The overall benchmark results of the Phenom II X4 940 were kind of a mixed bag, with the new CPU posting some very impressive scores, while others didn't reach the level we had expected. The memory bandwidth scores showed some improvement, as did the PCMark suites, but nothing that significant. The gains in the media encoding benchmarks were very impressive, and showed off the advantages of a larger L3 cache coupled with a higher clock speed. A few games, like Unreal Tournament 3 and ET: Quake Wars, displayed framerates competitive with Intel, while the performance in others such as Company of Heroes and Crysis did improve, but still remained well back of the Core 2.
Whenever a company introduces a core shrink of an existing processor architecture, the question of overclocking certainly comes to mind. AMD's 45nm transition is no different, and is actually more enticing due to the pre-release overclocking hype the company bestowed on the Phenom II X4. Intel processors are regarded as having built-in overclocking headroom, and valid or not, AMD's default clock speeds are viewed as running at the ragged edge, with not much room to maneuver. That is a mindset AMD is looking to change.
Our overclock testing included the standard reference configuration, supplemented by a Thermalright Ultra 120 Extreme + 120mm fan, which is one of the best air coolers ever built. At a default clock speed of 3.0 GHz and a standard core voltage of approximately 1.33V, our first step was to see how high we could take the Phenom II X4 940 right out of the box. Without raising core voltages, we easily reached 3.2 GHz, and then 3.4 GHz before finally hitting a wall at 3.5 GHz. We booted at 3.5 GHz with no problem, but after loading Vista Ultimate, it would inevitably crash.
After moving the core voltage to 1.4V, we were able to hit 3.5 and 3.6 GHz, but that was as high as we could take the Phenom II X4 940 and still keep the system stable. 1.45V allowed 3.7 GHz to boot, but we needed a bit more to go any higher and actually load Vista. In order to reach 3.8 GHz, the voltage was jacked to a full 1.5V, and even so, it took some additional tweaking to get it fully stable. We moved the voltage levels a bit higher, and did some experimenting, but that was as far as our particular Phenom II X4 would go.
That was the end of our overclocking experiment, and no matter what we tried, our review Phenom II X4 940 would not go any higher. Both 3.9 and 4.0 GHz would easily boot, but became unstable either loading Vista Ultimate, or when testing for benchmark stability. This didn't seem to be a heat-related issue either, as the motherboard sensors reported values hovering around the 50-52 degrees C range.
Total System Power Consumption Testing
In order to measure power consumption, we took each of the AMD and Intel reference systems, ran the outlet through a Power Analyzer, and then compiled total system consumption figures (in Watts) for Idle and Load scenarios. In order to keep the systems and results consistent, we used the same configuration for the AMD and Intel processors as listed on the Test Systems page.
The first test measures the total system power usage when the AMD and Intel processors are set to power-saving mode. For AMD, this means enabling Cool'n'Quiet, while for Intel we need to activate Enhanced SpeedStep. The test was performed after the operating system has loaded, and all of the various software and hardware components have initialized and we hit a consistent power reading.
This first set of power usage results are amazing, and show the Phenom II X4 940 posting the lowest wattage of any quad core processor in the list. AMD has made some improvements to their Cool'n'Quiet support, and it really shows.
The second test changes the operating conditions from Idle to Load, and the SANDRA Multimedia CPU benchmark pushes processor usage to 100%, while ensuring that no extraneous hard drive or peripheral activity artificially ramps up the power consumption rates. The results using SANDRA are also incredibly consistent, with the wattage numbers remaining stable through the entire test.
Now that the processing load was increased, we expected the Phenom II X4 940 to move up the list, but surprisingly it does not, and holds tightly on to second place, and top among quads.
Please keep in mind that these numbers relate to total system power consumption, of which the CPU is only one part. In evaluations like this, relative placing is sometimes more important than the base numbers, which can change based on the platform and peripheral mix.