Synthetic Testing
Windows Media Encoder 9 was used for SimHQ’s content creation test. The application is a free download from Microsoft and greatly benefits from systems capable of accelerating its multithreaded design. Falcon 4: Allied Force’s intro movie, a 91 MB AVI file, was converted into a WMV file with high definition video and audio settings and the total time required by each processor to convert the file recorded. Shorter conversion times represent a better score for this test.
Conversion Time
|
|
FX-57 | 24:02 |
FX-60 | 14:20 |
4800+ | 15:22 |
955 | 18:32 |
840 | 19:09 |
Both the FX-60 and 4800+ performed this file conversation significantly faster than Intel’s dual cores, with the FX-60 coming in a full minute sooner than the older AMD part. The 955 fell behind the FX-60 by over four minutes and behind the 4800+ by over three minutes. The older 840 obviously fared worse, taking over 33% longer to perform the conversion than the FX-60. Finally, the FX-57 finished in an expected last place, though for those who routinely make use of multithreaded content creation applications the fact that the unicore part took almost an additional 10 minutes compared to the FX-60 is noteworthy.
SimHQ also decided to throw a little multitasking testing into the mix to see how these dual core architectures differentiate themselves. Futuremark’s PCMark05 Multithreaded Test 3 runs four separate tests threads (file compression and encryption, a virus scan, and a memory latency test) simultaneously, making for a good multitasking scenario. The recorded score for each processor is the geometric mean of the four tests, with a higher number representing a better result.
Geometric Mean
|
|
FX-57 | 11.4 |
FX-60 | 16.8 |
4800+ | 16.3 |
955 | 18.5 |
840 | 17.1 |
These numbers represent the one test scenario in which Intel’s parts actually outscored the AMD processors, and it is likely that the strong performance shown here is a result of Hyper-Threading’s ability to run four threads on Intel’s dual cores. While the FX-60 scored similarly to the older 840, the 955 pulled ahead of all test parts with the highest score. As expected, the unicore FX-57 was significantly outperformed by the various dual cores in this scenario.
One of the challenges in building a solid performance desktop today is assembling a machine that does not run too hot or too loud for its owner’s personal preferences, a challenge steadily growing somewhat more difficult as the thermal envelopes of parts such as high-end processors and graphics cards continue to expand. To wit, AMD, in addition to their FX-60 review samples, has also made available to SimHQ the Extech 380801 Power Analyzer, a device that displays and records the power usage of the appliance to which it is attached. The power meter was installed between the wall outlet and each test system and the power usage recorded using Extech’s Datalogging software, which writes the measured power consumption to a specified file and at a specified sampling rate. Two scores were generated for each processor, the first representing an idle state of the test machine simply sitting at the Windows desktop for five minutes, and the second a load scenario created by running the full series of tests for PCMark05. Each score represents average wattage of the entire test system for each test scenario. Again bear in mind that both test systems were built using identical hard and optical drives, graphics and sound cards, and power supply units, so the only variables between the two systems were the motherboard, memory, and CPUs. Last, the power management features of both systems were ensured as properly functioning prior to testing for the idle scenario.
Idle | Load | |
FX-57 | 104 | 177 |
FX-60 | 107 | 188 |
4800+ | 102 | 182 |
955 | 155 | 218 |
840 | 147 | 230 |
At first glance it becomes rather obvious that the 955 is still a hot, power hungry part despite being manufactured with Intel’s 65nm process. While the increased cache sizes and higher frequency over the 840 most likely account for the similarities between the two parts, the 955 came dangerously close to throttling itself down during some of the more CPU-intensive testing conducted for this article. In contrast, the FX-60’s thermal envelope is almost identical to those of the other AMD processors, with a slight increase over the 4800+, particularly under heavy use. None of the AMD parts, however, broke over 200 watts under load, and the idle power consumption differences are truly significant variations that show just how hot Intel’s parts run; the consumption of the 955 and 840 at idle are actually much closer to the AMD parts under load than at idle.
Speaking of power draw, SimHQ, somewhat surprised that better thermals weren’t achieved with the 955’s use of a newer fabrication process, decided to install Intel’s system monitoring tool and take a brief look at the new part’s heat signature. It was almost volcanic, to say the least, spiking into 90 degrees C under load, and a sure contender for a more powerful cooling solution than what Intel ships with boxed CPUs.