Cooler testing method

[simon12]

I have recently become a distributor for Spire and am getting samples of watercooling systems and will be running some tests to see how good these product are and was looking for some comments on my planned tests below.
Test setup:
AMD Phenom II 550BE C3 unl;ocked to 4 core and overclocked to 3.55GHz (should get quite hot)
Asrock 880G Extreme 3 board
2 x 2Gb OCZ 1333 DDR3 RAM
Currently 2 x Radeon 4870s in crossfire. Is there any reason to run the tests without them?
Generic ATX case with top and back exhaust fans and front & 2 x side intake fans

Products to be tested (in order of price)
Coolmaster Hyper TX3
Spire Coolgate 09
Spire Verticool III
Coolmaster Hyper 212+
Thermaltake Frio
Spire Thermax Eclipse II
Generic water cooling (will test with 3 different size rads and 2 different pumps)

Test method
1. Install each cooler all using Arctic silver 5 (I thought this would be a better test than using the paste they come with).
2. Run Prime 95 Large FFTs for 20 minutes (temp currently seems to stabilise after 10 so 20 should be plenty)
3. Record CPU temp from speedfan and HWMonitor
4. Install next fan

I will do a test before I start to get some idea but should the results be based on temp above room temp or does the load temp vary differently to room temp? ie if the room was 5C hotter would the CPU temp peek exactly 5C hotter as well. I will try to open/close windows to keep the room the same.

Any comments, suggestions or advise welcome. Also if the water cooling is any good what temp drop from the high end coolers should I expect?

 

[davcon]

You'll have to establish an ambient room temp and then add increased temp values
as test are conducted.
Like test 3 unit 2C above ambient as starting temp.
I wouldn't use AS5 it has too long of a cure time.
You might want to use MX-2/3 or better yet the included paste.
I've noticed that alot of reviews that i've read the readers are now asking for the included paste to be used.
A good idea imo.

 

[simon12]

Thank you for the advise, the main reason I was using arctic silver as I only have enough of this or spires thermal grease to test everything I don't think the cure time is a problem as it will be a fresh installation for all so a fair test. I am considering using the included thermal paste but I thought what if some have 5 minute cure times and others have 200 hour cure times it then makes the test unfair again. Do you think this would be likely?

 

[davcon]

Valid point! so i agree if you use the AS5 it would be a fair analysis.
It would be nice to do comparative test with included paste though.
I know that is more work on your part though.
I'll be interested to read your test results as most members would i'm sure.

 

[simon12]

it looks like Sunday now for the testing and I may use the Spire thermal paste as I have loads on it and only 1 and a half arctic silver so to be on the safe side as I don't want to run out and need to wait for more (or to pay for next day delivery). I will post results on ths thread and open another with full results. Thanks again for your help Davcon.

 

[geofelt]

For a valid comparison, you will want to use the same standard fan at the same rpm. If the cooler comes with a stock fan, then include that also.

Delta cpu temp above ambient is a good measure. It is better if the ambient is kept relatively similar.

If you are testing water cooling blocks, then measure the delta cpu rise in temps over the water supply temp.

If you are testing the entire system, you will mostly have to deal with the supplied radiator and fans.

You might also want to measure the sound from the fans on each cooler type.

 

[simon12]

I am considering trying the same fans on the 120mm ones but its not the best comparison as I expect them to all have fans designed for them and most people will want to know the performance out of the box as buying 2 extra fans adds alot to the price. I think I may try them all with the Spire fans but really I am doing a comparison for them out of the box. As I said initially I will try to keep a constant room temp. I have no way of measuring the sound but I will mention any that sound alot louder. Thanks for the comments.

 

[lothdk]

Great suggestions from the above posters.

Another thing, however I do not know if you have access to do this, but coolers perform differently based on the CPU of choice, so one cooler may perform great on an LGA 775 but might only be mediocre on an LGA 1366.
Then next is overclocking, take a look at some of the current best air cooler performers for LGA 1366 when overclocking, they are more or less equal to lower priced coolers when at stock settings, they only truly show their potential when you get above ~3.7, so you would naturally have to test at stock and several overclock settings to see how far the cooler can go.

 

[simon12]

I understand the results will not translate directly to intel platforms but I only have the system mentioned above (except my old original Phenom x3). The system is currently using the Hyper TX3 and the current overclock is limited by the temps, as setting the voltage over 1.4125 puts the temps close to 60C on a hot day. I think I could get 3.8GHz ish stable just for this test on all fans. What stock settings do you think I should use ie stock dual or stock but still a quad or even as a 3 core? Due to time I will test at 2 different settings maybe 3 what settings do you think I should test?

 

[simon12]

Something came up and missed testing yesterday so maybe today I found overclocking I can get 3.7GHz stable so I will test at that and 3.1GHz 4 core on stock voltage unless anyone thinks I should do it differently.

 

[simon12]

If anyone's still reading its all got wrong and I got thermal paste on the pins on the first test. I cleaned it off and am leaving to dry overnight.

 

[lothdk]

Yes, still reading, and would like to hear back of your results once the testing is concluded.

As to the testing of stock and 3.7, it seems to be a good setting, considering the CPU.

For thermal paste you might want to consider Arctic Cooling MX-2, MX-3 was released a year ago (or so) and is better, but the consistency of MX-2 makes it easier to apply, which makes it better for your testing, price is equal to AS5.
A reason to go with MX-2 is that it is non-capacitive unlike AS5, and with you having to reapply so many times, it might be nice to know that any accidental spill would not cause a potential short.

