which Radiator for a cpu & gpu system

[Magicuser]

hi i have a pentium d 805 and a gefroce 6800gs and ive allready picked out the water block for them but i have yet to figure out an Radiator config or for that matter a Radiator at all i was thinking about a Black Ice GT Stealth 240 Radiator but im not sure if it is to big i what to overclok my pentium d 805 to around 3.6~3.8ghz range i dont know if this helps but here is the parts i have so far ( i already have a pump and what will serve as a resavior):

Swiftech Apogee-1U 775
Danger Den Acetal MAZE4 GPU
Clearflex 60 Premium Tubing 1/2" ID x 3/4" OD
Fluid XP EXT Extreme Non-Conductive Coolant - UV Midnight Blue 32oz
swifttech mcb-120 radbox

thanks for that help.

[NickS]

The Swiftech MCR-220 would be a good choice. The Black Ice GT 240 is nice too. Good luck :)

[Dragoon42]

I'd pick the mcr-220 over the GTS. I haven't really read too many good things about the GT 240

[Magicuser]

thanks i think i will go with the black ice gts thanks btw this is my first water cooler thanks agian

[BillA]

http://translate.google.com/translate?u=http%3A%2F%2Fwww.cooling-masters.com%2Farticles-38-0.html&langpair=fr%7Cen&hl=en&newwindow=1&safe=off&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools

EDIT: corrected link

[freakdiablo]

why go with just one? thats the great thing about watercooling:expandability. at as many radiators as needed, then when your pump gets overloaded, get one with higher output

[Senater_Cache]

wow, welcome back bill. great link.

[voigts]

the best pc radiators available are the Thermochill PA series.

[BillA]

not correct, just the most expensive
all rads of similar size and construction perform similarly
dig deeper

[FIZZ3]

http://translate.google.com/translate?u=http%3A%2F%2Fwww.cooling-masters.com%2Farticles-38-0.html&langpair=fr%7Cen&hl=en&newwindow=1&safe=off&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools

EDIT: corrected link

Nice article. Seems like they have used your data to put together a comprehensive review.

Maybe in one sentence for those who didn't read the review: the GT240 is only 'good' if you use two 86CFM fans at full speed (I have one fan near that rating and I find it too loud at 12v). If you have less airflow, it likely offers no benefit over BIX2-type radiators. It is also quite restrictive, but at least your Apogee block is not.

Redeemer: rads of the same rough size (i.e., 1, 2 or 3 fans) perform very similar in a real world system.

[nikhsub1]

not correct, just the most expensive
all rads of similar size and construction perform similarly
dig deeper
Bill, can you show us where the PA was tested against a CW rad of 'similar' size, FPI and headloss characteristics?

[BillA]

see Roscal's article
the CoolingWorks CR-22T and CR-32T are the same construction as the PAs
(NOT the CR-12T as it is 18FPI)

see the Swiftech article for similiarity of types

[[O-CuK]Marci]

Comparison tests end up with the usual "obviously Thermochill outperforms, it's thicker, so unfair comparison." Therefore, it can't be compared... apparently. *Shrug*

How do you compare a rad that has no "equals" when it comes to size, FPI and headloss??

Redeemer: rads of the same rough size (i.e., 1, 2 or 3 fans) perform very similar in a real world system.


ADDENDUM - JULY 2006 - Further independant comparative results vs other rads - Testing Performed by Radical_53 on behalf of HardwareLuxx Magazine - Germany

1 - Pressure Drop - flow set at 4lpm at pump, pressure measured after radiator. % loss shown - smaller the loss, better the head pressure... listed best to worst.

Least Pressure Drop...
ThermoChill PA120.3 - 6%
Watercool HTFX3-x - 14%
HWLabs BlackIce GT360 - 17%
HWLabs BlackIce GT360 XFlow - 17%
HWLabs BlackIce Xtreme - 17%
HWLabs BlackIce Xtreme XFlow - 17%
Watercool HTSF - 17%
Cooltek MCR320 - 18%
NexXxos Xtreme - 23%
NexXxos Pro - 27%
...highest Pressure Drop

2 - Performance order (best to worst) based on final coolant temp, cooled with YateLoon 120mm fan - lower the "k" figure (difference between coolant temp and air temp), cooler the coolant, therefore better the performance...

