Short comparison: Swiftech Storm rev.2, Aqua Extreme MP-05 LE, Nexxxos XP

[Radical_53]

I'm just about to finish a small test, seeing which of these 3 blocks would do the best job on my Opteron 170. As a testbed, I used my Venus board and collected the data from CoreTemp plus the onboard sensor. Water temperature was measured by 2 external sensors (one after the block, one before just to be sure).
The loop I used consisted of 7/16" tubing, EK barbs, Laing DDC+ with modded Watercool top (just like the Alphacool top), Thermochill PA .2, a filter and a little res I used instead of a t-line (it has the same shape as a t, just bigger). Also there was a flow sensor to check restriction, thermal paste used was standard Silmore stuff.

Restriction:

Storm Rev.2: 248-249 l/h ~ 4,15 l/min ~ 1,1 gpm
MP-05: 239-240 l/h ~ 4 l/min ~ 1,05 gpm
Nexxxos XP: 224-225 l/h ~ 3,73 l/min ~ 0,99 gpm

Then I checked temps with IHS, 2.9GHz at 7.2% added voltage (~1.41-1.42V):

Storm Rev.2: 38°C at 24.7°C water temperature, core 1 53°, core 2 47°C
MP-05: 39°C at 24.3°C water temperature, core 1 57°, core 2 47°
Nexxxos XP: 40°C at 24.3°C water temperature, core 1 57°, core 2 50°

After that, I took off the IHS and tested all 3 again. It was easy to set the Nexxxos right, as it only has one bolt holding it down to the socket, but much harder to get some good results with the MP-05 and the Storm. So, if you plan to use one of these on a CPU without IHS, I'd recommend the following:
Tighten down the block just a little so that it sits on the CPU without moving. Then check your temps and tighten it down accordingly. I had huge differences when doing this, both on overall temps and also on temp difference between the readings of core 1 and 2. If it's done right, you can decrease the difference between the cores plus you'll get more overall performance in the end. I never saw differences this big just from the mount, yet it does make sense as the core is so big.

Storm Rev.2: 34°C at 24.7°C water temperature, core 1 35-36°, core 2 37-38°
MP-05: 33° at 22.5°C water temperature, core 1 31-32°, core 2 37-38°
Nexxxos XP: 32°C at 24.7°C water temperature, core 1 33-34°, core 2 39-40°


As you can see, the Storm was best regarding temps on the cpu with IHS, restriction and temp difference between the cores, both with IHS and without. The Nexxxos XP does a great job when the IHS is off, yet it has a large difference between core 1 and 2 (and you can't do much about that, as it does only have one bolt holding it down). The MP-05 was quite close to the Storm on both setups, just a tad behind.
Both semi-impingment blocks showed the big difference between core 1 and 2, even though it got smaller when the IHS was removed. Their peak performance was good, still they didn't manage to spread this on both cores. What I know from this specific cpu is that it's quite picky considering temps, also it was always the hotter core that gave me a hard time when trying to get to higher overclocks. Personally I'll stick with the Storm, as it has the lowest restriction and best overall performance.

Testbed looked somewhat like this:







IHS after removal:








Now I know the testbed isnt's too exact, still I thought it might be interesting for some to see these numbers on a real world comparison.

[v0dka]

Good job.

I also have the temp difference between core 1 and core 2, but I was under the impression that this might also just be a bug in coretemp.

[Radical_53]

I thought so too until I changed the blocks and saw that there is a way to minimize the difference.
Another thing that I could see when doing this test: We all know how the temps are jumping around on a DFI board, right? That gets smaller and smaller as you raise flow on the loop (I had the option to turn down the DDC+, so at lower rpms the temps started to jump again).
As they were still jumping without IHS, there's still something left for me to improve

[Okda]

Thanks for this nice review

I love my Storm and Thank God i bought it

[Patriote]

Nice Little Review! I will definitely Get A Strom by the same time a second MCW60 along whit of course the X1900XT Master card of course

Oh and I never eard of Removing the IHS, Looks like it give better temps. SO basicly it's just removing the "Heatspreader" On the CPU ? Right ? How do you do this ?

Thanks!

[Radical_53]

Never heard of it, really? But yes, you "simply" remove the heatspreader. Can be done with some precaution and a razer blade or a very thin scalpel. I'll take a picture of my "tool" later on if it's interesting to you.

[Jupiler]

Thx for sharing the results, pretty interesting.
Think I'll be getting me a Storm (once they become available again of course).

[Radical_53]

It shouldn't be much difference to your MP-05, yet it's a "sure bet" I'd say. I liked the way it performed, very nice results on every run.
Also I do like the low restriction compared to the other two. Right now I'm using this setup to test some rads and added my two modded GPU-X 2.5 blocks. Flow went down from ~250l/h to 210l/h, so that's 0.92gpm for my full loop. Not much less than a CPU only loop with a Nexxxos

[Patriote]

Never heard of it, really? But yes, you "simply" remove the heatspreader. Can be done with some precaution and a razer blade or a very thin scalpel. I'll take a picture of my "tool" later on if it's interesting to you.

Yes, Im Interested in it

Also, Thanks for the Tips. I may try this some day. Fews Questions thought, If one day i decide to remove the Heatspreader from the CPU, Is there anyways to get it back on the CPU ? It's looks like it's Sticked whit some glue or something... Also Is there some Thermal Paste on the Core/s under the Heatspreader ? I bet no.. so i guess you just put some ?

