Performance and Overclocking Results
Before I even started overclocking, it was pretty obvious that my chip was ice-cold. The Vapochill LS looked like it was not even breaking a sweat at stock clock speeds and held an evaporator temperature of -53’C at idle. Again, please remember that my CPU temperature being reported via software is likely inaccurate. It is, however, still good for relative comparison as I start cranking up the clock speed and vcore.
The System Configuration:
- Asetek Vapochill LS [AC] in Black (115V) with 754/939/940/478 mounting kit.
- DFI Lanparty NF4 Ultra-D mainboard.
- AMD Opteron 148 (Socket 939): CABYE 0543FPMW
- 2x512MB OCZ Platinum Rev.2 (TCCD ICs)
- ATI Radeon X850XT
- 2xWD360 Western Digital Raptors (OS), 1x Seagate 7200.7 (Storage)
- Pioneer IDE DVDR Drive
- Lian-Li Internal USB Card Reader
- PC Power and Cooling 510 Express
- Lian-Li PC65B Case
Before using the Vapochill LS, I had a retail box Thermalright XP90 with a 47cfm fan. I had real temperature problems when using a vcore above 1.5V. This E4 revision Opteron was unusually hot running, and approached 60’C at 1.57V and 2.9GHz. Because of the high temperatures, I was not able to obtain a 24/7 prime stable 2.9GHz without keeping the ambient temperature in the room cool. Out of frustration, I stuck with 2.8GHz when using the XP90.
With the Vapochill LS, I was able to run the same chip at 3.1GHz without even needing a significant vcore boost. 3.3GHz was easy to obtain and 100% stable at 1.6 volts.
Below is a graphical representation of the results. All of these results should be considered ’32M SuperPI stable’. All temperatures are at load using Prime95 in-place large FFT testing.
Looking above, it is pretty clear that lower temperatures allow for higher overclocks. 3.3GHz was totally stable and did not need low ambient temperatures to stay stable like 2.9GHz on the XP90 did. I was able to boot into Windows at 3.4GHz, but even with 1.7 volts I could not maintain any degree of stability, even using memory dividers and all sorts of tricks I have under my belt. It is not unusual to hit hard walls like this on the A64 platform.
Temperature increased proportionally with clock speed and vcore, although more profoundly with vcore.
Click to enlarge
The final overclocking results were very impressive, and I was able to use higher vcore values that were never possible with my air-cooling solution. Even my high-end water-cooling system would buckle under these high vcore values.
When vcore is kept constant, it seems that the colder temperatures just about always permit a 300-400MHz gain over air-cooling. To say the least, I was very impressed.
Articles RSS