At 600rpm:
Thermochill PA120.3 - 7k
Watercool HTSF - 8k
Watercool HTSF3-x - 8.4k
Cooltek Maxistream / Swiftech MCR - 8.4k
Black Ice GT - 9.2k
Black Ice Extreme - 11k
Alphacool NexXxos Pro - 11.2k
Black Ice GT X-Flow - 12.1k
Black Ice Extreme X-Flow - 12.3k
Alphacool NexXxos Extreme - 14k


At 900rpm:
ThermoChill - 6.2k
Cooltek - 6.9k
BIGT - 7.2k
WC HTSF - 7.3k
BIGT-XFlow - 7.3k
WC HTSF3-x - 7.4k
NexXxos Pro - 7.5k
BIX - 8.5k
NexXxos Extreme - 8.5k
BIX-XFlow - 9.5k


At 1200rpm:
ThermoChill - 4.2k
Cooltek - 5.6k
BIGT - 5.7k
BIGT-XFlow - 5.9k
BIX - 5.9k
NexXxos Pro - 6.1k
NexXxos Extreme - 6.1k
HTF3-x - 6.3k
HTSF - 6.3k
BIX-XFlow - 6.6k

*shrug*

Define "similar"... in all other comparisons online (inc Bill's data used for Roscal's article) ThermoChill weren't approached to provide a rad to include. The testrig used for Roscal's article that Bill links to no longer exists, so once again, no direct results on the same testrig exist other than Radical53's above...

General consensus seems to be that cos it costs so much more than the others and doesn't comply to "similar" dimensions that it should be omitted / ignored.

I've given up trying... folks'll buy whatever they wanna buy. Leave em to it. Why the hell should we care?

[[O-CuK]Marci]

Confusing me bill...

Coolrad 22T
Double Row, optimized density louvered flat tubes for optimal heat dissipation at low noise levels (12 fins per inch)

PA has lower FPI, larger tubes, 3/8" BSP inlet and outlet, and 22T falls 19mm short on overall depth...

[nikhsub1]

I tend to agree with Marci here... Porshe/Audi/VW are the 'same' construction, too. I know which I'd rather have :D

[roscal]

Major difference with others rads is the extreme depth used in Thermochill rads. Even if the FPI is ~equal to MCR220 or CR-22T and process is identical, there's a little doubt about their performance because flat tubes are very deep (how much exactly Marci ? Tubes dimensions ?). We can reasonably think that more perfs is possible with that depth and the low FPI, airflow will be hurt a bit for sure, but surface will probably compensate well in this case. Bill is a bit expeditious in his decision ;) .

Thermochill PA are probably the best in performance at low-moderate airflow, but cost and oversized rad are 2 drawbacks. As I said and unfortunately, ratio perf/cost or perf/size are not in favor of Thermochill rads. Generally, people who want to buy a rad ask themselves these questions. Ex: Is 50% more cost worth 1-2°C lower at full load ? Can I integrate it in my tower ? Etc. Perf is one thing, but not the only thing ;)

Huge difference, no ? :D

http://www.cooling-masters.com/images/articles/gts240/images/taille.jpg

[nikhsub1]

Generally, people who want to buy a rad ask themselves these questions. Ex: Is 50% more cost worth 1-2°C lower at full load ? Can I integrate it in my tower ? Etc. Perf is one thing, but not the only thing ;)


Yes, yes, yes and yes :D I do think Bill is jumping to the wrong assumption here as well Roscal.

[roscal]

It's a pity that Bill didn't ask Marci a PA120.2, it would be interesting for comparison, but originally it wasn't decided to be an article, data were for CoolingWorks use. Another time...

[Dragoon42]

Yes, yes, yes and yes :D I do think Bill is jumping to the wrong assumption here as well Roscal.


Lol when I read that post I was going to ask him the same questions that you've asked. But it's already been done.

Anyways, OP, if you really have the budget, just pickup a PA120.2

I'm in the process of getting a new WC setup and i'm going with the MCR-220. But this is only because it's right next to my AC vent, and I don't plan on moving for a long time. If not I'd pick up a PA20.2

Yes this is the same as my heater, but i live in Alabama, so we keep the house around 70f year round. I have it like 1-2 inches away from the vent. It keeps everything really cool.

[[O-CuK]Marci]

Roscal - specs have never been published publicly, we had to move the design discussions out of public eyes due to Chinese interference (cloned the PA160 before we'd even released it)... will send you them by e-mail tomorrow. FPI is most definitely lower than both products Bill references... as is inlet/outlet bore...

Cost/perf ratio is mainly an issue outside of EU more than anywhere else. Within UK, the price difference is literally a handful of £'s... so all depends from what country you're looking at it for the price issue - altho this has been battled out all over the place... we didn't particularly ask for folks outside UK to buy our rads - when they were first released back in HE days it was solely for the UK market... they chose to buy of their own volition. Obviously there are those out there who purchase on price/perf, however, there are still those who purchase purely on performance. Should this mean the product line is discounted / overlooked??

The StormG4 still sells in high numbers, despite it's price/perf being over the competition...

