A challenge!

[Cathar]

Originally Posted by pHaestus

From Bill's data cited by Les this morning though it seems like the penalty from using the quieter fans is substantially lessened by increasing flow rates through the radiator...

Yah, I was already taking that into account, however that is also a per-radiator dependent property.

[dutchcedar]

Originally Posted by unregistered

the Exos ?
Korea
Corsair ?
Mexico (but designed by Delphi)

but it is the seductive phrase "you don't need a lot of water to cool" that has misled so many

Thanks Bill. Exos is the one I was thinking of.

I'm not surprised the Corsair units are made in Mexico, but just for fun I looked up the shipping tag and it indeed came from Delphi in Detroit. It was, however, a pre-production unit.

[knipex]

Originally Posted by pHaestus

If a greater premium is placed on quiet than on performance, that's fine. But why feel the need to state that you're "at the most a degree or two" away from the highest performance solutions available? There's no basis for that really other than wanting to maintain a big e-penis.

Sorry Ph again my fault.

I seem to have drifted a bit here. I started talking about low flow in general and ended up talking about my particular (current) setup.

I am the first to admit I have crippled my system. Its capeable of much better cooling than I have at the moment. I know temp readings from diods are notoriously wrong but I took a temp reading using good air cooling and took a temperature with my watercooling setup. Once I can keep it cooler than the benchmark set by the aircooling I am happy. BUT thats me.

The discussion started about low flow / small tubing V's high flow / large tubing. My system can never be used a comparision. Unless I am compltely wrong here the test will be run with a CPU block only so with all my extras we couldnt even run a comparision.

In my opinion (and again I emphasise its only an opinion) if you took an optimised (read not crippled like mine) low flow system with decent fans you are then in a situation where temperature differences will be small. Dial in a huge overclock and the difference will increase but I couldnt even take a guess as to how much. Thats why I would like see the test run.

Originally Posted by unregistered

knipex
you're tough, sparing with this bunch

Im Irish and tp that I grew up with a red headed mother and you can begin to understand why arguing is a national sport over here. All in good civilised fun of course. Life would be boring if we all agreed all the time.

Originally Posted by Cathar

Hmmm, it seems that a contradiction in goals has crept into the argument, with acceptable noise vs acceptable performance. The level of "acceptance" varies from person to person, and depends on one's overclocking goals.

knipex has just stated though that absolute silence is more important to him than cooling performance. In such a scenario absolute performance is always going to take a back-seat purely because one is not "allowed" to fit even moderately powered fans on the radiator, nor use any of the stronger/noisier pumps. Such goals are fine so long as the system is stable and meets ones stated goals, but to claim that it is within 1-2C of a high-end system without any hard evidence is stretching faith more than just a little.

I agree totally and hence my clarification above. I am in the middle of about three threads here and have a tendancy to drift so appolagies. I hope my comment above clears things up.

Originally Posted by Cathar

P.S. IMO, hard-drives don't need to be cooled with anything more than a low speed Papst 80mm NGL fan, which is what I use in my server case. That fan pushes a pathetic 19CFM or so in free-flow mode, is absoutely silent when in-case, and even when thrashing the disks (Seagate Barracuda's) they barely get warm to the touch. I can assure you that the disk drives are making more noise than the fan. For disk drives that do suffer from heat problems, these are typically noisy things too all by themselves.

With baracuda's I would agree with you. I used to use them alot but for speed I went for the raptors which a scorchingly fast but also loud and hot. (10K drives) To kill the noise I put them in an enclosure. To find a hardrive enclosure rated for 10K drives that doesn't scream isnt easy. Plus I got a good trade on the HDD block and it looks good.

Originally Posted by Cathar

Here is too where I get a little uncomfortable with 1-2C figures.

1-2C at stock CPU speeds is a totally different matter to 1-2C at very highly overclocked levels.

