The pinnacle of performance? [Archive]

Cathar

07-09-2007, 09:59 PM

So tricky to answer.

It helps to assess designs on the merits of their effective convectional co-efficient per unit area vs the hydraulic power required for the design to achieve that performance. If there was ever a way to determine "what is the pinnacle", that is the way to do it.

It does get clouded by implementation details, and CPU's change, IHS's change, mounting mechanisms change, and there's a mass of other factors that serve to affect the final outcome. All those factors can be controlled, but they are specific to each CPU type, and it's why some blocks that used to be the performance leaders don't do so well nowadays. It's not that the design is wrong, it's that the particular implementation (collection of external assumptions onto what CPU characteristics the design characteristics is optimised for) is now out of date with more modern specifications.

If we remove external factors, get back to the nub of the issue, being what I mentioned in my second paragraph, and just release a new block optimised to each CPU type, then we get back to the realisation that there's not a lot left to be gained in block designs. Things are largely just being rehashed, and shrunk down as new machining techniques allow.

The problem with too much shrinkage is that pressure-drops increase more quickly than the gains, and there really is a point where we can shrink too far, because there don't exist pumps strong enough to push the flow-rate through quickly enough such that the thermal capacity of the water is not a major limiting factor to the design's performance. We can't just go ape with the pumping power to ram flow through true micro-structure blocks, because pump heat dump becomes an increasing factor, which in turn limits the gains seen.

Pulling together all of those multifaceted elements, we really are coming up to as far as we can go with water-blocks.

sick_g4m3r

07-09-2007, 10:03 PM

NOOOOO!!! Excellent answer but, nooo!!!

Whats the next thing to be improved upon? Rads are the last thing arent they? They are also at a pinnacle too, with excellent C/W ratings and very low deltas as it is.

Wow, this hobby is just about dead, isnt it? That kinda sucks. I am never coming back to this forum. bye!!

Jedda

07-09-2007, 10:05 PM

IHS integration?
Intell hires Cathar? lol
Immerse the cpu within the block in a WC iteration socket design, maybe?

Cathar

07-09-2007, 10:30 PM

The estimated h(eff) for my Storm/G7 prototype in silver is ~110-120,000 W/m²°K with 5W of hydraulic power. That value may not mean much to many people, but it is the first product of measurement of what I stated in my first post. Most blocks on the market today would be around the 60-80,000W/m²°K sort of range. That gives you some idea of where we are today, and where we can get to. Need to keep in mind that those values are independent of the thermal interface material layer, and so while a ~50% higher dissipation value might lead one to think that temperature deltas can be reduced by 1/3, the reality is more like 1/6th. i.e. if CPU is 12C above water temp, then we can get to 10C, rather than 8C, because of the TIM layer.

Mind you, the Storm/G7 prototype is a block that costs around $300US just to even manufacture (takes a CNC machine center half a day just to make a single block), so it's not exactly the sort of thing that you'd ever pop down to Fry's to pick up.

I can't see h(eff) ever getting above 150,000W/m²°K @ 5W hydraulic pumping power, ever. I may be short-sighted in saying that, but I'll stand by it and challenge anyone to make a physical block that proves that statement incorrect.

What does that mean? Basically that the limit would be CPU-water temp deltas that are around 3/4's what they are today, with 5/6's really being a realitically achievable goal. Please keep in mind that this is independent of variables such as mounting pressures, bowed plates, bent boards, and so on.

The challenge is still out there to make such blocks, cheaply.

For radiators, there's also not a whole lot left. Part of the reason for my fan/radiator testing is to help me to understand at a finer level where the limits might lie. There is always though the difference between performance at any cost, and affordable performance. Still a ways to go with affordable radiators though, maybe improving by 15-20% for any given fan power over the best of what's out there today.

Pumps still have a little way to go, but my goodness the DDC2's with the straight inlets comes so close to ideal that it's quite freaky (disregarding any perceived reliability issues). Laing were so close to the mark it's not funny. The only other major thing to do here is to reduce heat dump & noise.

The TIM layer is another area of benefit, but that's not specific to water-cooling.

There is also the on-going hunt for the best & quietest radiator fans. This is extremely important because it's fan noise that places the real upper limit on what we can achieve with radiators within a given size constraint.

So, you see, there is still lots to do, and looking on these forums here I see that there are plenty of people who are still enthusiastically contributing towards digging deeper at the aforementioned issues.

sick_g4m3r

07-10-2007, 06:42 AM

Wow Cathar, thats amazing insight. Thanks!

nikhsub1

07-10-2007, 06:52 AM

The days of huge performance leaps are over. If you were around this scene when the White Water was developed, you know what I mean. Water cooling was terribly stagnant at that time, and not for the reasons it seems that way today. Cathar gave the community a good kick in the pants with the WW and got the mfgrs off their collective rumps. The reason for the perceived stagnation today is for exactly the reasons Cathar listed above, there just isn't that much further we can go in regards to absolute performance.

Levish

07-10-2007, 06:56 AM

If dies were bare and larger we'd probably see better temps, and that test on matching heatsink/waterblock mounting surfaces (by creating a pattern) showed some promise (to increase surface area and even out mounting pressure accross the IHS).

Unfortunately neither are particularly likely to occour any time soon.

Mankz_91

07-10-2007, 07:01 AM

Surely the nextrevalution in waterblock design would be having direct die cooling, no IHS or baseto the block,just a secure water-tight bond around the CPU.

aussie-revhead

07-10-2007, 07:06 AM

Turtle 1

07-10-2007, 07:12 AM

Modding PC will never die it will grow and grow water cooling will remain as part of that modding experiance even if its not needed. Whats really going to change over the years is the size of the PC itself . Someday you will be wearing your pc on your wrist . I suspect it will be cooled by some form of fluid. So from now till than their will be new innovation . I suspect the major one will be from some sort of fluid that when run say 10c below ambient threw some sort of reaction . Who knows but water cooling dead not for a very very long time . Its fun . Its cool looking and its one hell of a good hobby .