Look at the GPU market.... price/perf equally skewed... and the CPU market. Is this not something inherent to the "performance hardware" within the field that we all choose to reside?

One could debate the pump price/perf argument...

[BillA]

3 hr phone call caused the delay in my response . .

Hi Marci,

2 of these 'test rigs' exist, Swiftech has one and I the other (which is why the MCR-220 was tested with its Delta M fan, to confirm the similarity of results). Mine, shown in Roscal's article's photos, is presently crated as I am preparing to move.

You have in the past inquired about my testing your rads; but as I was at that time affiliated with CoolingWorks (no longer the case), and as it has been demonstrated over and over that professional objectivity and integrity is not going to be afforded me, I declined. (http://www.hardforum.com/showpost.php?p=1028738150&postcount=69 )

The collaboration with Roscal for his article occurred after the fact, I was testing CoolingWorks rads and tested those which I was sent. After several discussions it became apparent that the data was of interest so I sent Roscal the spreadsheet.

Some comments on ‘similarity’
Re tube size: the BIP and BIX have the same tubes, the BIX having an additional row; the CoolingWorks the same as the BIX; and the ThermoChill HE is also similar to the BIX (by external examination). I was/am not aware that the PE has different sized tubes.
The Swiftech rads have a unique tube size.

Re fin density: the BIP and BIX are 18 FPI (nominal), as is the CoolingWorks CR-12T, and also the ThermoChill HE (as I recall). The Swiftech are 14 FPI (nominal) and the CoolingWorks CR-22T and CR-32T are 12 FPI. Apparently the ThermoChill are below 12 FPI.

Re connection size: the size of the threaded hub in the header means nothing, it is the ID of the barb screwed into it that contributes to the head loss; this will vary depending on the supplier.

WRT ‘similarity’ and performance what must always be understood is that there is no BEST tube size, fin density, or any other single parameter. The number of rad elements is large, and they all are interactive. But loosely speaking, as we are doing here – no numbers eh ?, different sizes can be grouped, tube rows can be grouped, and fin density can be grouped. It requires no leap of faith to say that rads of similar construction will have similar performance, for that is what the data says.

What would be difficult to accept is that a rad of similar construction had significantly different performance. I can, and do, say this because I spent almost 9 months testing combinations and permutations of rad design variables, it takes a substantive change to yield a substantive difference.

I decline to comment on Radical_53’s testing, presentation, or analysis.

[[O-CuK]Marci]

You have in the past inquired about my testing your rads; but as I was at that time affiliated with CoolingWorks (no longer the case), and as it has been demonstrated over and over that professional objectivity and integrity is not going to be afforded me, I declined.

I don't personally think it was you that wasn't being afforded the integrity - more the company by which you were employed which would inevitably have gotten involved, but that's not worth debating here.

Would this mean you're soon going to be in a position to test...??

WRT ‘similarity’ and performance what must always be understood is that there is no BEST tube size, fin density, or any other single parameter.

Agree entirely... and I'm not sticking my neck out to say that we have the best combination available; we are already aware of improvements that could be made to gain further performance from our core... but I do believe we have a BETTER combination when positioned with the popular fans & pumping power in play today than many (if not all) of the competition.

PS: Missed ya bro!

[[O-CuK]Marci]

Roscal - specs have never been published publicly, we had to move the design discussions out of public eyes due to Chinese interference (cloned the PA160 before we'd even released it)... will send you them by e-mail tomorrow.

Done...

[FIZZ3]

Marci']
Define "similar"... in all other comparisons online (inc Bill's data used for Roscal's article) ThermoChill weren't approached to provide a rad to include. The testrig used for Roscal's article that Bill links to no longer exists, so once again, no direct results on the same testrig exist other than Radical53's above...

General consensus seems to be that cos it costs so much more than the others and doesn't comply to "similar" dimensions that it should be omitted / ignored.

I've given up trying... folks'll buy whatever they wanna buy. Leave em to it. Why the hell should we care?

Please don't misunderstand me, an AMD X2 5000+ is 'good enough' for a high performance rig, but an Intel E6700 is still a bit better. I would personally buy the latter (even if it was more expensive), but not everyone likes to spend beyond the pinnacle of price/performance... hence my comment.

I've seen Radical's tests before and I do believe the PA rads are very efficient at lower CFM. I have one on my shopping list, in fact. ;)
I wonder though why BillA won't comment on these tests? Is there some doubt about the integrity of these results?

[[O-CuK]Marci]

Doubts in the accuracy of the testing methodology... which has always been the issue with radiator testing and why so few will take up the job.

Bill's testing is without fault. Radical53's testing methodology doesn't (afaiK) match the acknowledged procedures / control etc used by Bill and a few others.

