HD Tach Benchmarks

HDTach IDE scores are almost all identical across the board, the biggest variation in the burst scores are under 1MB/s, and there is no variation in the average read scores. CPU utilization is also essentially as the measured processor usage is +/- 2%.
The SATA scores are a little more interesting. The Soltek K890Pro-939, the only non-nForce 4 board, scores about 5 MB/sec less than the nForce 4 boards in the burst speed test, which were all within 2.6 MB/sec of each other. Average read speeds stay identical across the board. The two Gigabyte boards had a higher CPU utilization (4% with our K8N Ultra-SLI) than every other board.
USB2 Testing

The USB 2.0 scores turn the tables once again. This time, the Soltek K890Pro-939 comes out on top, scoring about 4.2 MB/sec more than the four nForce 4 boards, which all hovered around the 32.6 MB/sec mark. The DFI LanParty NF4 Ultra had the highest CPU utilization out of every board (at 10%), with the Soltek K890Pro-939 and Gigabyte K8N Ultra-SLI at 7%. Our other two boards clocked in at 2%. This 5% difference is quite interesting, as both the Gigabyte boards are virtually identical.
LAN Testing


The Gigabyte K8N Ultra-SLI's nForce 4 Gigabit Ethernet controller performs on-par with the K8NXP-SLI's in throughput performance. While the two DFI boards both score considerably higher, ~950 Mbits/sec is still quite fast. Soltek's K890Pro-939 brings up a distant rear. When it comes to CPU utilization, the Gigabyte K8N Ultra-SLI comes out on top, using only 35.34%, as compared to the two DFI boards and Soltek board, which use around 37.5% and 38.16% respectively. Oddly enough, the Gigabyte K8NXP-SLI comes in last, at 42%. Processor usage for the nForce 4 network adapter should go down as we worked with NVIDIA in identifying a bug with their IDE driver that is interfering with Active Armor. We are told that the bug has been fixed but we are still awaiting updated drivers. The Marvell based GbE connection looks like it is PCIe based as it scored a very solid 946 Mbps with roughly the same processor usage