A stock AMD XP-M 2500+ (1833MHz/1.45v) under load is putting out under a third of the heat of the same CPU when it's being run at 2800MHz/2.00v, or even 2900MHz/2.15v.

What I'm saying is that small differences at stock and quiet settings take on a much larger scale of significance when pushing the (computer) system hard. I believe that many people far too easily dismiss the scale of the heat load problem when pushing a system very hard. When you collectively (pump, CPU, GPU, etc) start pushing >150W into a radiator you do notice rather quickly that ultra-low-noise fans just don't cut it any more.

Again I hope my comment above cleared this up for you. Those of us that fit low speed fans know its a tradeoff. Exactly the same trade off in aircooling. Lower CFM = less cooling. No argument.

Can I ask a question here ?
How much more or less of an effect (if any) would fitting low speed fans to a high flow / large tubing setup in comparison to a low flow / small tubing setup ??

Originally Posted by redleader

It sounds like you don't care about performance. With that in mind, why are you even here? No one doubts that quiet systems can be made to work, the (largely unrelated) question is how well can a low flow system be made to perform verses higher flow?

Lo Red didnt realise you were a member here.

About performace you are correct to a point. As long as its cooler than air I am happy. The whole quiet thing came from a tangent when I was talking about why I personally watercooled and the advantages I got. I admit it kind of derailed the thread and I appolagise for it.

As to why am I here.. well there are a few reasons but the main one is I knwo next to noting about the theory of watercooling but

Originally Posted by knipex

I am however open to ideas and want to learn so here I am.

Originally Posted by redleader

All data suggests "not very well". But this is not a concern to you, and only a concern to us because you posted what at best could be called questionable speculation as fact
.

I think this whole thread started so we could collect the figures and have facts. Once we agree what we are testing and how we are testing it (what criteria is going to apply to each system) we will have the data and the facts.

Sorry for the long reply and all the quotes but I am trying to eliminate any confusion created by me.

[knipex]

On a side note.

For the low-flow test what exactly were you proposing.

We supply pump, block, an tubing and we use a common rad and fans ???

Or do we supply a 240 rad and we use common fans ??

Or we supply a 240 rad and fans ?? (I assume you are no going to ask us to use low volted papst fans for a performance test...)

For the sake of interest in nothing else I would like to see the "middle type" included the convergance of the the two types, the AC cobra if you will that Pug offered..

What pump block and rad are you proposing ??

From my point of view (and remember I admit to knowing nothing) If we used a common radiator and fans then the only variables are block and pump (and obviously tubing) so noone can come crying afterwards....

[BillA]

"How much more or less of an effect (if any) would fitting low speed fans to a high flow / large tubing setup in comparison to a low flow / small tubing setup ??"

the reduction in air flow is the same, so the operating point of both will shift (to the right) causing a reduction in dissipation
the dissipation in the low flow rad is less than in the high flow (always), so while both are affected in the same proportion, the absolute magnitude of decrease would be greatest in the high flow rad

try this


and yes, the advantages of a higher flow rate in fact necessiate stronger fans - depending on the rad design of course
- only a single 'type' is shown in the above graph

[Blackeagle]

A comparison of both lower flow and extreme flow systems performance levels ( For the 3x areas of judging performance = noise, O/C & temps) sould be of great interest and value to all. Cost as Bill points out is also a 4th consideration, but is IMO, one area that can be very open to wide swings depending on how fugal a person is. Use of a auto's heater core, a Swiftech 6000 block (or second hand one) & a pump picked up with cost in mind and also perhaps second hand can reduce cost a lot, yet reach a high performance standard of cooling.

While knipex uses a number of water cooled elements that someone setting up a O/Cing / performance rig would not use, knipex is right in saying that refrigeration is the way to go if max O/C is the only goal. But most who are going for a performance system also include a GPU block to improve cooling & O/C of the vid card, and very few can afford a refigeration set up for both the GPU & CPU. And a extreme water rig cooling both the CPU & GPU can come real close to a system in which the vid card is air cooled but with the CPU cooled by phase change for overall system performance, and the water rig will still be much lower cost if a frugal person builds it.

