Video Tests

    First we will look at XMPEG 5.03 with DivX 5.2.  This program allows us to take DVD video and encode it into other formats.  For our tests we encoded to a DivX file, without audio and calculated the average fps.  Lets see the results.

    Here we see the video card doesn't make much of a difference in this test, at least in dual channel mode.  There is only 1% difference between the two cards, with both cards reaching 1.5 times realtime rendering.  In single channel mode the bandwidth needs of the IGP makes it 4% slower than the FX card based test.  However any of these configurations perform very well.  Overclocking the system gives it exactly a 1:1 increase for the clock speed increase, making this test extremely CPU/bandwidth limited.

    Next is VirtualDub 1.5.10, which unlike XMPEG, does not provide any optimizations for any CPU apart from MMX.  So this is a level playing field for most if not all CPU's.  The test will encode the HuffYUV based file to DivX exactly like the XMPEG test. 

    The results from this test somewhat mirror the results from the XMPEG tests, only slower.  The difference between single and dual channel mode is between 4.5-5% which represents1fps.  The difference between the two configurations in each memory channel are between 1-1.7% with the larger difference being in the single channel mode.  Overclocking the motherboard gives you a 22% increase for the 22% increase in CPU speed.  So far the video encoding tests give the standard ~4% increase moving to dual channel, and 22% increase when overclocking.  Lets see if MPEG-2 encoding, which is HT ready, shows the same improvements.

    Here is one of the two tests where we have the older system included.  Its surprising that the 400MHz slower system is about 2% faster than the newer motherboard, though it is less efficient with CPU time than the Albatron 915 board.  Now looking just at the 915G board we see that there is more than enough bandwidth available for TMPGEnc as there is no major increase gained from going to a non-integrated video card.  However there is a 2% increase in going to dual channel mode, but even that only translates to 0.4fps.  Overclocking the motherboard gives you a 12% increase compared to stock settings, and 10% compared to the 1GHz slower system based on the Northwood core. 

    After looking at the video tests we see that for the most part there is a 4% increase going to dual channel and a 22% increase in overclocking the CPU 22%.  The exception is TMPGEnc which only offers 2% and 12% increases in the same areas.  What happens when we render 3D images, or run all the office programs that you would run?

Other Benchmarks

    3D rendering is something that is very tough on most systems, and Truespace which is one of the more affordable 3D rendering programs is a good test.  The scene we used was one we have used previously.  So lets see the results of this test.

    Again like the TMPGEnc test we have the old Northwood processor as a reference.  Unlike TMPGEnc the older system isn't faster than the newer, higher clocked system.  The Prescott core and new motherboard give you a 11-12% increase compared to the the 16.7% slower Northwood system.  Moving to the PX915 results we see for the first time the IGP beats the FX, though its only in SC mode and then its by less than 1%, thus making the scores the same really.  Moving to the difference between SC and DC we see there is less than 2% difference between the two memory configurations.  Overclocking the CPU provides you with the same increase that DivX encoding gives you 22% for the same increase in clock speed.