A challenge!

[BillA]

where is this going now ?

high flow vs. low flow ?
or
high flow kit vs. low flow kit

IF it is a kit (and Swiftech has no 'interest' in this as 3/8" ID tubing is used), then it will need those things that kits provide (inst, cust. serv., warranty, etc.)

jeez, why not drop the 'kit' requirement ?
who offers a 6mm kit ?

[Blackeagle]

Bill I thought the kit thing was a dead deal as well.


I"m sure in favor of components chosen at both sites to best perform as a total "system" but not as a system pruchased as a kit.

That was a one problem in the thread that Pug was involved in a little while back.

[pHaestus]

Quite separate issues I think Bill getting intertwined:

1) This challenge (carefully worded by me, eh?) is for the wizdforums folks to back up their assertions made in that (now deleted/cleaned up) thread. It has mostly done it's job; looks like everyone is backing away from that "cools just as good" nonsense and migrating towards "I don't care how it cools because it's quiet" statements. Which is fine.

2) Procooling is going to begin testing kits; I don't really care for mix and match kits from vendors and would prefer to test only kits offered as such by mfgrs.

Pug and Wizd Designs have offered to send parts from AlphaCool for testing/review. I suggested that the parts should be chosen so that they resulted in one complete AlphaCool kit I think this is what he was going to send

To get at the "low flow vs high flow" thing and to further this discussion, Pug also offered to send the following stuff to swap into the above kit:

"Exchange barbs for 10mm plugin, 7mm ID tubing, AP900 instead of 1500 and HTF (continuous tube) rad instead of BIX"

For the most part we could crunch the numbers on this stuff w/o doing a lot of testing given your Thermochill 120.2 results and a (provided by me) dP vs flow curve for the AlphaCool block(s).

If we drop the kit testing then I can just return to the wb tests and leave all the noise testing for others to deal with too. I assumed that there was a fair bit of interest in a BIG scale "state of the industry" cooling kit comparison though??

[knipex]

I for one would like to see the test go ahead.

I still contend that low flow will cool decently I have a post on Wizd laying out my thaughts but I will copy it over here for those of you who dont want to register.

I have a feeling that you are under estimating ProCooling. I agree that some posters canot see beyond cooling performance but most see both sides.

I have a feeling that testing will go ahead IF we can agree on kits.

What i would like to see tested is.

Performance using decent fans on both systems loop with and without GPU block. (loop without waterclock will have no interest to 99% of us but it does to alot of procooling members).

Performance of both systems with quiet fans fitted. Same test as above. (I would like this to get as low as 7V Papst but I cannot see ProCooling going for this but I will ask)

Performance of both systems with extra blocks added. (tested as above)

On the results. On the CPU only I expect us to get beaten. As the overclock increases I expect the diference to get worse. The higher the temp the more they beat us.

With the GPU block included I see the results getting closer.

If I can get them to try "quiet" fans I expect us to give them a surprise.

The multiple block test isnt realy a fair one, its not suited to high flow but then again CPU only isnt really suited to low flow so fair is fair.

Any recomendations on the blocks we should submit. Remember 1048 pump.

[BillA]

Originally Posted by knipex

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 ??
. . . . .

far too simplistic a question to assess the impact of fans and tubing size

start with noise:
list all sources and levels
identify those which can be altered
so where does WCing fit ?
the fan on the hsf went to the rad - presumably bigger and slower making less noise
but a pump is added, how much noise ?
-> note that the pump size IS related to the tubing size, so for a low noise WCing solution the 1st choice is the pump - which if based on noise will be small
- so the line size is small

do you understand that a small pump and fat tubes are incompatible ?

so now you have small tubing and a small pump, and large tubing with a large pump
- now one could put a big pump on the small tubing, but to what purpose ?

now the rad size can be addressed, for Swiftech this is predicated on installation convenience, you would consider noise
-> the only low noise solution is a thin rad, the bigger the better
-> a thick rad with a low noise fan will be outperformed by a thin rad with the same fan

so the only true low noise solution uses a silent pump, which requires smaller tubing

ah, but how big can the small lines be ?
well, again it depends on the pump - but not less than 8mm ID for a low noise Swiftech kit

[knipex]

Sorry Bill. I meant to say low flow / 6mm or high flow / 1/2"

I guess it is a subjective comment.

Originally Posted by unregistered

do you understand that a small pump and fat tubes are incompatible ?

Yup I though that was pretty obvious.

On the issue of pump noise.

As stated I run a 1048. Its mounted on 4 rubber anti vibration mounts. Silent unless I have my ear within inches of the pump.

My question about low flow being better suited to slower / quieter fans was based on my earlier question about which would suffer more. The answer seemed to be that high flow would.