In fairness I'd say a test of water cooling only CPU systems is to tilted in favor of the large line, high flow systems. It grants the ease of maintianing a extreme flow rig, without offering the lower flowing type system a chance to show the advantages of it's results with a more restricted loop due to more than one block.

May I suggest a GPU block be added to both loops, let it also be chosen to serve well in each system. And in order to maintain accurate testing data, only add in the GPU block as a secondary "last test" to see the results in both system types after testing of the loops with only CPU blocks involved. This adds in a further needed component for both of the two systems to be tested.

As this is my suggestion, I'll pay for, or provide, the GPU block for the large line/high flow system. I'm hope'n Pug or knipex , or anyone else interested, can provide the small line/low flow GPU block. This is of course only if pH will consent to add in this extra step to the testing of the two system types.

I'll grant that adding just one additional block does not come close to the restrictions involved with knipex's multi block rig, but is more fair to the low flow type system than a CPU only type system. And it's a two block combination common for performance set ups.


Looking forward to seeing what others think regarding this. Most of all ph.

[Cathar]

Originally Posted by Blackeagle

knipex is right in saying that refrigeration is the way to go if max O/C is the only goal. But most who are going for a performance system also include a GPU block to improve cooling & O/C of the vid card, and very few can afford a refigeration set up for both the GPU & CPU. And a extreme water rig cooling both the CPU & GPU can come real close to a system in which the vid card is air cooled but with the CPU cooled by phase change for overall system performance, and the water rig will still be much lower cost if a frugal person builds it.

A lad over at OCUK forums has an interesting thread categorising peak overclocks of various users. Of course the sample size is still "too small" to draw any strong conclusions, but there are the beginnings of a pattern:

http://server6.uploadit.org/files/weescott-OCUK.jpg

The original thread is here.

Fairlt clear distinctions between the top-end air and top-end water, but phase-change doesn't pull that much of a lead over top-end water, and unless given a particularly good CPU, phase-change isn't always getting ahead of water (except on P4's - although I suspect that if the IHS's were removed that the results would be a lot closer there too). Add in the costs of buying and running a phase-change setup and it is pretty apparant that it is truly a brute-force solution with fairly minimal gains.

Totally agree on the water-cooling CPU + GPU thing. The ability to water-cool the GPU as well and boost its overclock will typically play a much larger factor in the computer's perceived speed for gaming, than merely phase-change cooling the CPU and leaving the GPU air-cooled.

Phase-change's ability to super-cool hot CPU's cannot be doubted though, just the budget required to achieve such.

Arbitrarily sticking a GPU block in the test though seems rather odd. Fine if it's just for a "complete kit test". Not everyone water-cools their GPU.

[Edit: Fix quotage]

[redleader]

Fairlt clear distinctions between the top-end air and top-end water, but phase-change doesn't pull that much of a lead over top-end water, and unless given a particularly good CPU, phase-change isn't always getting ahead of water (except on P4's - although I suspect that if the IHS's were removed that the results would be a lot closer there too). Add in the costs of buying and running a phase-change setup and it is pretty apparant that it is truly a brute-force solution with fairly minimal gains.

Sort of agree here. Most of the phase change systems listed there are commercial units, which are basically the "kits" of the phase change world. They tend to be not as effective as many of the homemade units of comparable size for many of the same reasons dicussesd in this thread WRT to watercooling (style, compactness, cost as well as some new reasons like refrigerant choice). I think if you looked at homemade phase change units, you would see different results as many of these can do 20 or 30C colder for the same investment in power and often much less money.

If phase change is worth it or not is pretty debateable IMO. I'm currently very interested in it, but is a huge investment in time even compared to watercooling.