Marci

07-10-2007, 07:13 AM

Surely the nextrevalution in waterblock design would be having direct die cooling, no IHS or baseto the block,just a secure water-tight bond around the CPU.
Played with years ago by many and abandoned as a fruitless chore.

The TIM layer is another area of benefit, but that's not specific to water-cooling.
Even that is potentially easily improveable, allbeit permanent... just whack an Indigo pad (http://www.enerdynesolutions.com/prod_indigo.html) on it, replicating IHS > Die bond between waterblock and IHS, or waterblock and die if brave enough. Or are you referring to Tim1 improvements at Die > IHS level as opposed to IHS > block?

entropy

07-10-2007, 07:23 AM

As the number of cores increase and the GPUs get more power-hungry, the bottleneck will be cooling capacity (i.e. radiator). I agree that water-cooling technology is probably not going to see major leaps in performance.

The next die shrink I expect would reduce the need for cooling capacity, but for people trying to cool an overclocked cpu + 2gpus there might be demand for bigger radiators (http://www.xtremesystems.org/forums/showthread.php?t=143393).

Oh, and I don't think watercooling is going to die so long as people need a quiet way to cool their machines.

montyshaw

07-10-2007, 07:27 AM

I think there is some room for improvement, probably more evolutionary than revolutionary, however. For me, as a relative new comer, I can only talk about the water cooling market as it exists today, and the areas that I would expect to see changes. I like the maturity of the products available today. It certainly helped me out in building a wc loop.

Large die CPUs like quad (hex? octo?) cores are going to very mainstream, so I'm thinking that water blocks that cool a larger area are needed. Like the new AquaCool duplex thingy. I don't know how well it does, but improvements like that. I doubt if Swiftech and D-tek are just sitting around. I bet we see some new blocks specifically designed for quad cores soon.

It also looks like currently, there is a need for a reliable pump that performs like the old DDC-2 w/top. The D5 is fine, but we need changable barb sizes.

Just a new comer's 2 cents of the state of water cooling blocks.

]Monty[

Marci

07-10-2007, 07:29 AM

zhangmaster12

07-10-2007, 07:47 AM

water cooling probably wont be able to performe better at some point in the future, but it can always LOOK better. We can strive to make the form pleasing, yet still have the pinnacle of performance, thatll keep us satified for some time. For example, look at cathars G7. Its probably to closest to the pinnacle of PERFORMANCE yet, but NOT (sorry cathar, but this is my opinion) remotely close to the pinnacle of form. An AC cuplex XT look much better that the G7. Also, we will strive to bring prices down, the best bang for buck and efficiency.

So, we might have almost reached to pinnacle of performance, but we still have a long way to go to achieve the pinnacle of watercooling.

Marci

07-10-2007, 08:05 AM

Pinnacle of form is always subjective to the enduser. Pinnacle of performance isn't. We'll NEVER reach the pinnacle of form - it isn't possible. For instance, personally I've always preferred the looks of the Storm G4 over most other blocks, including the Cuplex XT, which I think looks vile *shrug*

Polizei

07-10-2007, 08:27 AM

I happen to like the D-Tek Fuzion.

Just to re-itterate that perfect form cant be reached. I dont really have $0.02 to add.

dnottis

07-10-2007, 08:35 AM

Watercooling evolution in my mind when it comes to the radiator side of things is going to end up being small size high capacity chillers (600w) to replace radiators altogether, holding coolant temps above dewpoint at idle...

As RealEstate in a PC Case gets smaller and smaller, radiators getting larger and larger is a bit of a contradiction to how the rest of the technological world is moving.

I agree with this 100%.

I think what we need now are smart chillers. Something that can keep the water temp at ambient idle and under load via a controller. Keeping the water at ambient or as close as possible will eliminate issues with condensation.

I have another machine with the Vigor Gaming Monsoon lite. The thing works well on a machine that is running 3.0g Ghz @ stock volts. The little 50 watt tec isnt enough for my machine running 3.6 - 3.7 Ghz with voltages from 1.45 - 1.55v actual. But it's the same theory the tec kicks in when the temp hits a certain spot via a PCI controller.

However using something capable of cooling ~600 watts like Marci said would give you a setup capable of keeping a CPU and GPU near ambient temps in the room. Two problems come to mind... Power usage as tecs need 24v PSUs and price.

NaeKuh

07-10-2007, 08:37 AM

Watercooling evolution in my mind when it comes to the radiator side of things is going to end up being small size high capacity chillers (600w) to replace radiators altogether, holding coolant temps above dewpoint at idle...

As RealEstate in a PC Case gets smaller and smaller, radiators getting larger and larger is a bit of a contradiction to how the rest of the technological world is moving.

i cant disagree with this statement anymore marci.

I am currently thinkn of a way to drop a chiller right after my radiator, and right b4 my CPU, in the loop, so i could bring the temps of the coolant slightly lower then ambient/room temp.

Since i can almost safely assume water after the radiator would be the coolest, and water b4 the radiator would be the hottest, The radiator would bring the coolant temps near or as low as possible on your loop, and you would just need a chiller after the radaitor to compensate for the extra delta b4 it hits the cpu.

However this is a project that would take a lot of fine tuning, and im hoping i could somehow pull it off on a 60W tec, so i can have them scale with my RD-30. Higher i turn my RD-30, the cooler the TEC would run. The hotside remains a problem tho. I would most likely need to drop a dedicated loop to chill that hotside and that would require an extra loop in my already cramped case.

