This page has been added a few weeks after the main review.
As we mentioned on the cooling page, we felt that the I.C.E. heat pipe cooling system Shuttle have used could be improved by using a sealed water cooling solution... since the review we have been testing that theory and here are the results.
The plan was to take advantage of the fact that the SZ68R5 motherboard has standard Intel specification heat sink mounting holes and connect a sealed water cooler without the need to modify the Shuttle case. Firstly we brought a standard sealed water cooler from our local supplier which had a 120mm sized radiator. The Shuttle case is too small for a 120mm radiator so we sourced a 92mm one and replaced the tubes and radiator with the 92mm one. This is not for the faint hearted and refilling a sealed water unit is very difficult. We did succeed though and our customised water cooler fitted nicely.
It was only after all this work that we came across a ready-made product that is far more suitable. Asetek, who make OEM products for several manufacturers, have the 545LC which is a 92mm sealed water cooler. Unfortunatly Asetek are only providing this to OEM system builders so it is rather hard to get hold of. Asetek USA were kind enough to send us one for testing.
Details of the 545LC can be found here.
The Asetek 545LC has several advantages over our custom solution. The radiator surround is smaller and it does not block the Shuttle case Wi-Fi antenna mounting holes. The radiator is a little thinner and the fins are further apart allowing a single fan to move air more easily through the radiator and out of the back of the case (We needed two fans on our custom solution, one either side of the radiator to blow air through the radiator effectively). The radiator screw threads are the same as Shuttle use and this means that the I.C.E. thumb screws could be used to attach the radiator to the rear of the case. The Asetek CPU block is a little more cumbersome and it is more difficult to mount the block and keep the tubes untwisted. Both solutions required fitting a backing plate under the motherboard, which means dismantling the Shuttle and removing the motherboard. The backing plates both fitted without any issues.
So, after all this hard work, was the water cooling worth all the expense and effort? The answer is a most definite yes. We saw significant drops in CPU temperature in every test we tried. Please bear in mind that our tests are not very scientific and that there are some inconsistencies, but we think the results show a valid trend.
Prime 95 for 1 Hour. Run with the fan at full speed. Please note that the I.C.E. results (Left) were done before we had the Bios wattage override in place so the 4.8GHz was not maintained (Fluctuated between 4.4 and 4.8GHz) while the water test (Right) was at a constant 4.8GHz. We can see a drop in CPU temperature of about 10 degrees Celsius with the water cooling.
We also tried using a quieter and slower speed Asetek 92mm fan running at 3200rpm and this still kept the temperatures at about 8 degrees lower than the I.C.E. cooler and remember the I.C.E. includes throttling. Also as a side note we found that the CPU temperature reported in Speedfan is about 3c higher than Realtemp and CoreTemp for some reason.
Next we tried looking at some other benchmarks.
Battlefield 3 (This is a screen shot with the custom cooler and twin fans, but the Asetek with a single fan is about the same). As you can see temperatures and fan speeds were very respectable. It is also worth noting that the graphics temperatures on the MSI 570 GTX O/C were brilliant considering the cooling restrictions of the Shuttle case. The fan speed was controlled by speedfan.
Media Encoding... In these tests the fan was set to automatic with speedfan doing the adjustment.
Graphics Benchmarks... Here we look at how the CPU and graphics cope together. Both tests used the CPU for PhysX. The fan was controlled by speedfan.
Water cooling summary
We hope that you will agree that our results show that sealed water cooling shows a significant improvement over the standard I.C.E. cooling system supplied with the Shuttle SZ68R5. In our original review we were able to use the machine at 4.8GHz but recommended that 4.5GHz was sensible for real world use and 4.8GHz for the odd benchmark. With water cooling, however, 4.8GHz is perfectly achievable for day-to-day use. We are confident that the cooling would be quite capable of cooling our processor at 5.0GHz with the Bios being the only thing holding us back.