Page 1 of 3 123 LastLast
Results 1 to 25 of 75

Thread: CPU water blocks roundup

  1. #1
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505

    CPU water blocks roundup

    Hello!

    I would like to show first results from my CPU water block testing. I included some outdated water blocks but I guess it'll be a nice comparison to their successors.

    Group photo:


    Some info about testing subjects a.k.a. water block ID:

    Little bit different texture on top, really nice!


    Tweaking options! But it takes 3x more time. Worth if we find the best jet for our CPU.
    Supplied with top thermal compound.
    Great mounting.


    Easy to remount, not so easy to install easy mount (short screws).


    Lower end of EK CPU water block, you can get it even for 34,94? (non nickel version).
    No need for dedicated inlet port due to water block design.


    Shuriken? I don't wan't to drop it on my foot.






    Very nice mounting system.


    Bling factor included in box. Should check for backplate issues with your motherboard. On Maximus V Gene there are capacitor legs where the backplate comes.




    Wasn't really happy with 2011 mounting kit. Couldn't install it like that, had to sand down the area around holes.
    Thumb nuts provided with 2011 mounting kit are waay better than the original ones.


    I purchased two separate water blocks so I could test on both platforms (1155 and 2011). Water block base is the same in all versions so you can also buy rebuilding kit for each socket (13,95 ? for AM2/AM3, 13,89 ? for 1155/1156).


    Overall, it would be nice if manufacturers could give us prefered/best mounting orientation for a specific CPU. In results, you'll find how much difference can placement yields. In charts look for Koolance CPU-380I and Koolance CPU-380I@90?. For all other water blocks I have used the best possible mount orientation. In general, inlet port is labelled.
    I'll just stop chitchatting, those were just my biggest personal impressions on tested CPU water blocks.

    Let's hop to used methodology.


    Methodology:
    Water block was installed using original (supplied by the manufacturer) mounting with the use of Arctic Cooling MX-2 thermal compound. Gelid GC-Extreme is a bit expensive for so much remounting. MX-2 spreads very nice and should have a short curing time. According to personal tests with water cooling it's around 1?C warmer than Gelid GC-Extreme.
    Cpu was stressed with Prime95 v27.9, build 1 using smallFFTs test. After running for +30 minutes data was saved, water block remounted with a fresh layer of thermal compound and so on. Test was run five times, the one with the best and worst deltaT average between water and CPU were eliminated. Out of remaining three was calculated the final average deltaT between water-CPU and water-IHS.

    Test system:
    For the first batch of tests I chosed:
    CPU: Intel 2500K, overclocked @4700 MHz, 1,464 V, estimated dissipation ~187 W.
    VCCSA 1,112 V
    Vccio 1,160 V
    Vpll 1,862 V
    Mainboard: Asus Maximus V Gene, bios 1204
    Memory: 2x4Gb Samsung @1600 CL9-9-9-27 1T, Dram 1,601 V

    2500K with a hole for K-type thermocouple:



    Cooling setup:
    480 mm radiator with 4x120mm/38mm SCYTHE DFS123812H-3000 running at 5,089 V

    EK DCP-2.2 pump running at 12,26 V
    Power Consumption: 6.5W +/-10%
    Q-max: 400 L/hr +/-10%
    H-max.: 2.2m +/-10%
    Noise: 17.2 dB

    Norprene tubing 12,7/9,5, whole length ~1,5m

    Fittings: 6x EK-PSC Fitting 10mm


    Data acquisition:
    3x K-type thermocouples, 1x in CPU IHS (drilled hole), 2x for measuring water temperature (about 5 cm above input port of water block), then logged with Voltcraft 304 (sample time 1s)
    Real Temp 3.70 for logging core temperatures (sample time 1s)

    Probe check:



    Results:
    CPU-water average deltaT


    Relative performance index


    CPU-water average deltaT over time


    IHS-water average deltaT


    IHS-water average deltaT over time


    Conclusion:
    I think that charts say more than a thousand words in this section. But, we should be aware, that the big factor in those results is the pump (heavily depends on desing of water block, some will benefit, some won't or the difference be really small). I just wanted to simulate a standard 24/7 system that is kind of similar to mine (1x CPU + 2x full cover, 2x radiator running with the same pump).
    In future, if there's a demand for it, I'll try to repeat this test with a "bigger" pump. Also looking to hook up this system to a flow meter.

    In 1 or 2 days, I'll update this thread with some 3770K results, a week later with another socket, stay tuned.


    Um, no awards?

    Just one, a big thumbs up for all the reviewers.

    Sorry for my english, it's been a while.
    Last edited by Moonman; 06-02-2013 at 03:40 PM. Reason: added WC Heatkiller Rev3.0 LT

  2. #2
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    So, as promised, here are results with 3770K, everything else same as in 2500K test.
    Some changes in rankings, especially Koolance-380I. Was really surprised about it so I did recheck and reassembled everything but results are in the same range.