:\

Marci BTW~ thanks for coming on Anandtech. Your always welcome anywhere i post at, and please feel free to correct any statements i have said wrong in the past. :up:

sick_g4m3r

07-10-2007, 01:01 PM

Cathar

07-10-2007, 02:21 PM

Good stuff Cathar , following your study on coolant hose sizes what do you see as the bottleneck in modern watercooling ? If hose size doesnt help , increasing pump flow has minimal affect and rad size can only get you to ambient anyway does that mean everyone has to concentrate on water block design ?

I was trying to say that there is no one bottleneck now. It's no longer a case of targetting any one element and getting a huge gain there. It has to be done in a wholistic fashion nowadays. The big leaps won't come from changing a single block or a single radiator, it'll come from having a vision of where we need to get to and ensuring that all elements (waterblock, pump, radiator, fans, tubing, etc) are all in alignment. Select and put the sum of the parts together properly and there's still a lot to gain. Thing is that we're all just incrementally working our way there. If we could jump ahead 5 years from now, there will be a fairly large difference in the cooling performance of systems then and now. It's just the way it is.

Surely the nextrevalution in waterblock design would be having direct die cooling, no IHS or baseto the block,just a secure water-tight bond around the CPU.

CPU silicon is very slightly porous, so it's only a matter of time before enough water-molecules get in and kill the CPU. High speed jet-impingement or evaporative spraying would be the best ways to do the job though. The first will slowly wear away the silicon, while the second doesn't perform as well (to my knowledge.

Even that is potentially easily improveable, allbeit permanent... just whack an Indigo pad (http://www.enerdynesolutions.com/prod_indigo.html) on it, replicating IHS > Die bond between waterblock and IHS, or waterblock and die if brave enough. Or are you referring to Tim1 improvements at Die > IHS level as opposed to IHS > block?

My dislike of IHS's is well documented. They are a barrier to better cooling in and of themselves just because they do add an extra TIM layer, and extra heat-spreading material (copper) which really should just be integrated into the base-plate thickness of the cooling device if it needs it.

IHS's grant a measure of protection to fools who like to run their CPU's without heatsinks on them, or for badly designed mounting mechanisms that crush the dies.

I did mean between the waterblock and the CPU though, so the Indigo Pad would do the job if we never wanted to separate the block from the CPU again.

situman

07-10-2007, 02:42 PM

Just out of curiosity, is there a super durable material that can conform to the IHS like plastic wrap? THis super durable material can be the base of the water and regardless of shape of the IHS it will "conform" to it perfectly.

Cathar

07-10-2007, 03:36 PM

Great stuff! I agree 100% about the chiller thing and TEC junctions that would lower the water temp accordingly. These might be fun to make, as TECs cost relatively little, the problem is the hotside.

Let me run some numbers and see what would be plausible.

I can tell you right now what's plausible from direct experimentation.

1 x 226W TEC @ 5V, drawing ~45W of power, to bring water soaking up 90W of heat-load down to ambient
1 x 120.1 radiator on hot side per 2 x 226W TEC's to achieve this

This was the most power-effective balance point. Even though the TEC's are rated at 15v, if we raise voltage much above 5v, power draw goes up roughly proportional to the voltage squared, while the TEC's heat-pumping ability only slowly increases. We get to a point where we're feeding so much power into the TEC that the resultant hot-side heat-dump over-whelms the radiator.

Where that leaves us is that realistically we could build a 6 x 226W TEC chiller module that would draw around 270W of power, and cool up to 540W of total heat load with a 120x3 radiator on the hot-loop side, and bring the water temperatures down to around ambient. With lesser heat-loads than 540W, the water temps will decrease further. Under low heat loads (<100W), such a device would freeze the water in a loop solid quite quickly (matter of a few minutes), so it needs to be controlled properly.

By raising radiator fan speed, and increasing input voltage, we can get more cooling power out of such a unit, but the power-draw costs are high.

The issues are these:

- Finding efficient 5V power supplies that can provide ~300W of power (60A). They exist, and cost around US$70, but are not usually that efficient, and also come with their own fans that make a racket. A 300W x 5V PSU, will likely be drawing around 400W from the wall, and needing to convert 100W of it into wasted heat by way of its installed fan.
- Paying for 6 x 226W TEC's (~$40US each)
- Paying for the chiller blocks (~$100US)
- Paying for an extra pump (~$70US)

So, for around $500US, you could construct such a beast, and it'd drop your water temps under full load to ambient, or a bit less, if you were running an SLI + Quad core. Stick it on a typical dual-core + single-GPU machine and you'd likely get 10-15C below ambient water temps, and be having to worry about condensation.

The idea works, but it's a toss-up as to whether or not it's simply better to bite the bullet and go with phase-change.

santiagodraco

07-10-2007, 03:50 PM

I can tell you right now what's plausible from direct experimentation.

1 x 226W TEC @ 5V, drawing ~45W of power, to bring water soaking up 90W of heat-load down to ambient
1 x 120.1 radiator on hot side per 2 x 226W TEC's to achieve this

This was the most power-effective balance point. Even though the TEC's are rated at 15v, if we raise voltage much above 5v, power draw goes up roughly proportional to the voltage squared, while the TEC's heat-pumping ability only slowly increases. We get to a point where we're feeding so much power into the TEC that the resultant hot-side heat-dump over-whelms the radiator.

Where that leaves us is that realistically we could build a 6 x 226W TEC chiller module that would draw around 270W of power, and cool up to 540W of total heat load with a 120x3 radiator on the hot-loop side, and bring the water temperatures down to around ambient. With lesser heat-loads than 540W, the water temps will decrease further. Under low heat loads (<100W), such a device would freeze the water in a loop solid quite quickly (matter of a few minutes), so it needs to be controlled properly.

