The impact of tubing sizes

[IanY]

Remember my rant about the Stealth? I gave Gabe a hard time because it shows no temp improvement over the MCW60 even on the gpu core, notwithstanding the improved GTX/GT base.

I will walk through walls of fire for a 1 degree change in temperature. If it were not for my wife, I would run the room airconditioner in my personal study (a small room with a 12,000 BTU wall unit!! ) in the middle of winter where it is -5 degrees F outside, just for that extra lower load temps.

I understand many people are not willing to go to such lengths, but its up to them. I'll just continue doing what I need to do.

[NaeKuh]

Remember my rant about the Stealth? I gave Gabe a hard time because it shows no temp improvement over the MCW60 even on the gpu core, notwithstanding the improved GTX/GT base.

I will walk through walls of fire for a 1 degree change in temperature. If it were not for my wife, I would run the room airconditioner in my personal study (a small room with a 12,000 BTU wall unit!! ) in the middle of winter where it is -5 degrees F outside, just for that extra lower load temps.

I understand many people are not willing to go to such lengths, but its up to them. I'll just continue doing what I need to do.

ahahahahaha... so true.

I think your moto was, i dont care how fugly it is, as long as its king.


For me, i think looks has about 40% to do with parts. 60% on performance. If the RD-30 wasnt so blingy, i would of done 4DDC-2 2-alphacool 2in1 tops. :T

[IanY]

If the RD-30 wasnt so blingy, i would of done 4DDC-2 2-alphacool 2in1 tops. :T

You should see the latest on pumps and flow being over-rated... that you need no more than one DDC+ in your loop, and anything more would just be heating up the water... hey... I thought I held the patent for the four pumps into two Alphacools lol

[Cyprio]

This has me concerned. mCubed advertises the ability to run pumps via the BigNG. Is this inaccurate?

You could always use an Aquacomputer USB Poweradjust that was specifically designed to control a Laing DDC. You'll have to get one for each pump though.

[Cathar]

Like someone else mentioned, funny how Swiftech had it right for so long and how Gabe and co still used it in their machines this whole time.

They were right in the same way that if you wear some old bit of clothing that was once fashionable, it'll eventually come back into fashion again.

Historically there were justifiable reasons for going with 1/2" tubing. In the specific case of Swiftech though, their 5yo old MCW462 design, the inlet into the block was in itself a jet impingement effect. By using smaller tubing (3/8") the MCW462 saw a boosted jet speed onto the base-plate that overcame the flow-rate impact penalties seen with the pumps of the day. Also, Swiftech did the right thing, and used quick-fits, so these two factors combined meant that 3/8" was the better choice for their block design. The MCW5000 block followed, and that had a flow-agnostic thick base-plate, so once again the reduced flow had minimal impact. The MCW6000 followed, once again with a "the fitting is the impingement jet" design, coupled with a thick flow-agnostic base-plate, and so yet again 3/8" was justified.

The rest of the market pursued other designs where 1/2" was more justifiable, at least until fitting-independent accelerator plates and flow-agnostic internal structure was researched and more properly understood. Further, as block restriction was added to boost performance, this in turn increased the pumping power required, and soon it was discovered as people explored using more powerful pumps that pump-heat dump was playing a significant role in the final CPU temperatures. In time, the need for >6LPM flow-rates to achieve good performance, once pump heat dump is factored in, diminished. To a fair degree, it's quite possible to design a system & waterblocks to suit, where flow rates >4LPM are never required to achieve maximum performance.

At 4LPM flow-rates, even 5/16" (8mmID) tubing with quick-fits presents a minor hurdle to overcome for the pump. At 8LPM flow rates the resistance of the tubing quadruples and becomes very significant. Design every waterblock in the system to operate well at 4-5LPM, creating a level of restriction where pump heat becomes significant, and we have an ideal system.

Also, better designs in pumps, in part inspired by the better understanding of the levels of pressure & flow & pump-heat, and their impacts on a complete system, meant that we started seeing pumps on the market (whether created specifically ala Laing, or sourced from available stock ala Iwaki or Panworld) with more pressure and less peak flow. This aspect was almost critical for bringing the 3/8" and 1/2" worlds closer together. As we added a little bit of extra restriction, the pumps that are well designed can "shrug it off" and continue to plow through the restrictions because their PQ curves are now far more optimally balanced for the PC water-cooling scenario than the pumps we used to have.