    Final results from LGA-1155 (Ivy Bridge) testing:

    CPU-water average deltaT


    Relative performance index


    CPU-water average deltaT over time


    Final results from LGA-2011 testing (Sandy Bridge-E):

    CPU-water average deltaT


    Relative performance index


    CPU-water average deltaT over time



    Final results from LGA-1150 testing (Haswell):


    CPU-water average deltaT


    Relative performance index


    CPU-water average deltaT over time


    To sum things up, also combined relative performance index:



    Final results from LGA-2011 testing (Ivy Bridge-E):



    CPU-water average deltaT


    Relative performance index


    CPU-water average deltaT over time


    Combined relative performance index



    Best orientation for water blocks cooling Sandy Bridge-E, Ivy Bridge, Haswell and Ivy Bridge - E CPUs (in alphabetic order):
    Aquacomputer cuplex kryos PRO


    EK Supremacy


    EK Supreme HF


    EK Supreme LTX


    Koolance CPU-370


    Koolance CPU-380I


    Phobya UC-1 LT


    Swiftech APOGEE HD (inlet on top/left side)


    Watercool Heatkiller CPU Rev. 3.0


    XSPC Raystorm


    As you may noticed, no picture for DT 5noz, couldn't try the other orientation because of incompatibility with used motherboard (Z87X-OC Force). Those 3 capacitors are in the way of water block, or vice versa

    Thermal compounds test:

    As promised, here's the comparison between thermal compounds that are supplied with water blocks.

    Aquacomputer cuplex kryos PRO - Prolimatech PK-1
    EK Supremacy - Gelid GC-Extreme
    EK Supreme HF - Arctic MX-4
    Koolance CPU-370 - Koolance GRS-CP002P
    Koolance CPU-380I - Koolance GRS-CP002P
    Phobya UC-1 LT - Phobya HeGrease Extreme
    Swiftech APOGEE? HD - Prolimatech PK-1
    XSPC Raystorm - XSPC Thermal compound K2

    I also had Arctic MX-2, MX-3 and Ceramique in my stock, decided to test them aswell just to see where they stand.

    Testing methodology was the same as when I was comparing water blocks. 5 tests, each time cleaned and reaplied thermal compound.
    The best and the worst result was deleted, remaining three were put into final data processing.
    I chose to test on Haswell so the best possible solution was EK Supremacy with J2 (1,0mm). Their mounting provides the same
    mounting pressure for every mount so that was also a plus for this decision.

    Results:




    Last edited by Moonman; 02-11-2014 at 01:26 PM. Reason: added Ivy Bridge-E results

  3. #3
    Banned
    Join Date
    Oct 2012
    Posts
    566
    Nice effort. Thanks for sharing. One thing to be careful about with graphs is that they can visually exaggerate the results. As an example from your test data, with the exception of the Phobya UC-1 LT, there is only an ~3C delta spread in H2O temp from all of the water blocks tested. Depending on what graph range is used, it will alter the appearance of the minute changes in actual water block performance.

    I realize that some folks mistakenly believe a 1C change is a big deal but in the big scheme of life 3C is a very small range for CPU/GPU thermal cooling. By that I mean that pretty much any of the water blocks tested would work just fine even for extreme overclocking as 1C-3C isn't going to significantly impact system performance, i.e. you are highly unlikely to be able to achieve a meaningful increase in CPU clockspeed via a 1C-3C CPU temp change. That being said people should be able to pick anyone of these water blocks except the Phobya UC-1 LT and be pretty happy with the results.

    As you noted the results are based on the pump that you used, your lines, rads, fans, ambient temp, etc. and the results might and probably would vary with other layouts. That is important for folks to understand and is precisely why with a very narrow differential between these water blocks, they all can be considered a worthy option for any system except perhaps the UC-1 LT.
    Last edited by AMDforME; 04-10-2013 at 08:54 PM.

  4. #4
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    Tnx for input. Added relative performance index chart.

  5. #5
    -100c Club
    Join Date
    Jun 2005
    Location
    Slovenia, Europe
    Posts
    2,284
    Excellent work there buddy

  6. #6

  7. #7
    Xtreme X.I.P.
    Join Date
    Jun 2007
    Location
    Bend, Oregon
    Posts
    5,693
    Really nice work. You should post this up in the Water Cooling forum if you haven't already. Look forward to your 3770 results as well.

    Bravo!

  8. #8
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    Thanks guys, really appreciate your comments.

    Now I have to order a new large tube of MX-2, next on the line is CPU with quite a bit bigger IHS.
    But first need a day or two off, this really is a time hog.

    Pasted a link to this in the water cooling section

  9. #9
    Xtreme X.I.P.
    Join Date
    Jun 2007
    Location
    Bend, Oregon
    Posts
    5,693
    You may want to verify between MX-2 batches that you are getting the same result as before with one of the previously tested blocks. I have seen enough difference between two tubes to cause problems. Always try to make sure I have plenty to finish a whole round anymore.

  10. #10
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    Thanks for heads up. Now I'll test all with the second MX-2 but I'll also check for the differences bewteen both batches.