By raising radiator fan speed, and increasing input voltage, we can get more cooling power out of such a unit, but the power-draw costs are high.

The issues are these:

- Finding efficient 5V power supplies that can provide ~300W of power (60A). They exist, and cost around US$70, but are not usually that efficient, and also come with their own fans that make a racket. A 300W x 5V PSU, will likely be drawing around 400W from the wall, and needing to convert 100W of it into wasted heat by way of its installed fan.
- Paying for 6 x 226W TEC's (~$40US each)
- Paying for the chiller blocks (~$100US)
- Paying for an extra pump (~$70US)

So, for around $500US, you could construct such a beast, and it'd drop your water temps under full load to ambient, or a bit less, if you were running an SLI + Quad core. Stick it on a typical dual-core + single-GPU machine and you'd likely get 10-15C below ambient water temps, and be having to worry about condensation.

The idea works, but it's a toss-up as to whether or not it's simply better to bite the bullet and go with phase-change.

Very interesting!

From a control standpoint there's already a solution available if it actually fits (and I apologize to some of you in advance for using this name!) via Koolance's TMS card, of which I'm running one. The card can control fans and receives temperatures from 3 sources. It also has pump control. The software for the card is fully automated, that is it can control fan speeds and pump speeds based on temperatures of any of the 5 sensors. It can also control pump flow in the same way.

Koolance also just released the source code for the software.

Could this be a good solution that is available today to control such a beast? If the temps get too low the voltage to the TECs can be lowered, and visa versa?

Cathar

07-10-2007, 03:59 PM

What needs to be controlled is the power feed to the TEC's based upon the current water temperature and the current dew point.

santiagodraco

07-10-2007, 04:04 PM

What needs to be controlled is the power feed to the TEC's based upon the current water temperature and the current dew point.

Right, that's what I was referring to. It probably wouldn't work out of the box, but it might be close.

dinos22

07-10-2007, 04:04 PM

what about carbon based block base for watercooling...............we are seeing a lot of PR mumbo jumbo from OCZ about their air cooler but no results........................but if it is as good as it's cracked up to be surely future CPU waterblocks will adopt carbon and use in WBs

SiGfever

07-10-2007, 06:09 PM

As stated small efficient chillers or phase change will be the way to go. But they need to take compressor technology to the next level. A super small screw compressor (helical screw rotor) will give great performance. Or maybe even baby scroll compressors. Being able to chill the medium down below ambient and keeping above the dewpoint will give good performance but the additional components required for a chiller operation will make size an issue. Also the ability to add waterblocks to the chiller for vid cards, NB, SB and mosfets will be appealing. This is also where a screw will shine by being able to load/unload the compressor as the load changes.

I would think that phase change with a super small screw would be the optimum system but will be limited to the number of items that you can cool.

Turtle 1

07-10-2007, 07:27 PM

Marci I think your right about the chillers . My dothan has a chiller setup like what you describe. I can chill 400watts cooling -20c and heat 800 watts. It uses 230 V and its Very Very Expensive I usally have the control unit set for 10c unless it gets crazy humid. So its going to be awhile befor we see these things. I was smart tho . The control unit I bought can manage 800 watts cooling and 1600watts heating . But these are hard to work with. and the cost to turn on the machine and game with it blows holes in utility bill . But I agree this is the future but prices need to fall . 1000% before its anywhere near being marketable .

sick_g4m3r

07-10-2007, 08:47 PM

Cathar thats brilliant TEC stuff!!

Say we get two of those TEC junctions @ 5v and 90W. Say we wire them up to this PSU http://www.newegg.com/Product/Product.aspx?Item=N82E16817171007
which gets 32A on the 5V rail so it can easily handle them. Then we get a nice loop of a MCR120 and a d-tek db1 with some TEC junction blocks at $30 a piece.

$30(rad)
$60(2 x junctions)
$45(pump)
$30(PSU)
$20(extra WC stuff)
$40(2xTECs for cheap somewhere)

Thats about $200 for something that can take 180W to ambient. And on a 350W load it dry it to 170W needed to be cooled which is a big difference.

$200 dollars doesnt sound too bad, doesnt it?

Cathar

07-10-2007, 09:08 PM

Can also wire the pairs of TEC's in series off the 12v rail of a PC PSU, for 6v apiece, and that work okay too for a little extra oomph.

There is a trick to getting good performance out of TEC's for heat-exchange with respect to the blocks required to do so.

You're also wanting to keep a radiator on the cold-loop side, which you can do I guess. Then you'd just run the TEC's full-time, and ramp up the cold-side radiator fan speed if temps get too low.

Which brings about another project I've been wanting to do. A TEC powered air-conditioner. Double up the TEC's used, and just have a radiator on the cold-side loop. Run the hot-side loop outside to be cooled outside of the room, and have some fans blowing through the radiator on the cold side. Using 10 TEC's you can get around a true 1000W of recycled cooling power happening. Many air-cons actually suck air from the room that they're cooling and blow out out the hot-side, and so suck in warm air from outside the room which it then has to work to cool. While the BTU rating of the cold-side air may be high, they're really sucking more heat into the room as they do their job. At least, that's been my experience.

I've always been curious to see just how much cooling power I can get from TEC's for purposes of an air-condition with a 800W power budget. I do think that it's possible to give smaller phase-change air-con units a good run for their money provided the TEC's are managed properly.

sick_g4m3r

07-10-2007, 09:14 PM

So what your saying is to feed the cold air from the cold loop rad fans into the fans on the rad in the hot loop? Is that right?

Your comments are vey insightful and really excellent.

How about that $200 setup? Would it be beneficial?

Lord Powerlocke

07-11-2007, 12:50 PM

Moin !