The DDC's have best efficiency points (BEP's) within the 4-6LPM range. The "ideal" system to work with these pumps would have a single-CPU block design that is fairly flow-agnostic by around 4LPM and offers a level of restriction that results in 5.5-6LPM flow rates in a 1block/1pump/1radiator system. The end-goal for all the blocks in a 1cpu/2gpu/1chipset/1rad system would be 4-4.5LPM flow-rates, with blocks that are specifically designed to be fairly flow-agnostic by that point.

The final piece of the puzzle is the radiators. More modern radiators are now more flow-agnostic. The performance hit of lower flow-rates from smaller ID tubing is reduced over what once used to be.

So, Swiftech's CPU block designs specifically favored the use of 3/8"ID tubing because of their "the fitting is the jet" designs, and where that wasn't the case, the base-plates were thick enough to not care too much. In the meantime, greater knowledge and understanding in the market place has led to a selection of modern off-the-shelf components where the impact of lesser flow-rates is minimalised as well.

There is still room for improvement though. I still don't see any manufacturer taking a fully wholistic approach to a range of products from top to bottom. That's pretty expensive work though, carefully balancing an entire range of products. The impact of modern pumps and radiators makes the task easier though, but it's been a long road of discovery to get to this point. We're not at the end yet, but we're getting closer.

I personally consider the 775 IHS implementation to be completely flawed, and it has caused the market to take a back-wards step to work around it. I know I'm a fairly lonely voice in a market-place of enthusiasts who don't care about how getting better temperatures is achieved, so long as it works, but believe me when I say that things can still get better. It's taken me some while to understand just what the heck was really going on with S775, 'cos believe me, the theory of it assuming a proper mechanical mounting system made no sense. That's just what the problem was though. The mechanics of the situation are wrong, not the block designs that have been pushed to the wayside in the march for better performance, and we're currently seeing an artificial "bubble" in the forwards progress of block design as a result.

Well, that's my summary of the history to now...

[ranker]

There is still room for improvement though. I still don't see any manufacturer taking a fully wholistic approach to a range of products from top to bottom. That's pretty expensive work though, carefully balancing an entire range of products. The impact of modern pumps and radiators makes the task easier though, but it's been a long road of discovery to get to this point. We're not at the end yet, but we're getting closer.

How do you see watercooling progress in terms of new technology and capturing a larger audience than your typical pc enthusiast?

I'm interested to know as the gains over the past year have really been minuscule compared to the years before. It's gotten to the point where it's almost 'boring' where manufacturers appear to be focusing more on bling rather than pushing the envelope.

I'm really interested in hearing on what you believe the future of WC to be and how we can get there.

[SiGfever]

With the introduction of the DDC series and the use of Iwaki pumps designs such as the Storm should be optimum with the jet impingement. These pumps have the extra kick to make full use of the jets.

Do you think that a design where the jetting is divided into two distinct areas focused over each core would allow the Storm design to shine again?

[NaeKuh]

Cathar, send me a G7 block.

OR better yet, make me an offer. :P


Im really dying to own one of your original pieces. I would seriously love one.

[Cathar]

How do you see watercooling progress in terms of new technology and capturing a larger audience than your typical pc enthusiast?

I'm interested to know as the gains over the past year have really been minuscule compared to the years before. It's gotten to the point where it's almost 'boring' where manufacturers appear to be focusing more on bling rather than pushing the envelope.

I'm really interested in hearing on what you believe the future of WC to be and how we can get there.

Capturing a larger audience eh? Might be a good time to take a look at who makes up the respective audiences, and then look at the impact that the unified platforms (consoles) are making into audience that the pc watercooling market might hope to capture. Microsoft is on a mission to own the non-business home computing market, software AND hardware. What makes you or I a PC enthusiast? As in what is it that we actually used a PC for? I'd be surprised if >95% of the PC enthusiast market wasn't captured in the following 5 reasons: "games, internet, folding/distributed computing, music/video viewing/editing, specialised/business software".