[Les]

Originally Posted by unregistered

"How much more or less of an effect (if any) would fitting low speed fans to a high flow / large tubing setup in comparison to a low flow / small tubing setup ??"
........
the dissipation in the low flow rad is less than in the high flow (always), so while both are affected in the same proportion, the absolute magnitude of decrease would be greatest in the high flow rad
.......................

I prefer Bill's Thermal Resistance curves for manipulation purposes
Same radiator as above :


For a 100watt load increasing the Air-flow from 0.11m^2/min to 0.88m^2/min :
At 2 lpm the coolant temp is reduced by ~31c
At 10 lpm the coolant temp is reduced by ~ 28c

[knipex]

I would have no objection to donating a GPU block if required, if fact I would be delighted to.

The more charts I see the more sense it makes to me (again I could be wrong) that we use a common rad and fans. This will also have the effect of eliminating noise levels from the debate but :shrug:

Bill and Les

From your comments can I take it that a low flow system is better suited to low speed quieter fans or would the real world effect make no difference to how Hi-flow and Low-flow systems would compare ??

[pHaestus]

That graph is packed so densely with useful information that it's scary!

Note that the Delta EHE falls only atop the 0.75m^3/min graph. That noise level is FAR above my threshold.

Note how the Panaflo L1A @12V performs slightly better than the Delta EHE@7V; I think I mentioned something about not bothering to undervolt stronger fans if noise/performance was important earlier in this thread already.

And talk about the diminishing returns on adding noise: Panaflo L1A is rated at 21 dBA rated while the Delta EHE is rated at 53 dBA!

Great graph!

[BillA]

knipex

"better suited"
this is a value judgement, made (implictly or explictly) by every consumer
- will depend on the design goals

the only 'right' answer is in terms of cost effectiveness (my technical assessment)
not considered is size, extra features - or noise
so I plot noise vs. dissipation too

pH
you should delist the first rad article and sub the ThermoChill review.

[pHaestus]

The more charts I see the more sense it makes to me (again I could be wrong) that we use a common rad and fans. This will also have the effect of eliminating noise levels from the debate...

Well by now it should be clear that if you reduce restrictions (which WILL increase flow) that radiator and waterblock performance WILL improve. Now the magnitude of that improvement is going to depend greatly on the parts that you are using in the total kit.

The original question (and still the most interesting one imo) is whether one can design a kit with 6mm tubing and quiet, low pressure pumps that cools NEARLY as well as one with larger tubing and higher pressure pumps. That is what has been claimed by many of the Europeans in essense, right? You said yourself earlier that you were only a degree or two off from the highest performance. Now there are a few ways your statement could be true:

1) It could be that the radiator you are using is so much more effective that your water temperatures are much closer to room temperature than with the radiators the "high flow guys" use.

2) It could be that the waterblocks you are using are so much more effective than "high flow" offerings that your CPU stays closer to water temp at 1-1.5LPM with that block than a "high flow" wb can achieve even at 8-10LPM.

3) "1-2C off from Best performance" could merely be referring to "with the same waterblocks and radiators" and not referring to "with any available watercooling parts". This could well be true; there is not more than a degree or so difference in the Innovatek Rev3 at 0.5GPM and at 2GPM; one would expect blocks destined for 6mm tubing loops to have thick baseplates and not respond especially well to raising flow rates.

I think testing the complete kits is a far more interesting approach to getting at how tradeoffs made in the name of convenience and noise affect performance.

Is it also reasonable for me to test my own personal cooling loop first off? It's nothing special really (FedCo 2-342, Laing D4, 2 Panaflo M1As, 1/2" Clearflex, LR Cascade). It is plumbed properly though and is IMO a pretty good example of what performance is possible if one has an eye to limiting flow resistance everywhere except the wb (where it's very useful).

//Edit:

you should delist the first rad article and sub the ThermoChill review.