[Radical_53]

The grey stuff you can see on the spreader is the standard AMD paste (looks like some Shin Etsu compound). And yes, you can get it back on. I did that several times using glue my dad has for his model cars and trucks. Normally the black silicone stuff keeps the spreader on the chip, but you can glue it back on without too much of a problem.

[Patriote]

Alright Thanks! Nice Oc BTW! Opty 170 3Ghz Nice! I should have got a Damn Opty 165 and Oc it to what i curently have...2.76Ghz and Paid half less...lol In that time i was'nt into Overcloking.... I can't seem to pass the Magical 2.8Ghz on that 4800+ Currently i need to give the CPU around 1.55v to get Stable a 2.76Ghz... Just finished Prime95 Testing for 24 hours and it's Stable...Ill try 2.8 later this Morning. 234x12 should work if not ill try using some other CPU Multi.

Anyways, Sorry to Hijack Your Thread lol

[Jupiler]

Yep,
I always thought the nexxxos would be the less restrictive block of the 3, but looks like it isn't.

I'd like to see how the 3 blocks perform with the D5 pump.
Any chance you'll be doing a test?

[thunderstruck!]

G4 wins again!

[Radical_53]

@jupiler: To be honest, the Nexxxos was less restrictive on my last test But that was with like half the flow or so. I think with this much performance from the pump, the MP-05's straight through inlet takes the lead compared to the Nexxxos's cornering thing.
Testing with the D5, well Would be possible. I'd like to do the rad test first, but after that I can hook up the D5 and the DDC+ with the Radiical top to get some comparison on that.

@patriote: I'll have to see how much more would be possible now without the IHS. Opting for 3.1-3.2GHz

@thunderstruck: Yes, but it was close. All 3 are great blocks I'd say. Still to me, the Storm is the "best package" overall.

[OCme]

Great job Radical 53! I always enjoy your reviews.
Simple, easy to understand, no-nonsense information we can all benefit from

[creidiki]

Its not ShinEtsu... SE doesent dry and crack that badly. The company that makes it apparently calls "Themal Crap" internally...

Regardless, interesting results if I could trust CT... but after being told my core 0 idles @ 17c on a 23c ambient, I dont trust its results any further than I can thrown them with a limp wrist ;p

[Radical_53]

Well, I never trust idle temps anyhow Nothing as strange as that. Still, what else do we have? It gave me some information, may it be good or not, I was able to get some answers with that.
If those 35°C really are 35°, or rather 40 or 45, well? I can't tell. Yet they were moving according to water temps, not fluctuating too much and only with CT I was able to see some differences honestly.
After going through the data, comparing it with the structures of the blocks, to me it does indeed make sense how the results of the blocks came together.

[v0dka]

I'm tempted to try and reaset my Nexxxos - again - as it did have even temperatures at one point. However no remount I did after it got uneven solved the difference and like you said Radical the design of the block and the straight single screw down method would make it hard to set it up that crooked.

It's just weird, at first I thought it was the IHS, so I took it off and voila, even temps on both cores, after taking my block off once I couldnt get them in line anymore...

Idle both the cores report a more even temperature though.

[Radical_53]

I thought it was the IHS too before trying different blocks. The IHS did its part for that difference, yet the biggest improvement came from the matching block being mounted "just right".

[Marci]

Whilst IHS is off, groove it... install a thermocouple. Take independant readings. http://www.xtremesystems.org/forums/...ad.php?t=89468

Acceptable method is as per Intel's drawings (available in Intel's TTV Specifications and Testing Procedures - google - they're publicly accessible at intel.com), on an Intel CPU. For AMD, just ensure core centrality, but use same process.

To mimick Intel's "permanent" indium-soldered bond to core, use Coollabs Liquid Metal between IHS and core once grooving is done. Use the same to fit thermocouple into groove.

[Radical_53]

Great idea for a testbed. Yet this was "my" cpu, something I'd like to keep. I was just curious which of the best blocks around would come in first on this specific thing.
That's why I called it "short comparison" instead of "little review". There's no single thing to get more enemies with than a wrong cpu waterblock test result...

Btw., "Goliath"'s little brother is doing a great job right now. The comparison we talked about should be done within the week.

[creidiki]

Teases go to hell without stopping by the start cell to collect 200$, you know that right? ;p

[Marci]

Yet this was "my" cpu, something I'd like to keep

S'only the IHS you'd be messing with, not the CPU itself - can send you one to groove if u like so u can keep the original... skip the coollabs for IHS > Core. Just use AS5...

Will still give more accurate (read: useful) results than coretemp... and minimise the risk of results being challenged on basis of software / calibration issues.

[Radical_53]

@creidiki: Well, that's just the "revenge" for you letting us wait so long for the 140mm shroud performance thing, remember that?

But, just for you, some more info As you know, the PA .3 came out way in front of the competition, with a bigger lead to #2 than #2 had to the last of the bunch. Now there were people complaining, as the PA is much bigger and heavier than most of the others. Well... so I talked to Marci and now I'm trying to find out how a PA .2 would do against the competitions .3 rads. Not all of them, just a selection but still some of the best from that test (like Swiftech MCR e.g.). Numbers do look satisfying at this point

@marci: Yes, you're absolutely right about that. If you have a spare IHS with a groove, that would be absolutely great tbh

[el rolio]

great job radical as usual. good thing that method of mounting with a decapped cpu is the way i like to do it too. load up bios and start turning.

im happy cuz my storm was able to keep my hot 146 under control (with 2 68gts no less) with a avg 31º water temp at the rad. yay witer is coming and the ONLY thing i like about winter is that it drops my water temp closer to what you guys see in your normal usage, the 24º area. ha.


screwed long posts permanently into the backplate.