Euler3D, Windows Media Encoder 64, Power, and OverclockingEuler3D
This scientific based benchmark simulates fluid flowing around an airfoil. 20 steps were used in this particular benchmark.
This particular benchmark looks to have been compiled to take advantage of the Intel SSE unit, but not that of the AMD processors. As such, the single thread performance of the AMD processors fall well behind that of the E8500. This is also another application which leans heavily on L2 and L3 caches. The Athlon IIs do not stand much of a chance against the Phenom IIs, even when clocked identically.
Windows Media Encoder X64
I took a 250 MB 1080P WMV file and converted it to a 40 MB 480p, 1 mbps VBR file using WME 64.
This is another fine example where more cores will save time. Caches play a small part in the encoding process, but they do help a little bit.
I thought this might be the most telling benchmark here, as the L3 cache in the Phenom II processors are fairly notorious power eaters. Measurements were taken at the wall from the power supply of the computer. The monitor was not included in the power draw. The system was allowed to sit without use for 10 minutes and the idle power was taken. The system was then loaded with Cinebench R10 and max cores.
It is pretty interesting to see the idle figures. When at idle, all cores go into a low power mode and are clocked down to 800 MHz. The two Phenom IIs both had higher overall idle draws than the Athlon IIs, and this again is likely a measure of the L3 caches the Phenom’s have.
When loaded, the quad cores are the hungriest, and again we see what kind of factor that the L3 caches have on power draw at full load.
Since these are revision C3 parts, one would expect overclocking to be better. This turned out not to be the case.
The X2 260 did head upwards of 3700 MHz, but was not exactly stable past that. A problem with this is that since the multipliers on these CPUS are locked, the northbridge portion also gets overclocked. Once around 2.4 GHz things can get hairy with the northbridge (or “uncore”) portion. More voltage needs to be applied to the CPU northbridge to get it stable, and temperatures start to rise. Still, 3.7 GHz is not terrible, and it was an easy overclock to reach.
The X4 640 did not fare so well. It was able to reach 3.4 GHz and stay very stable, but going above that caused instability. I was not overly aggressive with overclocking, but when using the same amount of expertise on the Phenom II X6 1090T and getting it to 4 GHz, this was a trifle underwhelming to me. I attempted 3.6 GHz with 1.45v, but after a few seconds in a benchmark utilizing all four cores I would soon get a blue screen.
These certainly are not enthusiast parts, and if a user wants to overclock then spending the extra money to get a Black Edition Phenom II is their best bet.