Concerning the pinnacle of form thingy:
Just look at THIS (http://forum.effizienzgurus.de/showpost.php?p=43395&postcount=71) and call looks subjective again ! (Especially the last pic) ;)

sick_g4m3r

07-11-2007, 02:35 PM

Woah! that top is sweeetet!

santiagodraco

07-11-2007, 03:04 PM

TECs are great, but if you are looking for precision control, the circuits can get expensive.

I have messed with a few solid state and optical lasers that have TEC controllers, and they arent cheap.

But given the right controller if it had a GUI interface for the PC, you could just monitor say MoBo temp, and then control CPU temp to meet that, but then again it shouldnt be hard to hook a temp sensor up outside the cases either.

In the tool I work on, tempurature control is very very very important. When you are talking nano-meters, just a slight variation in temp, can cause that expansion.

The technique that is used in our tool, is using AC to cool below then actually heating up to the temp you want.

That's why I brought up the Koolance TMS setup as it's already has just about everything you need to do just that (software that uses hardware monitoring to control fans and pumps. Since fan and pump control is simply voltage, you'd think it could be adapted to instead control a TEC through one of those channels.

sick_g4m3r

07-11-2007, 03:18 PM

That's why I brought up the Koolance TMS setup as it's already has just about everything you need to do just that (software that uses hardware monitoring to control fans and pumps. Since fan and pump control is simply voltage, you'd think it could be adapted to instead control a TEC through one of those channels.

That does sound pretty good. I am done with this TEC thing though as I now realize it would be a HUGE PITA to construct with the double loops and extra WBs (2 per junction, and halfway decent ones to get improvement in temps) and I dont want to deal with it. Maybe in a year or two, but not now. I mean think of how much space this would take up... two more WBs in the core loop, TECs on those, and then 2 more WBs on those in another loop. Thats very hard.

Cathar

07-11-2007, 03:22 PM

What is the wattage rating per channel on the Koolance TMS device?

sick_g4m3r

07-11-2007, 03:41 PM

I got an idea:

http://www.imagehosting.com/out.php/i893625_TECidea.JPG

What do you guys think??

dinos22

07-11-2007, 03:41 PM

:ROTF: the image doesn't load :p:

sick_g4m3r

07-11-2007, 03:42 PM

Fixed! lol

Those copper things are high SA unrestrictive mazes BTW

montyshaw

07-11-2007, 04:04 PM

power to the pump? j/k :)

EDIT: Couldn't you use 2 fuzions or mcw-60's ?

sick_g4m3r

07-11-2007, 07:00 PM

power to the pump? j/k :)

EDIT: Couldn't you use 2 fuzions or mcw-60's ?

http://www.imagehosting.com/out.php/i893625_TECidea.JPG

No a custom copper maze would be a ton cheaper, less restrictive, and could be made with greater SA.

Cathar

07-11-2007, 07:15 PM

For those who haven't read it:

http://forums.overclockers.com.au/showthread.php?t=334206

The problem with using TEC's is NOT the block's surface area, it's clamping the blocks to the TEC's with sufficient evenly applied force. This is actually a major reason in why many people's attempts fail.

bobbobson

07-11-2007, 08:07 PM

sick_g4m3r

07-11-2007, 08:34 PM

Woah! Your OCAU thread is jacked with goodies! Did you continue your experiments and get any results? Andy pics? You have done all this before with great success!

Ambient was 22C, which was also the starting point of the cold-side loop. This is roughly what I record as the wattage being removed from the cold-side loop given the temperature of the cold-side loop. Very, very rough and dirty figures here.

22C => ~130W

So with your two 226W TEC's @ 5V you could bring 130W to ambient, correct?

Cathar

07-11-2007, 08:51 PM

Woah! Your OCAU thread is jacked with goodies! Did you continue your experiments and get any results? Andy pics? You have done all this before with great success!

So with your two 226W TEC's @ 5V you could bring 130W to ambient, correct?

Pics in the thread, well, ones that I have are there. No pictures of the TEC block's internals. That'll have to wait for another day.

Yes, 2 x 226W TEC's @ 5V would bring a 130W load down to ambient, meaning the water temperatures would be ambient, not the CPU.

This post (http://forums.overclockers.com.au/showpost.php?p=4065738&postcount=40) tells us (and reminds me) that I had gotten water temps down to 2C below ambient with ~120W total heat load (including pump), and drawing around 80-85W of power.

This was a bit less than I was hoping for. I was aiming for more like 180W of heat held to ambient with the TEC's off the 5v rail.

The project was left at a point where I had designed some better TEC blocks to help improve performance, but I haven't had them manufactured yet. One day...

sick_g4m3r

07-12-2007, 07:12 AM

So what would be the effect of having two TECs in the loop as well as a rad with a greater than 130W heat load, like 350W? Would the TECs take care of 130W and the rads deal with 220W? What might the temp drops be without the TECs?

EDIT:
http://www.employees.org/~slf/images/tecc3.jpg

Holy Cow Cathar!!! Thats exactly what I was envisioning, and there it is! Its beautiful! How come no one has seen this work before? Ever brought it to XS? What are the internals of those blocks made of?

EDIT:
Wow that OCAU thread is really really amazing!!!! I am only on page 3 and cant wait.

NaeKuh

07-12-2007, 01:10 PM

what about clamping a MCW6500 to an ApogeeGTX.

The apogeeGTX and MCW6500 have identical cold plate specs from what im seeing.

There both LGA775 platform, so you have a 4 clamp method.

Drop that right after the radiator so it needs to do less work to get the water lower then ambient, and then have the water feed from the ApogeeGTX to the CPU.

2 apogeeGT's with a TEC in the middle might be another possibility. Have the hot side on its own loop with a small radiator, and the cold side inline after the radiator fed to the CPU block.

Im just worried about how cold my water would get using this method. I dont know the electronic knowhow to build a switch that would automatically cut power once temps got to a low point.

