Update, April 5, 2011:

We have been contacted by Swiftech regarding the thermal performance of our MCW80. They feel the performance it showed is not representative of what they think the MCW80 can do. They have a few ideas why it performed below expectations and we’re going to further investigate the cause of its performance.

It’s possible there’s something wrong with our specific MCW80 or its mounting system. It’s possible there’s an error in our installation procedure or the instructions. It’s possible that there’s some other unknown cause for its deficient performance. It’s also possible that there’s nothing wrong with the procedure, the block, or the mounting system.

When defining this test, we intentionally chose a thermal paste that is extremely sensitive to contact, Spire SilverGrease SP-457. In our opinion, making good contact is extremely important.

We will retest the MCW80 and MCW60 with Arctic Cooling MX-2, a commonly used thermal paste that is less sensitive to contact. Depending on those results, we might also send the MCW80 we bought to Swiftech for them to examine for flaws. If they find our MCW80 is flawed, we’ll also test a new MCW80 with our original test spec.

Comparison Data

Flow / Restriction

Our first chart of comparison data for flow and restriction goes outside the tests performed on the bench itself and follows our pressure drop test methods, where we go well beyond typical loop flow rates and uncover the full spectrum of restriction often times up to 5.0GPM (18.92LPM / 1135LPH) or close to 25PSI whichever is reached first. I stop at 25PSI because I value my life and pressure drop testing can turn disastrous with high PSI values.

As you probably gathered from the individual data tables, the MCW80 takes the crown for being most restrictive followed by the Rasa GPU. On the other side of the coin, we have the MCW60 Rev2 taking the prize for least restrictive followed by the Enzotech VGA-IB. The drastic change in the blocks from Swiftech is quite surprising; the older block is extremely low in restriction, while the newer MCW80 has more restriction than it should. If you running more than one GPU with MCW80′s in your loop, you either have an extremely low flow rate or you are compensating with several pumps in serial. To put it another way, with an MCW80 and a D5, you cannot achieve over 1GPM, even if your loop does not contain any other components… Yes, the MCW80 is that restrictive.

Our next chart just compiles the average flow rates from our test loop into one chart, making it easier to compare the blocks against each other. After all, we all like seeing a showdown no matter what the competition is.

No real surprises here after seeing the pressure drop results. The only thing to point out is to remember than in some cases we had to use rotary 90′s in order to loop up some of the blocks. So, there is some difference to flow results if you are trying to correlate the pressure drop directly to the average flow rate of the blocks.

Temperatures

The next two charts are the ones probably most referenced; TEMPS! However, before letting you move into those, do not overlook the importance of restriction and flow rates. Higher flow rates equate to better loop performance overall, to a point. Nevertheless, this is not the place to get into a pump heat dump discussion, my overall point is higher flower rates mean better cooling performance overall. Okay, now you can proceed to the temperature results eye candy, first up being the stock clocks compiled results.

I do not think any of us should be surprised that one of the best performing CPU blocks modified for GPU mounting was going to come in with the lowest GPU core temperatures. However, the surprise is small margin of victory, which internal lab discussions and guesstimates had it much further ahead. Before discussing too much, let us cover the Overclocked chart.

Result order stays pretty much the same, as it should from stock to overclocked, just more separation amongst the results as the heat output is increased. Unfortunately, the GPU-210 and the Reference cooler were unable to tame the GTX480. Both Koolance blocks are low-pressure blocks, and a GPU with an IHS prefers more pressure, which is my theory on why the Koolance blocks trail behind others. I also expected to see the EK FC480 be closer to the top even with the full cover block capturing heat from the VRM’s, but the majority of universal blocks outperform the full cover. The other big surprise is the older MCW60 Rev2 outperforming the newer MCW80, which was not what I had expected at all.

Now that we have displayed all the individual TIM prints, data tables and the compiled results, it is time to wrap this up and cover final thoughts.

Update, April 5, 2011:

We have been contacted by Swiftech regarding the thermal performance of our MCW80. They feel the performance it showed is not representative of what they think the MCW80 can do. They have a few ideas why it performed below expectations and we’re going to further investigate the cause of its performance.

It’s possible there’s something wrong with our specific MCW80 or its mounting system. It’s possible there’s an error in our installation procedure or the instructions. It’s possible that there’s some other unknown cause for its deficient performance. It’s also possible that there’s nothing wrong with the procedure, the block, or the mounting system.

When defining this test, we intentionally chose a thermal paste that is extremely sensitive to contact, Spire SilverGrease SP-457. In our opinion, making good contact is extremely important.

We will retest the MCW80 and MCW60 with Arctic Cooling MX-2, a commonly used thermal paste that is less sensitive to contact. Depending on those results, we might also send the MCW80 we bought to Swiftech for them to examine for flaws. If they find our MCW80 is flawed, we’ll also test a new MCW80 with our original test spec.