The issue of testing methodology is a sore one for some - the debate of which resulted in the downfall of ProC... best avoided for now. Let Bill enjoy his return to public posting...

[BillA]

Marci']. . . .
Would this mean you're soon going to be in a position to test...??
. . . .

I'm always testing, I'm so old its all I know how to do.
I doubt I'll have the water loop ever setup again as I've moved 'beyond' WCing; new technologies at hand. And my new personal pc is a silent box, undervolted etc.; the aggregate MTBF of 3 fans will always exceed by far any WCing system.

If for some reason I resume testing with water I'll let you know. But that same issue remains, I'm not taking flack now because the data was presented by Roscal with his analysis.

BR

[rlsnyder99]

Not to highjack this thread, but you guys seem to have a lot of knowledge. How would the PA120.3 compare to a dual heater core, all other things being equal (i.e. pump, size of tube, fan speed, etc.)

[Dragoon42]

Not to highjack this thread, but you guys seem to have a lot of knowledge. How would the PA120.3 compare to a dual heater core, all other things being equal (i.e. pump, size of tube, fan speed, etc.)


it would perform better

[Cathar]

Wow. Getting some of the old boys to come out from the woodwork.

I've seen Radical's tests before and I do believe the PA rads are very efficient at lower CFM. I have one on my shopping list, in fact. ;)
I wonder though why BillA won't comment on these tests? Is there some doubt about the integrity of these results?

Radical_53's results roughly correlate with my own in-development competitive performance testing. If they did not, for sure that I would have piped up at the time that I became aware of his results.

In my mind, it is not the integrity of the results which can be called into question, but perhaps the testing methodology, for which the procedures and equipment used are not up to industry standard. Still, I have no issue with the results as presented on the proviso of accepting that the accuracy of the size of the differences reported are subject to some fair degree of error-based variation. The ordering of the results, and to a degree as per above mentioned error variations, the scale of the differences, all agree with my own testing.

It is somewhat true to assume that radiators of similar construction will perform similarly, if by similar we are referring to similar fin density, thickness, tubing, and so on. To make the statement that the Thermochill radiators perform about the same, but cost more, is to over-simplify that assumption. In my testing I became aware that for fan CFM ratings up to around 100cfm rated fans that the "balance" point of those parameters within the "heater-core" style radiator construction criteria for almost all radiators on the market were significantly less than optimal for close to peak performance levels with fans of <100cfm. The performance differences of the PA120.x radiator series are significant as almost any person who has actually employed one in contrast to a different radiator will attest to.

The significance of the size of the difference is, as always, a matter of personal priority. Different people place different degrees of priorities on such matters as is their wont. It is exactly as Marci says, people will buy that which they want to based upon their own personal priority schedule, so let people make the choice for themselves. Who are we to interfere?

My only personal goal in all of this was wanting to ensure that there was an additional available choice, and at the end of the day, that's what it all boils down to. There's some additional products to choose from. No one is forcing anyone to buy anything, but it is nice to have a choice that is different within a market that is otherwise hell-bent on cloning each other.

[nikhsub1]

Wow. Getting some of the old boys to come out from the woodwork.
Got you off yer bum! :D

[FIZZ3]

Radical_53's results roughly correlate with my own in-development competitive performance testing. If they did not, for sure that I would have piped up at the time that I became aware of his results.

In my mind, it is not the integrity of the results which can be called into question, but perhaps the testing methodology, for which the procedures and equipment used are not up to industry standard. Still, I have no issue with the results as presented on the proviso of accepting that the accuracy of the size of the differences reported are subject to some fair degree of error-based variation. The ordering of the results, and to a degree as per above mentioned error variations, the scale of the differences, all agree with my own testing.

It is somewhat true to assume that radiators of similar construction will perform similarly, if by similar we are referring to similar fin density, thickness, tubing, and so on. To make the statement that the Thermochill radiators perform about the same, but cost more, is to over-simplify that assumption. In my testing I became aware that for fan CFM ratings up to around 100cfm rated fans that the "balance" point of those parameters within the "heater-core" style radiator construction criteria for almost all radiators on the market were significantly less than optimal for close to peak performance levels with fans of <100cfm. The performance differences of the PA120.x radiator series are significant as almost any person who has actually employed one in contrast to a different radiator will attest to.

Thanks, very helpful. I was considering to post about "accuracy versus validity", but did not because I thought someone might take it personally.

In my mind it is very clear that a source of presumably random error is *no big deal*. If that is all there is to those results, then there is no need to worry.
Whatever you measure, there will always be some noise. Noise is only an issue if the effect size you are looking for is compromised by the noise-- something that doesn't seem to be the case presently. For the rest it's a matter of repeatability and validity (more tests get the same result and you are measuring what you want to measure). It's almost like real science, isn't it. ;)