Microsoft is targetting the first 4 of those reasons, and some elements of #5 as well. Make no mistake that as enthusiasts, we're really just a pack of die-hards holding onto a computing concept that is earmarked, if not for eventual total elimination, at least for near-complete marginalisation, some time in the next 10 years at a guess. About the only "friends" we have is the business computing market, and you can be sure that they only care about the cost of the end-systems. A stock PC with a quiet air-cooling is good enough. There's no water-cooling market there, or at least, not while water-cooling costs more than air-cooling.

No, I don't think that the PC water-cooling market will ever really take off for that reason. It's always going to be the realm of the die-hard enthusiasts, because everyone else will either be using a unified hardware/software platform solution (Xbox v4 or 5), or using a PC in an office for business.

As enthusiasts, our solutions don't fit in with an XBox or a business PC. Both those computing types may get water-cooling one day, but it'll be some OEM solution that the enthusiast market has no input in.

Sorry for the bleak crystal ball outlook.

As for pushing forwards within the realm of the enthusiast PC market, it's true that we've come a long way, and the gains still to be had are small. There's not going to be any big gains (with the exceptions of bowed blocks for mechanically flawed socket implementations that are more of an anomaly than anything else).

Pumps - The DDC/DDC+ are fantastic PQ/noise wise, but still some cloud over their reliability. There is still room for an astute competitor to come in with a variable speed 12vdc pump and clean up. The performance gains will, at best, still be less than 0.5C, but a pump with 6-8mH2O peak pressure head, 9-11LPM peak flow, and a BEP in the range of 4-6LPM, and as energy efficient as possible (no more than 25W), small, reliable, quiet, and less than $90US, will basically clean up the PC water-cooling pump market, period. The top-modded Laing DDC2 is oh-so-so-close, but it seems that Laing have been dropping the ball lately to jump the final hurdles to take it to a position of total market dominance. Take a pump like that, add on a "speed dial" like the Laing D5 to allow for switching between 5-25W pumping modes to suit the widest range of setups, and no one would ever need buy a different pump again. Buy one pump, dial in the power required that is in balance with your radiator's cooling power, and done.

Radiators - We are close. Very, very close. We can eke out another 5-10% performance or so in an affordable, marketable format, and we can still do a little more work to make them more flow agnostic. I have ideas on how, but won't share them here in public.

Tubing - fat tubing is a turn-off for many, and it makes it hard to route tubing tightly around motherboards and components without involving a lot of forward thinking, and trade-offs. This is part of the reason why I revisited that concept here. 10-12mm OD tubing, while I'm sure some will disagree out of principle, is about as large as is physically practical. There are some small hurdles to overcome here (what fittings are best, and so on), but to get to a point where people can using 10-12mm OD tubing and easily route it without grief, and not see more than a 0.1C performance difference to 3/4" OD tubing would be the final goal here.

Waterblocks - IHS's are here to stay. Some older designs (mine included) were designed at a time that they had to straddle the opposed duality of bare-die CPU's and IHS-capped CPUs. There are a number of very highly performing designs that have been passed over because the specific physical implementations of the design were not wholly optimised for IHS CPU's alone. The magnitude of this has been freakishly amplified by the mechanically flawed nature of the highly popular S775 CPU's. We can still conceptually do ~4C better (at Tjunction, not Tcase) than what's out there on the market today, but the existing IHS implementation on S775 CPU's is preventing much of that from becoming a reality. In short, we're somewhat stuck in a rut with some less than optimal (but still highly performing - make no mistake) designs until the mechanics of S775 changes. In short, the waterblock design market has been backed into a corner that doesn't allow the better designs to really do their thing.

Beyond that rather vague overview, there's a lot of research that needs to take place to get from where we are now, to the best possible. It's probably best, in a sense, that it still remains a largely enthusiast based marketplace, 'cos its the collective passion for the best that will ultimately drive the for the push for perfection, rather than stagnating in unified design mediocrity.

[ranker]

Hey Cathar,

Thanks for the reply as it was an interesting read. You're bleak outlook on WC'ing is disheartening. The hobby itself has been interesting for the past 3 years. To see that large leaps in terms of performance, ease of use, and reliability are over seems to have sealed the fate for WC'ing itself.