 

[simon12]

Thank you again. Just got the PC working again I think the problem was the old arctic silver. I don't have any other thermal paste and am now starting the 1st test using the Spire stuff. I hope to post the 1st result in half an hour or less.

 

[simon12]

OK 1st results in but not likely the most interesting:
Both at a room temp of 21C
Idle 3.1GHz Load 3.1GHz Idle 3.7GHz Load 3.7GHz
Spire Coolgate 09 31 48 39 63
Coolmaster Hyper TX3 29 44 36 59

So there no reason the buy a Spire Coolgate 09 over a TX3 (except maybe its slightly shorter) the 2 look very similar (3 heatpipes and a similar size heatsink) but the performance shows another story. Though the Spire fan spins slower it seems noisier. The only interesting point to note was running the same 3.7GHz setup but with cured Arctic silver 5 the TX3 only ran 2C cooler.

 

[simon12]

This morning room temp is 2C lower so just to check I ran the TX3 at 3.7GHz again and load temps are 2C lower exactly but idle temps are 4C lower. I don't really know what to make of this so I will put the heating on to try to get the same room temps for the next test.

 

[simon12]

Next result the Verticool III, I had high hopes for this one as its all copper but all results are exactly 1C higher than the TX3. There are 2 advantages over it though 1 is on an AMD system you can mount it however you want so if your PSU is at the top you can point it out the back so this is a major advantage if your PSU is mounted at the top. Also it includes a fan controller. Hyper 212+ results coming next.

 

[simon12]

The 212+ has a very good way of installing in my opinion just don't rely on the instructions. Its a bit fidley to install but I think it would be almost impossible to do wrong. Also it can be installed either way round like the Verticool III. As for the results at 3.1GHz idle was 31C and more importantly load was 41C beating the TX3 buy only 3C. Crank it up to 3.7GHz and you see its in another league peaking at only 51C a clear 8C better than the TX3 and the only 1 so far I would happily run this overclock all the time with. Idle at 3.7 is 33C. Note to install this cooler and the next 2 I had to remove 1 case fan which I think may account for the higher 3.1 GHz idle increase. I don't think removing this makesmthe test in anyway unfair as I have a standard width case and the fan fits on the smaller 3 but not the larger 3 and from results so far I don't think the smaller 3 are really in competition with the big 3.

 

[simon12]

OK next to test was the 212+ with the fans from the Spire Thermax Eclipse II (I only have 2 identical fans from the thermax or the frio). At 3.1GHz idle was 30C and load 38C (3C better than single stock fan). At 3.7GHz idle still 30C and load 47C (4C cooler than the single stock fan and a huge 12C better than a stock TX3). I have fairly loud case fans and the stock 212+ fan cannot be heard over them but the 2 spire fans are quite loud but not excessively.

 

[simon12]

1st thought on the Thermaltake Frio are its huge, so fart its the worst to install but its still not to bad, like all the others the instructions are not much good though unlike the coolmaster ones they do make sense. The next thing I found is though only a few mm taller than the 212+ it means I had to bend the case slightly to get the side on. Surely they could have lost a bit of the excessive plastic on the cooler to make it fit. Next thing is its very loud with both fans attached alot louder than the 212+ with the spire fans. Other thing to mention is the 212+ can be installed any way round but the Frio only goes front to back not top to bottom. With only a single fan its still what I would consider bordering unacceptably loud it does have speed controllers included but theres no scientific way for me to adjust them equal to anything else so all results are at full speed.
Results
2 stock fans 3.1GHz idle 29 load 37
2 stock fans 3.7GHz idle 30 load 44
1 stock fan 3.1GHz idle 31 load 40
1 stock fan 3.7GHz idle 31 load 49

 

[simon12]

I couldn't change the fans on the Frio without removing it so to speed testing up I moved onto the Thermax Eclipse II and plan to try it with the Frio fans as well as stock for comparison. The next thing I found was the Thermax Eclipse 2 is the most difficult thing to install ever the instructions appear clear at first but I am just not sure if how I did it was what they had in mind or not. The only way I could tighten the bolts was with a pair of long nose pliers so I have no idea how tight it was. It did look as though it would be fine on Intel platforms though. I have no idea if the installation will affect the results but here they are:

Results for Thermax Eclipse II
2 stock fans 3.1GHz idle 28 load 38
2 stock fans 3.7GHz idle 31 load 48
1 stock fan 3.1GHz idle 31 load 40
1 stock fan 3.7GHz idle 31 load 49
2 Frio fans 3.1GHz idle 28 load 37
2 Frio fans 3.7GHz idle 29 load 44

Conclusion so far (my opinion based on the results):
Frio vs Thermax vs Hyper 212+
The Thermax has nightmare installation on AMD, the frio won't fit in a normal case, and the Hyper 212+ needs you to buy uprated fans to get the same performance. As far as the results so far tell they run the same with the same fans and cost around the same (the 212+ is cheaper but if you add the fans for performance it will cost more). So there is no winner if you want a quiet PC and not the ultimate cooling then the 212+ as stock is a no brainer. If you don't care about noise at all and have a wider than normal case then the Frio is for you. If you have an Intel system or don't mind complicated installation and don't mind some noise and want good cooling in a standard case then the Thermax is a good bet.

I will do some graphs and make a new thread with the results soon and consider if I need to do any more tests (ie the 212+ with frio fans or the Frio with spire fans). ny comments welcome.

 

[simon12]

Results now posted here http://www.tomshardware.co.uk/forum/261270-11-coolers-tested-hear-results