Benchmarking (continued):
PiFast is a command line style test that will calculate the value of Pi to as many decimal places as the operator requests. In the process, it records the amount of time the calculations take (lower is better), and the results are reported in terms of seconds. The variables set for testing purposes include; calculating Pi to 10,000,000 digits, using the Chudnovsky method, and using an FFT size of 1024 K. From the results below we can see that more memory at lower latencies can shave two to four seconds off of the computation time.
![](./reviews05/xms3500ll/large/pifast.gif)
The next two tests were actually conducted to simulate the performance in two of today's more popular games. The playable demo versions of Quake 4 and Battlefield 2 were both analyzed using a program called Fraps. Fraps can monitor/record the framerate in video games, and this functionality was used to create an average frame rate value for the two games tested (higher is better). Once the games were loaded to the scenes to be tested, Fraps was enabled and allowed to record for 300 seconds as I tried to duplicate the same motions / interactions each time.
Quake was tested on the "Air Defense Bunker" map with the following settings; 1024x768, High Quality, and no Anti Aliasing. The numbers presented in the chart below were automatically generated by Fraps based on the 300 seconds of monitoring. Although the results show that the Radeon X600XT isn't a gaming powerhouse, you can see that the 2048MB kit of memory does provide a slight increase in the frames per second (fps). In reality, based on the somewhat subjective nature of this test, and the fact that the results are so close, the differences seem insignificant to me.
![](./reviews05/xms3500ll/large/quake4.gif)
Battlefield 2 was tested in the same manner, with Fraps calculating an average fps value over a 300 second monitoring period. The game was loaded to the "Gulf of Oman" map with the following settings; 1024x768, High Quality, and no Anti Aliasing. From the results below, we see that the higher capacity, lower latency memory was of no benefit. But as with Quake 4, the somewhat subjective nature of the testing and the minimal range in scores have me thinking the difference in results are insignificant.
![](./reviews05/xms3500ll/large/battlefield2.gif)
In speaking with a representative from Corsair, I am told the Battlefied 2 performance could be greatly enhanced with the same test setup, simply by switching from an ATI to an nVidia based graphics card. Corsair claims that going from 1024MB to 2048MB of DDR with a high end nVidia card will provide a performance increase of 15% or more. I have all ATI cards, so unfortunately I could not verify their claim.
Overall, the testing showed that an upgrade to a 2048MB kit of Corsair's XMS PC3500 Pro low latency dual channel DDR can offer a nice performance increase. The gaming results were somewhat inconclusive, but the graphics card is no doubt to blame as much as anything else. Low latency DDR such as this is generally targetted for people that may not be interested in overclocking, but want optimal performance any way. These modules proved that they definitely fit the bill.
|
|