Can I take from this that for silent cooling that low flow is the way to go ??

Also my assumtion the fact that multiple blocks are not really suited to high flow (read 3 blocks minimum) is this correct ??

I am open to change and am considering a high flow system on my next system.

[pHaestus]

1) You will never improve cooling performance by decreasing flow rate
2) You can only increase flow rates up to a certain point by going with larger tubing; above that point air will not clear the lines

[redleader]

Originally Posted by knipex

Sorry Bill. I meant to say low flow / 6mm or high flow / 1/2"

I guess it is a subjective comment.



Yup I though that was pretty obvious.

On the issue of pump noise.

As stated I run a 1048. Its mounted on 4 rubber anti vibration mounts. Silent unless I have my ear within inches of the pump.

My question about low flow being better suited to slower / quieter fans was based on my earlier question about which would suffer more. The answer seemed to be that high flow would.

Can I take from this that for silent cooling that low flow is the way to go ??

Unless the 1048 is much weaker then my 1250 (and i've read that this is not the case and so i will assume so even though i only have experience with the 1250), up to 1/2 would be an option with that pump. Given that 1/2 will perform better with that pump then smaller tube sizes while adding no additional noise, I think its pretty obvious that 1/2 inch is the better option.

If the rest of your system will accomidate it is another matter.

Also my assumtion the fact that multiple blocks are not really suited to high flow (read 3 blocks minimum) is this correct ??

??? The number of blocks is irrelevent. Assuming they'll take a larger tube size, and that your case can fit it, then the larger size will perform better at a given noise level. Again assuming all parts in your system can accomidate the larger size.

----

2) You can only increase flow rates up to a certain point by going with larger tubing; above that point air will not clear the lines

I'm not sure i agree. Given an effective res you should be able to clear lines of any thickness, provided you don't mind waiting a while.

Or am I misunderstanding you?

[BillA]

silent, low noise - jeez, we need to put numbers on these things
silent is easy to understand (inaudable),
what is 'low' (and, more importantly, at what level of background noise) ?
N.B. there are big low noise pumps
-> if a big pump meets the 'low noise threshold', then a big pump and fat hose CAN be a very effective low noise solution
- so the answer to your persistent tubing question lies with the initial pump selection

low and high flow rate - same problem, except few know the system flow rate
suggest avoiding the term flow rate

re wbs: 1 in series plus 2 in parallel - works fine

[pHaestus]

For my purposes PWMing a 12V pump along with 2 120mm fans has been working very well. It's controlled by a MatrixOrbital LCD and linked to a DOW sensor epoxied under the CPU core. Idling/wife using PC for windows/whatever it throttles everything down and the HDDs are the loudest part of the system. Under load everything goes into "full speed ahead" mode and it gets too loud for full time use. If I want, I can turn it down (in the winter when I am running a distributed computing app 24/7) and temperatures suffer a little. If I want, I can run it wide open just to overclock or mess around with it. I am willing to sacrifice a little bit of noise for a lot more flexibility.

[pHaestus]

So knipex brings forward a rather interesting proposition really; the best idea I have seen in this thread. This is actually a very good idea that would spawn a great article I think:

1) Wizd forums puts together their ideal cooling system using their principles: smaller pumps, smaller tubing, low flow rates, quiet operation
2) Procooling forums does the same thing
3) I test them both out, and then I make any changes/substitutions that I would personally do to both and test again

We would have to both cool the same amount of items (CPU+GPU seems to be reasonably popular in both places), and the radiators should have similar facial surface area (sized for 2x120mm fans would be my recommendation).

I would limit the part choice to currently available commercial waterblocks (so no Cascade SS guys sorry) and would prefer to stay out of parts selection on the Procooling side of things. That way I should have some comments/changes to make for both the Procooling and Wizdforums cooling loops.

Evaluation would presumably be based upon:
Noise
Performance
-CPU cooling
-GPU cooling
Appearance/Quality ??
Don't think ease of use really matters as these are both DIY setups at this point

This would be more fun than arguing in the forums, wouldn't it? Note that wizdforums could easily put together a kit that looks a lot like a US system: 10mm ID tubing, larger pump, etc. to improve performance. That would be, in my opinion, conceding defeat on this whole "low flow is just as good" argument. Procooling could also make a lot of concessions to lower noise that would degrade cooling performance. Gotta be careful there or your foes would end up at that "1-2C difference" number that, in my opinion, would be a win for the "smaller is just as good" argument.

Does this sound like a worthwhile expenditure of effort?

