AMD FX-60 Review - PAGE 12William Henning - Tuesday, January 10th, 2006
Overclocking Tests
The FX-60 turned out to be a solid overclocker. In spite of being the fastest clocked AMD dual core chip, we managed to eke out 2.85Ghz (10x285) with ease using air cooling (in fact, using a stock cooler). Our feeling is that we've not reached the limits of this CPU yet. We were able to get higher overclocks but could not complete out stability tests. The next step? Slap on better cooling to see if we can push this chip even further. We'll report back later on our results, but for now, we have plenty of overclocked FX-60 results to share.
If you want to overclock your system, here are some guidelines we use so you can do so as safely as possible:
- You should reduce the Hypertransport multiplier to 3x or 4x from the default 5x, as you should keep the Hypertransport frequency to 1.1GHz or less
- You will probably have to raise the CPU core voltage to 1.5V or even 1.55V
- You better have excellent cooling too; we had some temperature induced crashes at 2.48GHz with the standard cooler until we added extra cooling.
- Make sure you have good thermal compound applied in a thin even layer, and that the heatsink is making good contact with the processor
For this initial round of testing, we used a stock heatsing and fan, and we were able to obtain totally reliable operation at 2805MHz (11x255MHz) by raising the processor voltage to 1.45V and dropping the Hypertransport multiplier to 4. Memory voltages were set at 2.8V.
Doom 3
Doom 3 once again demonstrates that memory bandwidth and processor speed can give quite good results.
What is really interesting is that the 250x11 and 275x10 tests run the processor at the same speed, but the memory at a higher bandwidth - and purely due to the memory bandwidth, we get a 12.9 fps increase!
We also have a whopping 37.7fps increase between the stock 2.6GHz/DDR400 and the 2.8GHz/DDR570 results.
WinRAR
WinRar also shows a significant influence from memory bandwidth.
Note how the same processor speed 250x11 and 275x10 tests still show an improvement of 31KB/s due just to increased memory bandwidth, and a huge 126KB/s difference between the stock and maximally overclocked results.