No way I am delisting OCers article 481. That treatise (though it DESPERATELY needs the revised graphs) completely changed the way I thought about wcing and testing and put several observations I had already made in my piddling into the proper scientific context. If it weren't for that article I'd probably be...

more productive at work?
more connected with wife?

hmm left-handed compliments taste the bestest I am told

[infinity9]

With baracuda's I would agree with you. I used to use them alot but for speed I went for the raptors which a scorchingly fast but also loud and hot. (10K drives) To kill the noise I put them in an enclosure. To find a hardrive enclosure rated for 10K drives that doesn't scream isnt easy. Plus I got a good trade on the HDD block and it looks good.

If your raptors are getting hot then you have some serious flow problems in your case. Mine gets 1C hotter than my Seagate when I run it hard awhile. I think it actually idles lower because of the big heatsinks they have on them. Watercooling hard drives is a waste of time. The only drives that might even be considered would be the 15k seagate scsi drives. If you have a poor case, 99/100 times the cause, then watercooling your hd's is simply making up for a poor case design. We really should address simple things like getting a good case for airflow before we start adding watercooling.

[Blackeagle]

Originally Posted by Cathar

The ability to water-cool the GPU as well and boost its overclock will typically play a much larger factor in the computer's perceived speed for gaming ......

Arbitrarily sticking a GPU block in the test though seems rather odd. Fine if it's just for a "complete kit test". Not everyone water-cools their GPU.

This will be a test of two differant "system" approaches to water cooling of a rig. Pumps chosen along with blocks and rads are often differant than those used to build a DIY performance rig. So a addition of the GPU seems reasonable to get a added bit of perspective and information. knipex's system with it's many added water cooled points, has I think, a number more water cooled elements than most do in low flow systems. Yet the use of a higher average number of water cooled components does seem to be a trait of low flow rigs, from what I've read of differant peeps systems. (please correct me on this if I'm in error.) And some performance peeps include the NB in rigs as well, although most do not. So the addition of a GPU, seems to make for a "fairer" and more complete test between the two system types.

Other means to add restriction to the loop could be used to simulate the use of added blocks effect on flow rates, but such would not include the added heat to the system of the actual extra blocks.

But as you point out, not everyone water cools the GPU, which is why I suggested it be left as a last test so as not to distract from the CPU only results.

I'm also interested in seeing the selection of the pumps, rads, fans & max line size to be allowed and used.



BTW,

I enjoyed reading this thread much more than the earlier one regarding this sort of comparison.

Edited:

To remove redundant ideas, comes of posting without reading what was posted while away from my puter.

[freeloadingbum]

Originally Posted by Cathar

Which brings us back to the original challenge.

Can a waterblock at low-flow rates (~1-1.5LPM), as would be typical of a system coupled with 6-7mm or ¼" ID tubing, and, using with a low-powered quiet pump (<1.5mH2O peak pressure, <10LPM peak flow) provide block-level performance that is within 0.02C/W of a 6+LPM flow rate system, as is typical of a ½" ID system when using a well powered pump (>3.0mH2O peak pressure, >10LPM peak flow rate).

JoeC's results show the hydrostream HS5 performing .134c/w at .3gpm while the mcw6000 performed .136c/w at 1.5gpm

[Blackeagle]

pH,

Your posted system makes use of the Cascade, matched by no other, but no longer available for purchase, so will limit how usefull the resulting data will be to those reading it. Rest of your rig would be great. Do you still have the Dtek White Water, RBX or 6002? They are all top performers and also still available to buy. Also, what do you think of the idea of adding in a GPU block as "add on" last test?

Unless Cathar would like to send you the latest version of the Storm block? Could also give you the chance to be the first one to test and post results of the STorm block in a seperate article, as well as make use of the very latest & best of high flowing performance blocks.
The Storm could also be of use in the low flow rig if it's design allows for it, as I suspect it willl.