  11. #11
    Xtreme Enthusiast
    Join Date
    Feb 2009
    Posts
    531
    Wow, the 3770 is a very hot cpu, 10?C difference from the 2500k...
    Quote Originally Posted by NKrader View Post
    im sure bill gates has always wanted OLED Toilet Paper wipe his butt with steve jobs talking about ipad..
    Mini-review: Q6600 vs i5 2500K. Gpu scaling on games.

  12. #12
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    Yes, Ivy Bridge running considerably hotter compared to Sandy Bridge.

    Updated 2nd post with LGA-2011 results.

  13. #13
    Xtreme Enthusiast
    Join Date
    Jul 2011
    Location
    San Diego
    Posts
    845
    Nice work - I love to see correlation so it's good to know that my 380/supremacy results look similar and that your ivy results look similar to st0ned's from what I remember. Your Phobya block seems terrible on every socket though. Mine performed essentially the same as the 380 on a 3930K but yours is 6C worse than the 380 block. It seems high to just be manufacturing variance. Maybe we should trade phobya blocks and see whether things change.

    +1 for posting in the water cooling section, I don't really check out here lol.

  14. #14
    -100c Club
    Join Date
    Jun 2005
    Location
    Slovenia, Europe
    Posts
    2,284
    @stren: It has already been posted in the water cooling section, last week IIRC, maybe it needs a bump? http://www.xtremesystems.org/forums/...blocks-roundup

  15. #15
    Xtreme Member
    Join Date
    Jun 2007
    Location
    Bay Area
    Posts
    164
    great charts! thanks for testing! i just removed my ek supreme hf last weekend and toyed around with the idea of replacing it with something newer, but had no idea what. your post was quite timely and actually tells me i dont really need to swap anything. =)
    1. INTEL E5200 M0 (200x12.5 @1.175), Abit IP-35Pro, 2x2GB DDR2-1000, GT240
    2. INTEL i7-920 C0 (200x19 @1.275), Gigabyte EX58-UD4P, 4x4GB DDR3-1333, GTX680 FTW+ 4GB


    heat

  16. #16
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    @stren
    Yeah, it's really nice to see that results are comparable. It really gives motivation to test more and more.
    As for the Phobya block, could it be related to relatively low end pump? The "sound" difference is really noticeable, just like the pump was really struggling.

    @darkchart
    Nice to hear that it helped with decision.

    EK responded to this small roundup and send Supreme LTX CSQ to add it to the charts


    Update follows.

  17. #17
    Registered User Utroz's Avatar
    Join Date
    Nov 2002
    Location
    Maine
    Posts
    67
    Ivy bridge has a smaller thermal footprint (the area the heat comes from is smaller) and how the waterblock spreads the heat throughout its cooling area effects the results significantly. Another factor is the Sandybridge cpu's have solder for TIM and Ivybridge has paste for the TIM (the interface from the die to the heat-spreader). These reasons are the bigger parts of why Sandybridge and Sandbridge-E cpu's tend to have different results from Ivy bridge cpus. I am highly interested in Ivybridge-E or even Ivybridge-EX (if you hit the megabucks) and Haswell results when they become available. Also I would be interested in seeing a regular Ivybridge cpu with the heatspreader removed or with better thermal paste. Nice work man, keep it up..
    File Server


    Super Old system
    [SIGPIC][/SIGPIC]
    http://valid.x86-secret.com/show_oc.php?id=371866

  18. #18
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    Added EK Supreme LTX results.

  19. #19
    Xtreme Enthusiast
    Join Date
    Aug 2009
    Location
    Louisville, KY
    Posts
    582
    Would be awesome if you could do some lg 1366 tests :O

  20. #20
    -100c Club
    Join Date
    Jun 2005
    Location
    Slovenia, Europe
    Posts
    2,284
    Thanks Moonman! If you wish I can lend you my pressure drop / flow rate testing equipment.

  21. #21
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    @Kenjiwing
    At the moment I don't have 1366 system. Will see if I can do something about it.

    @tiborrr
    That would be awesome. Contact me.

  22. #22

  23. #23
    I am Xtreme
    Join Date
    Dec 2008
    Location
    France
    Posts
    8,826
    I see you tried rotating Koolance 370... What would you say is the best orientation for 3770k?

    With inlet and outlet parallel to RAM, or orthogonal? Thanks.
    Last edited by zalbard; 04-26-2013 at 04:26 PM.
    Donate to XS forums
    Quote Originally Posted by jayhall0315 View Post
    If you are really extreme, you never let informed facts or the scientific method hold you back from your journey to the wrong answer.

  24. #24
    Xtreme Enthusiast
    Join Date
    Dec 2003
    Posts
    505
    With inlet and outlet orthogonal to RAM.

  25. #25
    I am Xtreme
    Join Date
    Dec 2008
    Location
    France
    Posts
    8,826
    Quote Originally Posted by Moonman View Post
    With inlet and outlet orthogonal to RAM.
    Thank you! Nice review, by the way!
    Donate to XS forums
    Quote Originally Posted by jayhall0315 View Post
    If you are really extreme, you never let informed facts or the scientific method hold you back from your journey to the wrong answer.

Page 1 of 3 123 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •