When it comes to testing a thermal compound, we want to try to heat up the processor as much as possible. To give us an indicator as to the effectiveness of each of these TIMs, we'll run a series of tests designed to stress out the processor so that we can try to put these compounds to the grinder. If they fail, so does the processor, but then that's the purpose of testing isn't it? To find out if a TIM can handle the workload?
I'll begin with the test system:
Mainboard: Soltek 75MRN-L (Supplied by Soltek)
Processor: AMD Athlon XP 1700+ (Thoroughbred "B")
HSF: Cooler Master X-Dream SE (Supplied by Cooler Master)
Video Board: Sapphire Radeon 9700 Pro "Ultimate Edition" (Supplied by Sapphire)
Memory: 2x 512MB Crucial PC2700 DDR
Hard Disk: Western Digital 80GB "SE"
Operating System: Windows XP Professional w/ SP1
Testing will start off at default speeds of the processor. This equates to 1470MHz for the Athlon XP 1700+ model. With the VCore setting of 1.65v, we'll be working with a mild mannered 59.9 watts of heat output. After boot up, the system will idle for 15 minutes and a idle temperature result will be recorded. From there, we'll run the SiSoft Sandra CPU burn-in wizard with only the CPU tests applied. CPU usage is set to 100% so we can stress out the processor as much as possible. The highest recorded temperature during the burn-in will be used. After the system has had the chance to cool back down again, I'll run it through a continuous loop of the 3DMark2001 Demo. This should give us both artificial and real-world results for maximum temperatures.
But since we're not talking about anything major as far as thermal output, we'll go ahead and crank up the speed a bit on the processor and test it under more stressful conditions. A moderate overclock speed of 2030 MHz and a VCore setting of 1.8v will bring us up to a more manly 85.2 watts of heat output. Running through the same series of tests should bring a TIM to its knees if it can't make the cut.
Ambient temperature during all tests was controlled at a constant 21C so all tests will be run on equal footing. Both of the silver based compounds were allowed to burn in for three days and the Ceramique was allowed to burn in for the recommended 12 hours.
Results - Idle
We're on pretty even ground here. Whether at default speeds or overclocked, all three compounds are able to provide very respectable temperatures. The half-degree difference between the OCZ compound and the AS3 is probably caused by the slightly smaller amount of silver content by weight. The OCZ is 70% silver by weight and the AS3 is over 75%.
Results - Sandra CPU Burn-in
Considering the quality of the Arctic Silver III and its proven performance, I am very impressed with the results. Whether a mild 60 watts or a much more aggressive 85 watt posture, all compounds are still giving admirable results.
Results - 3DMark2001 Demo Loop
After the last test, the results this time around don't surprise me at all. We are looking at three very capable products here. I was rather impressed with the Ceramique results, though. Considering that it has gone away from the silver based formula, it is still able to produce temperatures that are about a half-degree cooler than the competition.