//edit: Perhaps if we were to set up something like this with a clearly defined goal then the "winning" side could have both cooling loops sent to them when testing is complete? Like racing for car titles

[knipex]

Ph

That sounds like fun.. But if we won what would we do with all those fat pipes and if Procooling won what would ye do with all those plug & cool conections

[pHaestus]

Well....
I guess I'd get a case painted pink with a flower-shaped window and install your "girly-man cooling" into that

j/k j/k

[dutchcedar]

That sounds great, pH. The results would be worthy to anyone interested in water cooling... no matter what the outcome is.

[knipex]

Ph

I am curious as to how you would define a win.

Looks are subjective.
Performance so a forgone conclusion the only question is the margin.
Sound levels are quantifiable but how would you trade dB for Degrees ???

[Cathar]

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

Sorry - you made a mistake somewhere.


For the HS5

At 0.3gpm the HS5 is 1.2C warmer than at 1gpm.

At 1gpm, at 70W heat load, the die rise must have been 70 x 0.122 = 8.54C

So at 0.3gpm, 8.54C + 1.2C = 9.74C

9.74C at 70W is a C/W of 0.139

For the MCW6000

At 1GPM the C/W is 0.141. At 1.5 GPM it is 0.5C cooler.

70 x 0.141 => 9.87. 9.87 - 0.5 = 9.37

9.37 / 70 => 0.134

I admit that they are fairly close, however the MCW6000 is not a true "high flow" block, it really is an intermediate block, something that straddles the middle ground, and quite intentionally by Bill too I believe.

Pressure Drop

However, if we start to take pressure drop into the equation, and compare to, say, the Danger Den TDX. Refer to math above for the C/W's derived below.

TDX @ 1.5gpm => 0.123 C/W
TDX P.D. @ 1.5gpm ~= 0.76mH2O (rough approximation)

Hydrocool HS5 @ 1gpm => 0.122 C/W
HS5 Pressure drop @ 1gpm => 2.47mH2O (measured by JoeC)

Gets interesting doesn't it?! The HS5 requires over 3 times the pressure drop to perform even as well as the DangerDen TDX.

Okay, let's drop the HS5 back to 0.5gpm and look at the figures there:

HS5 @ 0.5gpm => 0.133 C/W
HS5 Pressure drop @ 0.69mH2O (rough approximation)

Look at that! At about the same P.D. the HS5 is actually quite a way behind the TDX. Now look at the TDX results at 1.5gpm next to a number of other blocks in Phaestus's tests.

Look, the HS5 is a good block, no doubt about it. It does represent what is good about micro/mini-channels in that it has nice low-flow properties. It is perhaps the only block out there that we have enough evidence to show that it does fit the bill of providing within 1-2C of a high-flow setup's performance against certain blocks, but it does come at a very high pressure drop cost that MUST be not be ignored.

What we do need to be very careful of JoeC's OC.com results is that he does give you enough information to make a decision, it's just that it is more than a little confusingly presented, and for the average person, almost misleadingly presented.

Given the relative level of (un)availability of the HS5, and that it is silver, perhaps a more fair comparison would be when I get the silver version of my Storm block out and about in a couple of months, or perhaps at the very least, comparing it next to a Cascade SS and do so with taking pressure-drop into account.

[pHaestus]

I would say that if your "low flow engineering over brute force" kit performed within 1-2C of the brutish American jobs then that is clearly a victory. Noise/performance ratio is also pretty hard to deal with because everyone's threshold of annoyance is different. Surely we can iron out something though. The "winning" and "losing" is sort of incidental as it would be a really good chance to discuss design choices and how to improve cooling/decrease noise levels for the DIYer in any case. But it would make the process more interesting, wouldn't it?

[BillA]

a rather elegant test regime could be defined if the design parameters were considered as a whole instead of a division between 2 'types'

[freeloadingbum]

Originally Posted by Cathar

Sorry - you made a mistake somewhere.

As I understand it, JoeC uses 70 watts to compare with 100 watts cpu load according to "Radiate". I thought that this was what BillA was doing too with his 70 watt loads as well. Maybe I'm wrong but if that's not what he is doing then why not just use a real 100watt load with his tests.

[Cathar]

Originally Posted by freeloadingbum

As I understand it, JoeC uses 70 watts to compare with 100 watts cpu load according to "Radiate". I thought that this was what BillA was doing too with his 70 watt loads as well. Maybe I'm wrong but if that's not what he is doing then why not just use a real 100watt load with his tests.

That may be, but regardless of the relationship being attempted, the C/W figures presented by OC.com are relative to an actual 70W, not a fantasy "radiate" 100W, and you simply cannot just substitute 100W into there to derive a fantasy "radiate" C/W.