I also believe the ApogeeGT has the base area to cover a 226W tec, You might need to use the GTX for the larger flow area. :T

dual Stinger V8's is also another possibility. But would the 226W TEC freeze the water when its inside?

Cathar

07-12-2007, 03:20 PM

NaeKuh, remember that the thermal capacity of water is quite high. For example, even if you "suck" 300W of heat out of 4LPM of water flow, the water temperature will only drop by ~1.08C from the entry to the exit of the TEC's. You won't freeze the water on its way through the TEC chiller.

Putting the cold side in the same loop with a radiator just has the radiator trying to warm the water back up to ambient if the water temps get below ambient.

Friedchicken

07-12-2007, 03:27 PM

On the radiator front, there is still room to go. Yes, they are getting larger to accomodate larger fans and hold more volume. But untill we see a watercooling system and case designed together, there will always be big variences between same rad/fan combos in different PCs (for example).

A friend of mine designs heat exchangers for the supercharged autos and so much engineering has gone into fin count, # of passes, core depth, materials, tubing & fin thickness, air resistance, effective cooling area... etc.

And even after all that- you still need to consider at what speed the car must be moving to achieve optimal air flow, and how effective the heat exchanger reacts in startup and in heatsoak conditions.

My point is, you can improve any number of areas in watercooling if you can standardize as much of your system and environment variables as possible.

sick_g4m3r

07-12-2007, 03:56 PM

So what effect would having two TECs in the cold side that can pull 130W to keep ambient, and a 120x3 rad have with a 350W heat load?

Cathar

07-12-2007, 03:58 PM

It'd roughly mean that the radiator would be acting as if it had to only cool 350-130 = 220W, and the water temps would be whatever they would be at a 220W heat load.

sick_g4m3r

07-12-2007, 05:38 PM

Wait there is one thing I dont get. If the TECs are drawing 45W of power, how can they pull out 65W each at ambient?

Cathar

07-12-2007, 05:52 PM

Wait there is one thing I dont get. If the TECs are drawing 45W of power, how can they pull out 65W each at ambient?

They're a heat pump. The peltier effect easily allows you to pump more heat than you consume.

It's not too different (in energy consumption concept) to an air-con. An air-conditioner can happily exchange more heat energy than it is drawing power from the wall in order to perform that heat exchange.

It's referred to as the Coefficient Of Performance (COP). Read here (http://en.wikipedia.org/wiki/Coefficient_of_performance) for some further explanation of the concept.

Air-conditioners typically run with COP's of around 2.0-3.0. A TEC can be made to do the same, if you power it right, and cool it right. That's what I was trying to prove/achieve with my tests. In practise I got a COP of ~1.5, but I still believe that a COP of 2.0 is achievable, which would have it compete with phase-change for efficiency.

sick_g4m3r

07-12-2007, 06:07 PM

OK thats awesome! I saw your 12V in series to get 6V to each TEC. I bet you got a 2COP with those, you think? That looked great! But that thread ended in the disappointment of no progress. What happened???

Cathar

07-12-2007, 06:18 PM

What happened???

I have new designs ready for the TEC blocks, but never went ahead. I was working with a guy in Canada who was making up a TEC controller that monitored ambient and dew point, and adjusted the power to the TEC's dynamically. User could turn a knob and effectively say "Keep the water at xxxC above/below dew-point", or flip a switch and it'd just give you full-power cooling full-time.

The project got stalled on waiting for him to complete the power controller. At this stage I'm starting to think that it's a lost cause (waiting any longer for it).

sick_g4m3r

07-12-2007, 06:22 PM

Yeah thats probably the case. That stinks though, but at least you have one for yourself. I am thinking of making one in the future.

Look at this, this is from the TEC section:

226 watt / 15.2 VMax / 24 IMax
5 volts: 24.5 watts; 7.89 amps

If it draws 25W, wouldnt your COP be 65/25 = 2.6? 2.6 is amazing! 25W seems too low though...

Cathar

07-12-2007, 06:28 PM

Someone's screwed up their maths there.

P = VI, so P = 5 x 7.89 = ~39W. I measured 41W myself at 5V. I have no idea what the 24.5W value is referring to.

sick_g4m3r

07-12-2007, 06:32 PM

Wow! lol thats sad. Its a sticky too............
http://www.xtremesystems.org/forums/showthread.php?t=67927

Every single number is messed up......

sick_g4m3r

07-12-2007, 06:39 PM

So from that chart (A x V), 6v on a 226W TEC comsumes ~ 57W.

You showed very considerable improvements moving from 5v to 6v on your 226W tecs in your OCAU thread. What do you think your COP might have been there. How many watts could be removed while staying at ambient do you think at 6v?

NaeKuh

07-12-2007, 06:40 PM

Cathar

07-12-2007, 06:46 PM

What do you think your COP might have been there. How many watts could be removed while staying at ambient do you think at 6v?

COP for TEC's drops as voltage is increased. This is why I was trying to make it work at 5v. The 6v would have worked better though, because the maximum temperature delta between the hot/cold sides of a TEC increases as voltage is increased, and the impact of this is affected by the cooling power of the radiator on the hot side. i.e. there is a balance point dependent upon the radiator cooling power.

At 6v, I guess around 80W moved.

sick_g4m3r

07-12-2007, 06:50 PM

So if 80W is moved, the hot side radiator has to dissipate that 80W, correct? Or the 57W of input?

Cathar

07-12-2007, 06:56 PM

So if 80W is moved, the hot side radiator has to dissipate that 80W, correct? Or the 57W of input?

The hot-side radiator must dissipate the amount of heat moved (80W) PLUS the TEC power draw.

