say 0.9mm wide walls with 0.6mm wide water channels would yield slightly better results, and would also increase the essential copper density right above the CPU core. However, just try getting something like that machined! (EDM would be the only option) I went with a 1:1 ratio and 1mm anyways because that was the smallest that the machine shop could do cheaply, and I was also concerned about flow rates. Having said that though, a 1:1 ratio still performs very well on the simulator not being much different (0.15C) to the 0.9:0.6 option.

[Auraka]

That plan was something I threw into paint in only a few minutes, block volume and in fact all design specifics were secondary. Mostly I was interested in a star shape of fins directly above the core to increase surface area and direct water flow. Fins could feature radius corners where they meet the base. This design would eliminate jet dead spots due to it splitting the flow into multiple parts, i.e. the core of the star would be where the dead spot would normally form. This star shape I think would lend itself better to a circular interior anyways, with dual opposing outlets from the main chamber channeled into a single outlet barb.

What I see lacking in current commercial block design is sufficient, effective surface area and I think this design would address this in a direct impingement style block. The downfall (aside from unproven results) is machining complexity, and might even prove more economical to cast if it reached volume production.

BTW cylindrical smooth bores like the interior of a bard fitting are considered nozzles, at least in the fire service. Go to your local station and have a look at a deck gun on top the engine to get an idea of what I mean. Of course that doesn’t mean your average garden/hobby pump will be able to generate enough pressure and volume to produce the desired stream.

I’ll be watching morphling’s work since he’s been developing a similar concept.

[bigben2k]

I'm gonna cut you off Auraka, to put up a link to another, very exciting thread!

Here it is!

(Big thanks to BillA, for providing
this link! )

[Cathar]

Originally posted by bigben2k
Did you calculate any improvement in increasing the fin height?

Yes. A higher fin height works better for lower flow rates. As we lower the fin height, we lower the cross-sectional water area, and this correspondingly boosts water velocity with the caveat of requiring a higher pressure pump to get adequate returns for doing so.

Les had an interesting series of graphs over at OCAU that plotted pump pressure vs water flow vs performance for some different wall/channel thicknesses/heights and the graphs were roughly showing that the differences are minimal. As we boost the pump pressure for a "tighter" block to bring its performance up, the block doesn't perform that much better if the same pump pressure is applied to a block of "more open" proportions.

So it seems that no matter what, it hinges on the pump. We can build a block that needs a high pressure pump to perform, but will it perform better than a freer flowing block when the same pump is attached? I suppose that then becomes a factor of the pump's rated flow rate at various pressures. Basically we need positive displacement pumps to guarantee our flow rates.

At this level we all do tend to get rather muddy on what will work and what won't in practise.

I'm just at the stage of where I can safely say that micro-channels is one way to achieve better than flat-plate/wide-channeled designs which all perform about the same, but predicting what can be done around the micro-channel concept to push it further is still definately experimental.

[bigben2k]

Originally posted by Cathar
Yes. A higher fin height works better for lower flow rates. As we lower the fin height, we lower the cross-sectional water area, and this correspondingly boosts water velocity with the caveat of requiring a higher pressure pump to get adequate returns for doing so.

Les had an interesting series of graphs over at OCAU that plotted pump pressure vs water flow vs performance for some different wall/channel thicknesses/heights and the graphs were roughly showing that the differences are minimal. As we boost the pump pressure for a "tighter" block to bring its performance up, the block doesn't perform that much better if the same pump pressure is applied to a block of "more open" proportions.

So it seems that no matter what, it hinges on the pump. We can build a block that needs a high pressure pump to perform, but will it perform better than a freer flowing block when the same pump is attached? I suppose that then becomes a factor of the pump's rated flow rate at various pressures. Basically we need positive displacement pumps to guarantee our flow rates.

At this level we all do tend to get rather muddy on what will work and what won't in practise.

I'm just at the stage of where I can safely say that micro-channels is one way to achieve better than flat-plate/wide-channeled designs which all perform about the same, but predicting what can be done around the micro-channel concept to push it further is still definately experimental.

Some good ideas. I'll ponder this for quite some time!

What channel/fin height would you recommend for a flow rate in the range of 100 to 150 gph? (375 to 550 L/h) Would a 1.5:1 ratio of copper/water be relevant?

[Les]

Maybe relevant to channel/fin dimensions,pump head, and cooling performance considerations..
A couple of my thereotical designs for maximising cooling from the "deliverable pump head" in another thread.::
http://forums.procooling.com/vbb/sho...9&pagenumber=4

Although these are thereotical designs for 50x50mm Peltiers they scale down to 17mm designs with 30fins.