If I'm reading this right, you're stating that there's only a limited amount of performance left to capture. Most enthusiasts will usually demand more and more performance and make the leap to the next level. In time, phase will become cheaper. What would prevent phase cooling from eventually cannibalizing the WC'ing domain?

I'm interested in knowing what you're personally working on (besides stuff for Thermochill) and what you see yourself contributing to the WC scene in the near future. Will we be seeing more Cathar designed waterblocks or will you be focusing on radiators as you were recently?

[HiJon89]

I think there's still room for improvement. Just look at something as simple as air cooling and there are still huge leaps being made. The Ultra 120 Extreme blows away the XP-120 which was king of the hill 2 years ago. I think the future is going to be ultra restrictive blocks and pumps pushing ridiculous amounts of head.

[RickCain]

If I'm reading this right, you're stating that there's only a limited amount of performance left to capture. Most enthusiasts will usually demand more and more performance and make the leap to the next level. In time, phase will become cheaper. What would prevent phase cooling from eventually cannibalizing the WC'ing domain?

OCZ was supposed to enter the phase market with a $300.00 unit but it never materialized. That would have pushed a majority of the watercooling people to phase overnight.

Cathar - wouldn't this be a good time to get your chilled water solution moving forward? That should result in nice gains in watercooling performance.

[nikhsub1]

What would prevent phase cooling from eventually cannibalizing the WC'ing domain?

1 werd. CONDENSATION. Ok, one more word. RELIABILITY. Gee, one more - HEADROOM. Oops, another. GPU's. Meh, a few more? NB, SB, mosfets? RAM? HD's. You can phase all that.

[Burn]

IIRC Stew said a while back that the possibility for a performance increase is quickly becoming impossible, even using silver. Because the die is so small, trying to remove 100+W from such a small package becomes difficult to do better and better, and at some point you reach a ceiling.

[NaeKuh]

welll, theres this concept i kinda wanted to try out. Basically a inline chiller, but not something that would drop major voltage on.

Rad -> TEC chiller -> Cpu

Basically the tec chiller would only be used to knock off 5-6C on water temp b4 it was shot into the CPU. So i figure not a very high voltage TEC would be required to do that.

And since the rad will most likely bring the coolant down to the lowest possible, it makes more sense putting it right in front of the TEC chiller.

Only problem is, how to make the chiller. :P

[MaxxxRacer]

Some interseting info Stew. Glad to see you back on the forums.

I stickied the thread and will be updating the main guide with info found in here when I get some time.

[ranker]

1 werd. CONDENSATION. Ok, one more word. RELIABILITY. Gee, one more - HEADROOM. Oops, another. GPU's. Meh, a few more? NB, SB, mosfets? RAM? HD's. You can phase all that.

The condensation part is definitely a good point, but lets not forget that people use to balk at the idea of mixing water with electronics. I think we really need to get back to the purpose of WC'ing. Does any other component depend on better cooling to get great OC results? Besides the CPU, everything else can't really be OC'ed other than GPU's. In addition, GPU's are typically volt limited rather than temp limited.

The same argument you make against Phase have been said during the early days of WC'ing. Most people were just happy to have WC'ed the CPU. People began to speculate about GPU's only to have the naysayers say it was impractical and ineffective. Fast forward years later and now we have people like me WC'ing everything that has a waterblock made for it and then some. Instead of truly focusing on the important aspects of WC'ing, manufacturers have begun to "accessorize" the hobby by making blocks for mosfets, chipsets, HD's, etc and other stuff. It seems as if progress in general is headed towards bling rather than performance as I've stated.

If phase is the future, I can see manufacturers developing something that might be "expandable" to be able to cool multiple blocks. I'm not sure how it would work but possibly the size of the unit itself might shrink to the size of a say a PA120.3. Once the size and price point are lowered, I can see multiple phase "loops" being run within a PC.

As an engineering grad, for someone to say it's impossible or impractical is usually a challenge and not a final answer.

However with that said, I'm hoping that there's still room left for WC'ing to grow and evolve. It's disheartening to have read the past few posts by Cathar and really drives home how bored I've been getting with the hobby in general. I was hoping for a shining light at the end of the tunnel.

[HiJon89]

Does anything else depend on cooling to get great OC results? Besides the CPU, everything else can't really be OC'ed other than GPU's.