[freeloadingbum]

Originally Posted by Cathar

Look, the HS5 is a good block, no doubt about it. It does represent what is good about micro/mini-channels in that it has nice low-flow properties. It is perhaps the only block out there that we have enough evidence to show that it does fit the bill of providing within 1-2C of a high-flow setup's performance against certain blocks, but it does come at a very high pressure drop cost that MUST be not be ignored.

At .3gpm pressure shouldn't be an issue. One of the advantages of low flow blocks is that most pumps ability to deal with pressure increases as the flow rate is reduced. I would expect a 5 watt c-systems pump to be able to produce .3gpm with this block and a doublefan rad

[freeloadingbum]

Originally Posted by Cathar

That may be, but regardless of the relationship being attempted, the C/W figures presented by OC.com are relative to an actual 70W, not a fantasy "radiate" 100W, and you simply cannot just substitute 100W into there to derive a fantasy "radiate" C/W.

You're missing the point. the hdrocool block is 13.4c (at .3gpm) over water temp with 70 watt load while the mcw6000 is 13.6c (at 1.5gpm). At 100 true watts the hydrocool block will have a higher improvement over the mcw6000. (.286?). Its not my fault they did it that way with regards to Radiate. Blame BillA and JoeC. Its their fantasy, not mine.

[BillA]

no, you do not understand what is being manipulated - public perception
and apparently you are not too familiar with the notion of thermal resistance, aka C/W
C/Ws are flat, 40W to 225W (my current max.), same value; so the heat load at which it is measured is moot

70W (actual, IxE) was selected by me several years ago to nominally equate to a 100W CPU heat load as characterized by Radiate
this nominal equilivance has been 'verified' by several others

so if JoeC tested with 70W, be assured that he also used 70W in his calcs

[Cathar]

Originally Posted by freeloadingbum

At .3gpm pressure shouldn't be an issue. One of the advantages of low flow blocks is that most pumps ability to deal with pressure increases as the flow rate is reduced. I would expect a 5 watt c-systems pump to be able to produce .3gpm with this block and a doublefan rad

What I think you're really getting at here is trying to use the unavailable silver HS5 as proof that a low-flow system could get to within 1-2C of a high-flow system.

Let's use your example and assuming a CSP on the HS5 with a heater-core, some small ID tubing that is contributing to the total pressure drop, and the system as a whole is pushing 0.3gpm.

Let's assume that the radiator is a Thermochill 120.2 for convenience sake since we have some good data on it, and that the air-flow rate is a moderate 2m³/min with some tolerable noise fans (perhaps a pair of Panaflo L1A's)

Now let's compare that to what is a fairly decent example of the "high flow" approach, being a TDX with 1/2" ID tubing with a Laing D4 and the same radiator. The predicted flow rate of the TDX system is about 2gpm.

So where does that all fall?

Using this graph for the radiator:

At 0.3gpm (1.1lpm) and 2m³/min the radiator has a C/W of (5C/115W) => ~0.044
At 2gpm (7.6lpm) and 2m³/min the radiator has a C/W of (5C/160W) => ~0.031

So already there's a 0.013 C/W advantage to the high-flow system.

From JoeC's tests, the HS5, at 0.3gpm has a C/W of 0.139
From JoeC's tests, the TDX at 1.5gpm has a C/W of 0.123. Let's be a little pessimistic and extrapolate this to 0.119 C/W at 2gpm.

So the TDX is ~0.02C/W in front of the HS5.

Add that up, and the 0.02 + 0.013 = 0.033 advantage to the TDX.

For a typical unoverclocked system, that would border at around a 1.5-2C difference.

For a hot overclocked system, that would be closer to a 4C difference.

I now guess it depends on whether you think that 2C on an unoverclocked system, where the level of cooling being applied in both scenarios is FAR more than adequate for stability is important, or whether when overclocking that the differences grow to ~4C or so where such a difference is undeniably significant and important.

[Cathar]

Originally Posted by freeloadingbum

You're missing the point. the hdrocool block is 13.4c (at .3gpm) over water temp with 70 watt load while the mcw6000 is 13.6c (at 1.5gpm). At 100 true watts the hydrocool block will have a higher improvement over the mcw6000. (.286?). Its not my fault they did it that way with regards to Radiate. Blame BillA and JoeC. Its their fantasy, not mine.

No, I'm not missing the point at all.

It would behoove you to take the time to understand what is meant by C/W, what is actually being measured by JoeC, and how to use that to extrapolate his values to the different flow rate settings he's measured.

Please do reread by post above about interpreting JoeC's results. It is clear to me that you have been mislead by the way in which they have been presented. I do not blame you for this, but please re-read my post above before commenting further, and it would help you to drop the 100W figure from your mind while doing so. We are manipulating C/W's here.