For the 5v scenario ('cos the figures are handy), the hot-side radiator has to deal with 65W (heat pumped) plus 41W (TEC power draw) = 106W of heat-load.

sick_g4m3r

07-12-2007, 07:01 PM

so 106W per TEC? Thats 212 W to deal with on the hot side at 5V? wow!!

How could a MCR120 (just for example sake) handle that?

Cathar

07-12-2007, 07:04 PM

With a ~20C air-water temp delta with a quiet fan, or a ~10C air-water temp delta with a noisy fan.

sick_g4m3r

07-12-2007, 07:22 PM

So lets say 12C delta with a 70CFM fan (medium yate) (numbers are from swiftech's site). We have 12C over ambient, 25C as an example, for about 40C water temp. So if the hot side of the TEC is at 50C (10C more for the WB c/w), the cold side should be at -10C for this load, correct?

What was your TEC chiller made out of, its hard to tell in the pics? Aluminum?

Marci

07-13-2007, 01:16 AM

Heh - this all takes me back... Thermochill's first ever "product"... altho it never went to market and development on it was abandoned to focus on the HE Series Radiators...

http://www.over-clock.co.uk/marci/tec1.jpg
http://www.over-clock.co.uk/marci/tec3.jpg
http://www.over-clock.co.uk/marci/tec-a.jpg
http://www.over-clock.co.uk/marci/tec-b.jpg

Marci

07-13-2007, 02:01 AM

I was working with a guy in Canada who was making up a TEC controller that monitored ambient and dew point, and adjusted the power to the TEC's dynamically. User could turn a knob and effectively say "Keep the water at xxxC above/below dew-point", or flip a switch and it'd just give you full-power cooling full-time.

I'm sure this could all be done easily with the controllers (Asetek ChillControl) for the Vapochill LS / XE / PE series systems, which had ability to adjust compressor speed as well as fan speed... compressor was 12v DC, fans were 12v... controller also had ability to control 110/240v AC compressors too... therefore, you could use it's 12v dc compressor controll to control small number of TECs in series, or the 110/240 control to possibly manage an external PSU... (??)

aussie-revhead

07-13-2007, 03:03 AM

Im glad to see all of this renewed interest in TEC cooling as I have thrown ideas around in my head many times , but my concern is that if we are talking about mainstream cooling solutions does that mean that every day computers will take up the space of a server cube case with one side dedicated to radiator-pump-TECs-water blocks-power supply-controller etc etc ? This may be cool for an enthusiast pc but for everyday use the cooling system will take up more space than the pc hardware .

Sounds odd now but it may be the way of the future.

Good work on the TEC stuff Cathar !!!

:up:

Marci

07-13-2007, 03:10 AM

We're not talking mainstream - far from it. Watercooling isn't even "mainstream" yet, and never will be in it's current form due to it's cost and size. The allinone units that we all laugh at will become mainstream long before what we here constitute as watercooling becomes mainstream, and even then it'll be short lived as a decent thermosyphon would slaughter it.

For everyday use, stock heatsink / aftermarket heatpipe or thermosyphon. Anything beyond that is an ePenis thing... and thus not mainstream.

aussie-revhead

07-13-2007, 03:25 AM

sick_g4m3r

07-13-2007, 03:32 AM

Yeah thermosiphons do look very interesting. However why havent we heard of many that work well? Doesnt the mounting have to be near ideal to get the vapor to rise and liquid to flow? Also, the low pressure needed to be created in the tubes is extremely low in order for the TS to be effective (for the water to boil at low temps).

Marci

07-13-2007, 04:17 AM

I was looking ahead to when multi-cores are mainstream as someone above said. Its quads now but 8 core is probably in the pipeline too and Im sure these will need revised cooling solutions , unless future manufacturing processes reduce emitted heat significantly.

Roadmaps already in place from CPU manufacturers thru til 2015 (and have been in place since 2 years ago), with everything based on either microchannel liquid cooling via all-in-one setup with microchannels combined into IHS and pump/rad mounted directly on top ala the kingwin thing that's knocking about, but done by someone with intelligence... or refrigerant-based thermosyphons (not water). CPU Manufacturers know exactly how it's gonna go for the mainstream market for all CPUs planned.

Quadcore still isn't a mainstream product... it's a highend server & enthusiast product - still only counts as a fraction of 1% of the total CPU market as it stands, thus quadcores aren't mainstream. They're mainstream within the enthusiast scene, but not the global PC scene... yet...

aussie-revhead

07-13-2007, 04:24 AM

Quadcore still isn't a mainstream product... it's a highend server & enthusiast product - still only counts as a fraction of 1% of the total CPU market as it stands, thus quadcores aren't mainstream. They're mainstream within the enthusiast scene, but not the global PC scene... yet...

I know that sir, but they will be eventually.

CPU Manufacturers know exactly how it's gonna go for the mainstream market for all CPUs planned.

Im sure, lets hope they coem out with decent factory coolers then.

I havent read at all on thermosyphon , going to read some now.

:up:

Edit: Is thermosyphon more or less what goes in inside a heatpipe cooler ? ie passive circulation of coolant ?

;)

sick_g4m3r

07-13-2007, 04:33 AM

see my new thread for thermosyphon stuff

Cathar

07-13-2007, 04:35 AM

The allinone units that we all laugh at will become mainstream long before what we here constitute as watercooling becomes mainstream, and even then it'll be short lived as a decent thermosyphon would slaughter it.

Queh? You sayin' that a thermosyphon exists that beats water? :hm:

I still perceive one of the major limiting factors with thermosyphons as being the fans.

Marci

07-13-2007, 04:52 AM

Edit: Is thermosyphon more or less what goes in inside a heatpipe cooler ? ie passive circulation of coolant ?

