Restriction isn’t as exciting as thermal data but it is a necessary part of the results.  For restriction testing we use a Dwyer Manometer and a King Instruments flow meter.  As we vary flow we measure the pressure drop across the CPU block in question.  A good rule of thumb I like to quote is that a component should be less than 1 PSI of restriction at 1GPM.


All three Koolance blocks we are looking at today have quite similar results.  It’s noticeable though that the 390 and 390C seem to be about 10% apart the entire time.  I had expected the 390I to have lower restriction because it is a two part construction without o-rings and I assumed that it would therefore leak from input to output.  But instead the 390CI is lower restriction.  This is perhaps due to the different channel design in the inlet port as that is really the only difference between the two designs.

Let’s now take a look at the results vs the other top performing water blocks that we also did thermal testing on:

Even with this few number of blocks it can get confusing as to which is which.  Five blocks appear to be the same while the Alphacool seems higher in restriction, and the XSPC is far lower.  In order to compare to more CPU blocks that we’ve tested we’ll have to simplify things and only look at the 1GPM datapoint:


It’s clear then that the CPU-390 is a medium restriction block.  While it could certainly be higher flow, it’s not so bad as to be a problem.


  1. #1. I use water blocks without jet plate…
    You should tests without jetplate. ¿Makes pump last longer? Do I sacrifice 10°C ?
    For non critical applications I think is better… I’m I wrong?

    #2. Koolance is getting cheaper by the minute….
    CPU-390AMD has removed support for G34 socket “server” like my Tyan S8232,
    Is just 1 centimeter of metal.. 2 small coins. How cheap is that?

    #3. CPU-390intel does not have 2011-v3 Narrow ILM, I want to buy a new server, but I’m stuck with Supermicro SNK-P0050AP4.
    ¿How cheap is not having an optional Narrow ILM?
    Square ILM is 80x80mm
    Narrow ILM 56x94mm (similar to AMD, but not compatible WTF?) Intel finally realized AMD mount is not that bad, if want to make board small.
    But intel decided not to make a compatible mount? ¿Why there is not an Universal Mount?
    Why does a few milimeters hurt so much to a billion dollar company?
    Cheap bastards. LOL. Jajajajajajaja

  2. Koolance should not include thermal paste, if want to save costs, instead of cutting features, nobody use it anyway…
    How good is it anyway? I use Antec formula 7 or 6, I’m happy. Won’t change for another brand.
    not electric conductive…
    Diamonds dust is cheap.
    very thermal conductive.
    Does not dry and crack like OCZ freeze…
    Is not as hard as “the other diamond thermal paste brand.” = Does not need as much pressure.
    Can be reused.

  3. Another thing I don’t like about the 390 are the arrows, the 380 arrows are smarter, less confusing…
    390 is crazy.

  4. I currently have a CPU 380i, with only 1C improvement in temps on this benchmark, I’m not about to dump $80 on the 390i. Very nice review though, very thorough.

    • It’s worth noting that no other reviewer or manufacturer had a CPU-380 as good as mine – other’s (e.g. thermalbench) have shown a bigger increase between the two. So if my 380 was just a lucky golden sample and yours wasn’t then you may see more of a difference. However all I can do is test what I have as best I can 🙂