Direct Comparison GTX360 vs PA120.3 results

[Okda]

thanks for sharing

and as i thought PA120.3 is only the best deal and deserve paying 120$+ for ti when u have 3 or more of HOT components in ur loop, but for cooling a cpu and a vga only the MCR320 is the best bang for buck imho

[Sideroxylon]

Thanks for the comments guys.

I love some of the conclusions that people come up with. I think there's only a few conclusions we can get from my tests...

1) The GTX and PA rads are both very good rads and perform similarly.

2) The PA does indeed work better with low cfm fans than the GTX does.

Any conclusions drawn beyond that is just speculation. The one of thing I was trying to eliminate with these tests.

[Marci]

Long n' short of it - the lesson to learn when testing large format radiators (Triple or Quadruple Fan Footprint) is that you must be able to produce a seriously large heatload for any differences between the two to become distinguishable... If the heatload is below the capacity of the radiators being tested, then both radiators will appear to perform very much the same, and any advantages of one over the other will be very difficult to detect. The miniscule differences in water temp vs the larger difference in temp at the CPU highlight this - a miniscule change in the temps of variables at the radiator can have a MUCH larger (relatively) affect on things elsewhere in the system - hence when Bill does radiator testing, he uses thermal probes to measure the coolant temps accurate to 6 decimal places. Flowrates measured in GPM to 3 decimal places.

And hence radiator testing is a bit of a minefield for anyone to step into...

[Sideroxylon]

Long n' short of it - the lesson to learn when testing large format radiators (Triple or Quadruple Fan Footprint) is that you must be able to produce a seriously large heatload for any differences between the two to become distinguishable... If the heatload is below the capacity of the radiators being tested, then both radiators will appear to perform very much the same, and any advantages of one over the other will be very difficult to detect. The miniscule differences in water temp vs the larger difference in temp at the CPU highlight this - a miniscule change in the temps of variables at the radiator can have a MUCH larger (relatively) affect on things elsewhere in the system - hence when Bill does radiator testing, he uses thermal probes to measure the coolant temps accurate to 6 decimal places. Flowrates measured in GPM to 3 decimal places.

And hence radiator testing is a bit of a minefield for anyone to step into...

I quickly realized that I didn't have nearly enough heat load when I started the testing on the PA. I almost didn't post them but I figured I'd throw out my results to you guys anyways since I spent the time to gather them.

Thanks for the info and chiming in Marci.

[SDatl404]

What do you consider a seriously large heat load?

So there fore to see a “difference” in the radiators you have to have so much heat in the system that you wouldn’t be using that radiator to cool the system anyways. So what’s the point; just plain being “extreme”?

You see the impression I am getting here is that no person would want to push any of those radiators into that range, because everything in the system would be to hot. If the heat load is lower and if most radiators, of equal size, perform about the same at lower heat loads why buy the most expensive one?

I am a noob at this but I have been reading a lot about it lately and all this radiator hoopla isn’t sitting well with me. I honestly think the only thing that matters is the surface area of the radiator. Hince the reason BIX and the Thermochill work about the same, because they are the same size and thickness.

[Robump]

Thanks,

Thats great help.

Rob

[Sideroxylon]

What do you consider a seriously large heat load?

So there fore to see a “difference” in the radiators you have to have so much heat in the system that you wouldn’t be using that radiator to cool the system anyways. So what’s the point; just plain being “extreme”?

You see the impression I am getting here is that no person would want to push any of those radiators into that range, because everything in the system would be to hot. If the heat load is lower and if most radiators, of equal size, perform about the same at lower heat loads why buy the most expensive one?

I am a noob at this but I have been reading a lot about it lately and all this radiator hoopla isn’t sitting well with me. I honestly think the only thing that matters is the surface area of the radiator. Hince the reason BIX and the Thermochill work about the same, because they are the same size and thickness.

The point that was made was that to get a good comparison between the two rads I needed a much higher heat load to really push them. Not for real world applications but just for testing. Of course you want to give yourself some headroom when picking a rad.

Surface area of a rad is important, but there's other things to take into consideration too. Air and water flow play a large part. The reason my results look so similar is because I couldn't put enough load on them. Don't draw too many conclusions from my testing. I think my results show a definite trend on how the two compare but certainly haven't shown what the two are truly capable of. And just for the record, the rads are not the same size. The PA is thicker and longer, has better flow characteristics, less fins per square inch, etc...Two very different rads.

[SDatl404]

The point that was made was that to get a good comparison between the two rads I needed a much higher heat load to really push them. Not for real world applications but just for testing. Of course you want to give yourself some headroom when picking a rad.

Surface area of a rad is important, but there's other things to take into consideration too. Air and water flow play a large part. The reason my results look so similar is because I couldn't put enough load on them. Don't draw too many conclusions from my testing. I think my results show a definite trend on how the two compare but certainly haven't shown what the two are truly capable of. And just for the record, the rads are not the same size. The PA is thicker and longer, has better flow characteristics, less fins per square inch, etc...Two very different rads.

Well when you get 10,000 watts of heat to put into the system and an 800gpm pump so that you can see the difference, please do retest them for us.

I honestly think that to see the difference you would need to put so much heat into the system that the water would be to hot to use for pc cooling. I could be wrong but I have yet to see real world proof to make me think other wise.

[virtualrain]

Long n' short of it - the lesson to learn when testing large format radiators (Triple or Quadruple Fan Footprint) is that you must be able to produce a seriously large heatload for any differences between the two to become distinguishable... If the heatload is below the capacity of the radiators being tested, then both radiators will appear to perform very much the same, and any advantages of one over the other will be very difficult to detect.

Given the propensity around here to overkill things to the nth degree (i.e. triple rad per block), you are really saying that nearly any triple RAD is as good as the other when cooling just a CPU... so all the endless arguments about which Rad is better is hardly relevant with the low heat loads they are subjected to... they will all perform about the same.

[MrToad]

Given the propensity around here to overkill things to the nth degree (i.e. triple rad per block), you are really saying that nearly any triple RAD is as good as the other when cooling just a CPU... so all the endless arguments about which Rad is better is hardly relevant with the low heat loads they are subjected to... they will all perform about the same.

Hmmm... I suppose is subjected to interpretation.

The way I understand it, you can put more blocks in the loop with the TC than with the GTX before reaching the rads maximum cooling capacity.

In fact, if I understood correctly the data on the tests, with a single CPU block the comparison should be PA120.2 vs GTX320.

[virtualrain]

I guess so, but if someone insists on cooling a single block with a triple rad, it doesn't appear to matter what rad you use.

[Senater_Cache]

cut

[SDatl404]

I am curisous to see how much heat the triples can take before overall temps start to rise. Meaning that if the guy who did this test added more stuff to the loop would his CPU temp rise or not and if it didn't rise how much heat can the radiator take until the CPU temp starts to rise.

If you added heat to his loop i feel temps all around would go up. Maybe this is where the thermochill would shine, but then it would mean a thermochill is unnecessary for people doing multiple loops.

I am curious to see how a BIX III would perform in this test.

[migueld]

What do you consider a seriously large heat load?

So there fore to see a “difference” in the radiators you have to have so much heat in the system that you wouldn’t be using that radiator to cool the system anyways. So what’s the point; just plain being “extreme”?

You do bring up a very interesting point because regardless of how the radiator performs in extreme situations, it seems to me that the extreme theoretical situations are inapplicable to most people. I often hear people recommending PA to every single user, even noobies, and I wonder, will they really notice a difference? What Sideroxylon’s results show is that a GTX performs very similar to a PA in a real world set up, a set up that most people can relate to. The extreme theoretical tests can show that one product is better, but how does that translate to a real world set up?

On the other hand, Sideroxylon, one test I’d really like to see is the CPU at 100% and the fans at 35%. That may be a test where the PA might really shine. If not, then… I guess it’s time to stop the PA mania!@

[Grinch]

You do bring up a very interesting point because regardless of how the radiator performs in extreme situations, it seems to me that the extreme theoretical situations are inapplicable to most people. I often hear people recommending PA to every single user, even noobies, and I wonder, will they really notice a difference? What Sideroxylon’s results show is that a GTX performs very similar to a PA in a real world set up, a set up that most people can relate to. The extreme theoretical tests can show that one product is better, but how does that translate to a real world set up?

On the other hand, Sideroxylon’s one test I’d really like to see is the CPU at 100% and the fans at 35%. That may be a test where the PA might really shine. If not, then… I guess it’s time to stop the PA mania!@

PA's are the KING of radiators...period!

[dinos22]

Given the propensity around here to overkill things to the nth degree (i.e. triple rad per block), you are really saying that nearly any triple RAD is as good as the other when cooling just a CPU... so all the endless arguments about which Rad is better is hardly relevant with the low heat loads they are subjected to... they will all perform about the same.

that's what i gather from his statement too

i'm happy with my GTX240 anyways......CPU in the loop it can't produce that much heat.....maybe with 8800GTS things will get funkier

[Eastcoasthandle]

Wow, this reminds me of the GTX480 thread...

[Jedda]

I think the rad threads would improve considerably if contributors considered rads on their merits. Exchange rates and shipping fees aren't absolutes. Yet many factor their local circumstances into performance discussions as if they were.
I can buy PA120.3 for much less than 360GTX. That doesn't affect their performance in any way or justify a sour grapes attitude towards the 360GTX. Yet we continue to see exactly that re the Thermochill rad.

[Sideroxylon]

On the other hand, Sideroxylon’s one test I’d really like to see is the CPU at 100% and the fans at 35%. That may be a test where the PA might really shine. If not, then… I guess it’s time to stop the PA mania!@

It's bothering me that I didn't do that test. I didn't even think about it till I had already switched rads. I'll tear back into it this weekend and do that. Any other tests people would like to see while I'm at it?

Stay tuned.

[Cronos]

It would be nice if anybody doing such a tests does at least little analysis of the data, including consistency and some simple math relations which constituent watercooling basics.

The temperature deltas, air flow, water flow, and total heat load relates each other with very simple equations, and knowing them can sometimes easily tell
if there is something wrong with the data.

[Jedda]

The temperature deltas, air flow, water flow, and total heat load relates each other with very simple equations, and knowing them can sometimes easily tell
if there is something wrong with the data.

Going to provide them, since it's a given Sideroxylon would've if he could've?

[Sideroxylon]

It would be nice if anybody doing such a tests does at least little analysis of the data, including consistency and some simple math relations which constituent watercooling basics.

The temperature deltas, air flow, water flow, and total heat load relates each other with very simple equations, and knowing them can sometimes easily tell
if there is something wrong with the data.

I know your not completely happy with the results I've posted. I've made them as accurate as I can. Short of buying some very expensive test equipment and spending way more time than I'd like to, that's as good as they're gonna get. I'm not going to be checking flows and heat loads and whatnot. I'll provide the basic results as I did before and let people like you, who seem to have a grasp of the math involved, figure out whatever results your looking for. I know it sounds lazy but truthfully, I don't really want to know all the math involved. I just wanted to show you guys my results with these two rads and maybe help answer some questions that have been floating around here since the two rads came out.

[Marci]

You do bring up a very interesting point because regardless of how the radiator performs in extreme situations, it seems to me that the extreme theoretical situations are inapplicable to most people. I often hear people recommending PA to every single user, even noobies, and I wonder, will they really notice a difference?

Half the problem is the way in which folks blindly make their purchasing choices.

As I've always said whilst data has been available, quantify the heatload of the items being cooled, then decide at what noise level your aiming to cool it at, THEN choose the radiator needed.

Folks overcomplicate this.

1) Quantify the heatload: Go here - http://www.extreme.outervision.com/p...ulatorlite.jsp - stick in only the items that you intend to stick waterblocks on. If there is the option to stick in your level of desired overclock, do so. Leave all other fields blank, other than the pump if it's listed. Your total is your approximate heatload. Add a touch more for an overhead.

2) Decide whether you want a quiet / silent system, or whether noise is indeed an issue at all... now find the fans that match your desire.

3) Consult radiator testing data to find which radiators combined with those fans has a Heat Dissipation equal to or greater than 1) at between 5 & 6lpm (common flowrate for most loops using a D5 or other common pump).

If 1) comes to 200w, and you're using a PA120.3 with Nexus fans at 12v, the radiator can happily dissipate in excess of 300w of heat, so we know you can reduce the fanspeed down to 7v happily. Any other radiator and fan combination capable of cooling in excess of this heatload will produce very similar results, but as you drop the fan speed, the Heat Dissipation of other rads falls faster... eg: a BI GTS360 with Nexus fans at 7v will have a Heat Dissipation of LESS than 200w, so the air>coolant differential rises, thus producing a higher temp. To compare two triple rads with a 150w CPU as the heatload, we know that you need to start reducing the fan speed to find out which radiator will allow you to do so at the lowest airflow... as that would be the only way to compare those two rads with a decent visible difference. You can't measure airflow with what you have available... so, the easiest way to do this is to use the same fans on all rads. Put the CPU at the same clockspeed and voltage for all tests, and give it 100% load. Now, reduce the fanspeed whilst measuring the fan's voltage with a multimeter. At somepoint, one rad will start to produce higher temps than the other rad. It may be the case that you can continue to lower tha fan voltage by another volt or more with one rad than you can with another before temps start to rise drastically.

The "best" watersystem is in effect the most efficient. The most efficient can be defined as the one that moves the most heatload with the lowest "effort", with the fan airflow being the "effort", or the liquid flowrate being the "effort". If using fan airflow as the basis, then it becomes a test of noise. Which can cool the heatload at the lowest noise. This is the case for the majority of users. Most people are using D5 or DDC-based pumps. The variance is in the fans. Folks blindly go out and get a PA120.3 and Yateloon because it's the trend at the moment, without seeing exactly what airflow they NEED. They may need less airflow as their heatload is low. They may actually need MORE airflow as their heatload is high.

Say a 120.1 rad needs 50cfm of airflow to cool an 80w CPU. This 50cfm is achieved by running fan X at 9v. If you move to a 120.3 rad, you can reduce Fan X to 5v, as 3 of Fan X at 5v produces the same total combined airflow as a single of the same fan at 9v... so the larger rad allows u to cool the same heatload with less noise.

So the average choice of "which radiator" becomes a matter of "what's your heatload" vs "how quietly do you want to be able to cool your heatload". The ThermoChill series, when comparing SAME FOOTPRINT against another brand (120.1 vs 120x1, 120.2 vs 120x2, 120.3 vs 120x3) allows one to do so with less airflow than the rest.

Now, if the PA120.3 can cool a relatively low heatload with LESS airflow (ie: less voltage to the fans) than the GTX480 which has an extra fan at the same voltage, the PA120.3 is obviously the winner at low airflow / low heatload cooling...

It isn't all about "extreme" heatloads... it's just extreme heatloads can produce very noticeable differences at a fixed airflow when looking at output temps. When testing with non-extreme heatloads, the differences are more noticeable by looking at the airflow required, rather than solely the output temp.

The above is all awfully explained... it's hard to word it... hopefully you can dig yerself some clarity out of it all and grasp what I'm trying to get across *shrug*.

[dinos22]

It's bothering me that I didn't do that test. I didn't even think about it till I had already switched rads. I'll tear back into it this weekend and do that. Any other tests people would like to see while I'm at it?

Stay tuned.

FANS OFF COMPLETELY UNDER LOAD

[Sideroxylon]

FANS OFF COMPLETELY UNDER LOAD

That's not a bad idea. I'll check it out.