In our last article we looked at the new Intel platforms with respect to their overall configuration as well as with regard to their memory performance. Briefly, some of the results came unexpectedly but most of what we did was concerned with theoretical memory benchmarks. Overall, the streaming performance was comparable with the older DDR(1) platform, whereas the latencies were higher, as expected. The write performance of the new platform was unexpectedly high, however, keep in mind that chipset buffering as well as the streaming algorithms used by AIDA32 may not paint an overall relevant picture.

Today we'll move on from theoretical benchmarks to more practice-relevant applications. That is, we are going through an entire slew of office and content creation applications, from Sysmark2004 to Cinebench2003 and 3ds max amongst others. We’ll round up the picture with a number of gaming applications. One issue we were running into concerning the benchmarking was the fact that we had one nVidia GeForce 6800GT (PCI express) but no equivalent counterpart with the AGP interface. We solved this issue by borrowing a GeForce 6800 from Sudhian and underclocked it to run at the same clock and memory speed as the PCIe GeForce 6800 GT for a 1:1 comparison between platforms. In most cases, the results were within 1 fps margin of those obtained with the ATI RADEON 9800 Pro – we were running at low resolution to test the CPU and memory, therefore this was to be expected. Wherever this held, we are comparing the new Intel ‘925 DDR2 chipset with our benchmark data base, if the deviations from the GeForce to the ATI card were greater than 1% we only show a comparison between the ‘875 AGP-set and the ‘925 PCIe-set.

http://www.lostcircuits.com/cpu/intel-9252/