Look around the table. If you don't see a sucker, get up, because you're the sucker - Amarillo Slim
A FEW WEEKS AGO, we had a quick Computex look at the Asetek LCLC (Low Cost Liquid Cooling) system - an interesting zero maintenance water cooling job for those who don't like messing around with leaking water tubes, evaporation, re-fills and such.
With its 1/4-inch tubing compared to the 3/8-inch tubing commonly seen in the overclockers rigs, the integrated LCLC was never exactly promoted as a performance cooling platform, though - Asetek has its Vapochill freezer systems for that purpose.
Although the thermal paste is pre-applied on the CPU heads, we replaced it with a brand new high efficiency grease from the Hong Kong company Gelid. The new kid on the block, set up by two senior ex-Arctic Cooling partners, seems to have produced a great thermal paste, GC1. In the BIOS thermal margin tests, the GC1 gave us a few degrees benefit compared to the original one.
The platform of choice was, of course, Skulltrail, at least until we fix the LGA771 mounting issues on the Asus Z7S WS. As you can see, the setup looks fairly simple, the big benefit in my mind being the absence of those old plug-in LGA775 socket heat sink mounting pads: here, Asetek provided the more stable bottom plate plus screwed top holder retention rings, and it worked like charm without bending the mainboard noticeably.
Even without overclocking, and assuming the overstated TDP, the two QX9775 3.2 GHz FSB1600 CPUs on this board should still consume something close to 180 watts total at BIOS boot. And at dual 4 GHz, let's say 260 watts total TDP just at "idle" - can the LCLC handle it? Well, it should - in our Computex review, it handled a CPU plus two hot GPUs simultaneously, with a far greater total load...
This time, I set up a power-guzzling Skulltrail configuration with the new BIOS 1140 - which, at the end, disappointed me a bit due to no measurable overclocking improvements over the old version. I added 4GB of Kingston's overclockable heat-piped FB-DIMMs, still the only ones of that kind around, as well as Asus EN GTX280 TOP - the factory-overclocked 'teraflop peak' version of the Nvidia's current flagship card running at 670 MHz GPU and 2.35 GHz GDDR3 settings. As for the LCLC, I added two Thermaltake smart 120mm fans onto its radiator - at 2,700 rpm they are not the speed champions, but fast enough to thoroughly cool it without making any significant noise.
And the QX9775 CPUs? I stuck with the 4 GHz CPU / FSB 1600 settings, to compare against the Zalman 9500 fans that came with the previous Skulltrail test kit. Rather than rely on BIOS's funny thermal margin readings, I used the Sandra XII environment monitor to check the CPU temperature changes as I ran a variety of tests - 3DMark06, Sandra CPU arithmetic and so on.
Keeping in mind I ran the stuff at a 33℃ ambient temperature here in Singapore. Look at the idle temperature difference: 34℃ for CPU 1 and 40℃ for CPU 2 on the LCLC, vs 46℃ for CPU 1 and 45℃ for CPU 2 with the Zalman. Not bad for a start - the idle temp is barely higher than the ambient one, for a pair of heavily overclocked quad-core processors using a single liquid loop.
When checking the full load operation nearing 100 per cent, the CPU 1 temperature jumped to 69℃ at one point, while the CPU 2 went to 56℃ at the same time. The CPU 1 quickly went down to around 60℃ till the benchmark run was complete, after which both processors were down to the usual idle readings.
With Zalman air cooling, the temperature went to 77℃ / 73℃, and, during the benchmark, it never went below 70℃ until returning to the idle readings after the benchmark completion.
As for the 3DMark06 run, you can see the - pretty decent - result here:
In summary, LCLC can handle a dual 4 GHz Skulltrail setup well: at 1.4 volt CPU settings for both CPUs, it managed to dissipate in excess of 400 watts total when the processor twins are under full load. It'd be interesting to try it out on a Nehalem Gainestown Tylersburg duo next, maybe this time with a bit thicker - say 8mm - tubing system and, why not, more pump strength to provide the extra flow. µ