Yep

Have a look at http://www.apialliance.com/pdf/Archive_semiwest_05/AMD_Touzelbaev.pdf for where AMD were looking 2 years ago (Stew - you'll remember this from ProC TIM1 discussions - micro capillaries etc)

Queh? You sayin' that a thermosyphon exists that beats water?
A refrigerant-based thermosyphon resembling visually Thermalright's Ultra120 Extreme layout would beat anyone's implementation of this breed of watercooling (http://anandtech.com/casecoolingpsus/showdoc.aspx?i=3032), which is what will become "mainstream" watercooling for the average home PC to replace the "stock heatsink"... imo... that's where I'm making the comparison... heat dissipation vs surface area vs noise vs weight, I reckon the refrigerant-based thermosyphons would have water beat when looking at the potential mass-replacement for the stock heatsink for high thermal-output packages... (???)

Also, consider the Thermal solution becoming part of the CPU - the IHS being the evaporator, with condenser suspended above Thermalright style... preaffixed out of the box on "retail" cpus... eliminates TIM2 entirely...

Open to debate, but it seems to have been hinted to head this way by IBM, Intel, AMD etc for a year or so...

But anyways, this doesn't affect us... we aren't the mainstream users... we're the enthusiasts, and we'll still be buying OEM CPUs without integrated cooling, and whacking on our own largefootprint industrial scale cooling appliances regardless.

I'm just trying to make the distinction of what we're referring to as mainstream and how what mainstream gets is a different rendition entirely of anything we'll all be playing with...

but my concern is that if we are talking about mainstream cooling solutions does that mean that every day computers will take up the space of a server cube case with one side dedicated to radiator-pump-TECs-water blocks-power supply-controller etc etc ?

aussie-revhead

07-13-2007, 04:55 AM

Would the radiator area of a Thermalright Ultra be enough for this implementation? Of course a self contained cooling solution like this would be much easier to deploy in retail packages than a complete water cooling kit.

http://www.apialliance.com/pdf/Archive_semiwest_05/AMD_Touzelbaev.pdf

Interesting stuff.

:up:

Marci

07-13-2007, 05:01 AM

With the use of refrigerant within... sure... depending on the fans in play and the choice of refrigerant and the quality/efficiency of the evaporator...

Everyday CPUs run at stock... these may not be good for overclocking, but for your everyday home pc running stock, that kind of footprint would be fine.

Cathar

07-13-2007, 05:06 AM

A refrigerant-based thermosyphon resembling visually Thermalright's Ultra120 Extreme layout would beat anyone's implementation of this breed of watercooling (http://anandtech.com/casecoolingpsus/showdoc.aspx?i=3032)

Cheers, thanks for clarifying. When I read the statement, it read like you were saying that by the time our industrial water-cooling solutions became mainstream, that thermosyphons would beat them, not just the all-in-one jobbies. If referring to beating all-in-one jobs, that makes perfect sense.

Marci

07-13-2007, 05:31 AM

God no... if you upscale thermosyphon's to our performance levels, the equivalent ain't a syphon... it's a singlestage phasechange... if anything mainstream is just filling the middle ground with the gap between heatpipes and compressed phasechange... and that already exists. The silent market could have a lot to benefit from thermosyphons but they're a different breed to us... they UNDERCLOCK *spit*

At the enthusiast end of the market - ie: for us, everything will remain the same but with refinements & improvements of existing techniques...

sick_g4m3r

07-13-2007, 05:40 AM

Yeah Marci thats a great way to sum it up. Looks like TEC chiller now is the next step, correct?

aussie-revhead

07-13-2007, 05:41 AM

Yes I have seen TEC threads starting up again in other forums too.

:up:

Scarlet Infidel

07-13-2007, 11:28 AM

I was always keen on TECs but whenever I did the maths I came to the conclusion of high cost and wasted power. I like this idea of running the elements at 5-6v.

I intend to do some reading on the subject, but to save some time, what constitutes a good TEC block?

Would clamping a couple of peltiers between two HDD water blocks such as these (http://www.xspc.biz/photo/HDD/) be any good? That would lend it self to a relatively cheap, easy and high flow system but i dont know how it would fair performance wise.

sick_g4m3r

07-13-2007, 08:09 PM

Cheap? Those things are like $50 a piece in the US, $100 total! I can get all the copper for the two pieces and mill a nice maze with a drill press, the copper costs like $30 total.

Scarlet Infidel

07-14-2007, 12:36 AM

True, they are very simple designs and a lump of copper with some copper pipes attached would be cheaper (or mill a maze out of thicker pieces). You can usually get this kind of block for £10-£15 though (if you look hard enough) which isn't hugely expensive.

Scarlet Infidel

07-14-2007, 09:18 AM

Having just checked ebay (in the UK) for current peltier prices I find they have different models from the ones im used to. They still have the 320w models i was expecting but now there are ones at 250w and even 543W whcih i havent seen before (probably old news to many).

Im not overly sure on how the sums work so id be glad if someone could help me out. However, using P = (V^2)/R and the resistances claimed on ebay; the 543w element would draw between 80w and 93w at 5v (values for 50c and 25c temps respectively) and the 250w elements would draw 33.3w (at 25c).

Edit: ignore that, I think its rubbish. I hope someone can tell me how to work it out properly.

Marci

07-16-2007, 04:15 AM

Would clamping a couple of peltiers between two HDD water blocks such as these be any good? That would lend it self to a relatively cheap, easy and high flow system but i dont know how it would fair performance wise.

Not really - clamping is by 2x screws halfway down length of each side of block. You'd struggle to get sufficiently even clamping pressure to get the desired performance... as cathar said earlier in the thread...

The problem with using TEC's is NOT the block's surface area, it's clamping the blocks to the TEC's with sufficient evenly applied force. This is actually a major reason in why many people's attempts fail.

If it were as simple as using HDD blocks, that'd have been done already...