Performance
To test the performance of the VapoChill, I naturally did what any of you would do. I rounded up the fastest components I had available and tested how much faster they went under the VapoChill.With the absence of water cooling hardware, the VapoChill goes up against a pair of heatsinks. A stock AMD XP3200+ rated heatsink that you'd get with a retail XP3200+ and a Swiftech MCX462 with Panaflo L1A fan are the air cooling culprits.
The test system in more detail:
• AMD Athlon XP3200+ 'Barton' Processor, 2200MHz, 512KB L2 cache, 11 x 200MHz, Socket A
• Asus A7N8X Deluxe 2.0 motherboard, nForce2 Ultra 400 chipset, Socket A, UBER 1004 BIOS
• 2 x 256MB Corsair XMS3200 memory, CL2, 5-2-2, DDR400
• Enermax 431W PSU
• Asetek VapoChill XE with Socket A mounting kit, phase change CPU cooler
• Swiftech MCX462 with Panaflo L1A 80mm fan 5V/12V
• AMD XP3200+ Socket A cooler
Cooler testing isn't too exciting, you just see how fast they make things go and monitor their temperatures.
To that end, here's a single graph that should tell you everything you need to know about phase change cooling in something like a VapoChill XE.
Since that graph references maximum stable overclock, here's that list. CPU Vcore was 1.85V for every test.
| Cooling | Clock (MHz) |
| Stock Cooler | 2400 |
| Swiftech MCX462 | 2420 |
| VapoChill XE | 2730 |
| VapoChill XE Quiet | 2700 |
Quiet mode for the VapoChill means the condenser fan at 2500rpm rather than 3500rpm. I can't really detect the difference with my ear, given the other components in my system.
So what does that graph tell us? Basically that we get 300MHz more CPU overclock versus decent air cooling, and no real difference in overclock with the two tested VapoChill operating modes. That overclock also comes with a near 60 celcius temperature difference. We are definitely running up close to process limits here, checking around the web shows that ~2.7GHz for a Barton core AMD processor is close to maximum, at any cooling.
What the graph and table doesn't show is evaporator temperature. While a table of evap temps isn't too useful, it's worth nothing that you can measure whether the VapoChill is working as hard as possible, by seeing if it matches your 'Hold' temperature that you set in the ChillControl. Set to -30c (the lowest possible), and the CPU to stock clocks (2.2GHz), could the 180W XE setup drag its evaporator temperature down that far?
It gets very close. It's hard to quantatively test things like this, with a single processor and setup. Ambient air temperature, fitment, motherboard etc, all conspire to make cooler testing a less than scientific affair.
Performance Conclusion
While it's hard to come to concrete conclusions about cooling hardware, given the variables involved, I can say this: 180W of performance is enough to run an Athlon XP 'Barton' at 2.7GHz load with only 2C reported diode temperature, in a hot room.
Do you need 200W from a Prometeia in that case? I wouldn't say so. I think XE gives you enough power for current processors. If any CPU comes along that benefits from that extra 20W of heat removal, I'm sure Asetek will have an answer for it. They seem very receptive to providing solutions for their customers, and they were keen to stress that in conversation with me. I'd have to agree, I don't think a Mach II would take me any futher on this highest of high end AMD chips, 180W is plenty.
That runs the OcUK SETI benchmark in under 1.8 hours, solid as a rock, 1 celcius load, evaporator temperature of -20 celcius, and gives over 20,000 3DMark points in 2001SE with an overclocked GeForce FX. It's currently the fastest AMD and NVIDIA combination on the FutureMark ORB, out of nearly 7 million submissions.