Well high FSB does overclock the NB in a way. And at high FSB, watercooling the NB can become essential to maxing out the CPU clocks. (I'm still a little dubious about the SB and mosfet blocks )

For the most part, GPU's are volt limited rather than temp limited.

And after you volt-mod the cards and kick up the voltage, they're right back to being temp limited

[Grinch]

Well high FSB does overclock the NB in a way. And at high FSB, watercooling the NB can become essential to maxing out the CPU clocks. (I'm still a little dubious about the SB and mosfet blocks )


And after you volt-mod the cards and kick up the voltage, they're right back to being temp limited

good points..

[afireinside]

I will walk through walls of fire for a 1 degree change in temperature. If it were not for my wife, I would run the room airconditioner in my personal study (a small room with a 12,000 BTU wall unit!! ) in the middle of winter where it is -5 degrees F outside, just for that extra lower load temps.

Man your stuff better come pretty damn fast at stock to need a 12k BTU AC to keep it cool...

[sick_g4m3r]

Its never ever over!! Ever!! Muahaha. Truly it wont be. Thats like saying the gasoline engine was over as soon as it was made.


And that TEC Chiller sounds like a good idea. maybe a 45.6W TEC sandwiched between an HSF and a maze-like copper block.

cooling the copper block which cools the water with a lot of contact from the maze-like block. Might work

[Grinch]

it seems like I am upgrading every 2-4 months...not including cases..LOL

[Cathar]

Do you think that a design where the jetting is divided into two distinct areas focused over each core would allow the Storm design to shine again?

There's nothing wrong with the Storm design even now. It focuses on cooling a central patch, and that is exactly what it should be doing.

The "problem" is that the S775 CPU's mechanical mounting characteristics results in less pressure being applied at the center of the IHS between the die and the water-block, and so what instead occurs is that the heat from the die instead spreads out more over the entire IHS, rather than remaining centralised above, which is completely opposed to how the Storm does its job.

i.e. the problem is a mechanical socket issue, not a waterblock design issue.

Now, the mechanical socket issue can be addressed in two ways.

1) Increase the size of the cooling patch to cool the entire IHS. The impingement aspect of the Storm design is better per unit area than the mini-pin designs. I know this to be true (in my own testing - others may disagree) based on bare-die performances. Since the mechnical flaws of the IHS-core mating spreads the heat out over a larger area than the existing Storm implementation is designed to handle, the solution is to expand the cooling patch size.

2) Create a way in which we can properly sandwich the IHS between the die and the block base with a high amount of pressure. One (ugly) solution is to bow the bases. Another solution that I believe would work well is to machine a 22x22mmx0.5mm high step into the middle of the baseplate of the block. On a Storm block this would be achieved by milling away the baseplate by 0.5mm around a central 22x22mm square, a bit like how some GPU blocks are, where they have that raised middle portion to fit inside the shim. This still wouldn't cover the entire IHS, but it'd apply pressure in a manner that would ensure that the IHS gets sandwiched with strong force between the actual die and the wb base. This raised section would also be mostly covered by the Storm's cooling patch. This would be how it would be easiest to modify an existing Storm block to be competitive again on an S775 CPU, and bring the cooling patch design back to where it expects the heat-load to be.

Doing #2 would "fix" the current Storm for single-die (dual core) S775 CPU's.

Doing #1 would fix the Storm to a point where it'd compete, or even outperform, once again anything else with an unbowed base.

Doing both of the above would "fix" the Storm design for quad-core (dual-die) CPU's (but with a 26x26mm raised section instead).

[migueld]

There was a time where the Storm was seen as the absolute best block. The general consensus at the time was that water-block design had reached its peak, and that performance wouldn't get any better, at least not significantly, not even with the mythical silver storm..... anyone remember those days?

Just when the situation seems hopeless some Fuzion thing comes out and blows everyone out of the water. It defies the "laws" of the days, it's several times less restrictive, and several degrees cooler.

I say as long as we don't get clouded by the old and are willing to consider the unknown, there's plenty of universe left for exploration

[Cathar]

There are always engineering reasons for why things are the way they are, or why things changed from what was expected. The true way forwards is through scientific analysis, understanding, modification and refinement. Not romance.