[BillA]

Originally Posted by freeloadingbum

JoeC's results show the hydrostream HS5 performing .134c/w at .3gpm while the mcw6000 performed .136c/w at 1.5gpm

yes, this IS a low flow wb
and can perhaps be selected to win the point by the low flow group
but this is not the wb/system design that is the basis for this discussion

[Blackeagle]

JoeC's results stop short of where they should.

For the quality of his test bench he tells relatively little.

What would the results be of the HS5 at x4 or x6 flow rates ect.

How many mountings per block tested does he do now?

Edit:

beat'n to the post.........

Good points I didn't think of as well Bill.

[BillA]

the designer of the HS5, Herr Montag (sp ?), on OCAU had some comments
to the effect that with microchannels there was only so much fluid that can be pushed (at any pressure)

it is a low flow wb
is it commercially available ? as in: any one bought one ?

yes, JoeC could provide more info - but he has no desire to test WCing gear full time

[pHaestus]

How many mountings per block tested does he do now?

JoeC rather cleverly gets much more reproducible mounting pressure than I do with springs and wingnuts (though I am now getting good )

He presumably does at least 3 mounts because he reports a std dev.

[knipex]

Originally Posted by infinity9

If your raptors are getting hot then you have some serious flow problems in your case. Mine gets 1C hotter than my Seagate when I run it hard awhile. I think it actually idles lower because of the big heatsinks they have on them. Watercooling hard drives is a waste of time. The only drives that might even be considered would be the 15k seagate scsi drives. If you have a poor case, 99/100 times the cause, then watercooling your hd's is simply making up for a poor case design. We really should address simple things like getting a good case for airflow before we start adding watercooling.

Sorry but I have allready tried to explain this twice.

The drives are hot becuase they loud. To get around this they are in a soundproofed enclosure. An enclosure means no airflow. I looked for a harddrive enclosure rated for 10K drives and got a good option on the watercooled one. If it was not in an enclosure I wouldnt be cooling it.

Baracudas run quiet so I did not need to put them in the enclsoure so I did not need forced cooling.

Originally Posted by Blackeagle

BTW,

I enjoyed reading this thread much more than the earlier one regarding this sort of comparison.

As did I



So can I assume we will supply the rad and fans with the blocks and tubing. Also will we be including a GPU block in the test ??

I like the idea of simulating multiple blocks (or even fitting mutiple blocks) just to give another perspective. (but then I would)

Now for the problem. Pug does not stock a complete kit that fits all your requirements. Would you settle for parts from the same manufacturer ??
Also would a 1048 pump satisfy your requirements ??

Secondly I havent spoken to Pug about this yet (if you have been Ph then I will leave it up to ye to decide.)

[knipex]

Originally Posted by unregistered

knipex

"better suited"
this is a value judgement, made (implictly or explictly) by every consumer
- will depend on the design goals

the only 'right' answer is in terms of cost effectiveness (my technical assessment)
not considered is size, extra features - or noise
so I plot noise vs. dissipation too

Sorry Bill I missed this.

My question was this.

Lets assume someone was going the route of watercooling for noise reasons and planned on running low volume low noise fans which solution would work better 1/2" or 6mm ??

Ph would it be possible to try this when you are running the test ?? I for one am curious...

Here is my theory (admittadly uneducated).

In a low flow system water will spend longer in the rad. This will give the water more time to cool. Therefore less air flow will still cool the water.

[Blackeagle]

Big PlayerMaking Big Money??????????

What's up with that pH?

Last I knew it was more like you were spending Big Money........on testing gear. ???

Another ...........uninformed comment on another site????

LOL!!!

[knipex]

Originally Posted by Blackeagle

Big PlayerMaking Big Money??????????

What's up with that pH?

Last I knew it was more like you were spending Big Money........on testing gear. ???

Another ...........uninformed comment on another site????

LOL!!!

Long story best forgotten.

One missinformed poster problem resolved. One good thing it did result in this thread...