VIA Chipsets slow down PCI cards

Update:Conclusion

VIAs problem with PCI performance affects motherboards for Intel- and AMD-CPUs alike. VIA only adapts the northbridge to new processors and normally uses the same southbridge for all chipsets of one generation. However, with VIAs new V-Link-Designs, the southbridge also features all of the PCI interface. With our tests, the same PCI problems were found up to the latest southbridge VT8233A. The older chipsets, without V-Link, showed the same effects.

The whole problem is the bad PCI performance of VIAs chipsets. In theory, and according to the datasheets, there is a maximum bandwidth of 133 MBytes/s. In practice, there is 65 to 90 MBytes/s left of it. Intel chipset offer between 110 to 119 MByte/s. Chipsets by SiS and ALi fall in between these two extremes.

Especially for ATA/133 cards this makes VIAs chipsets a "no-go". With a single drive, you may not see the effect as heavily as here. However, these affordable RAID cards are very popular for setting um cheap stripe sets, optimized for performance. With two hard drives delivering 40 MBytes/s, this setup already gets slowed down by VIA chipsets. Depending on the motherboard, there's not 80, but around 60 MBytes/s left.

SCSI-RAIDs with Ultra160 drives are no solution either, as the problem lies within the PCI bus and not the hard drive interface. We found significant performance decreases here as well.

VIAs quietly released Promise RAID Performance Patch only helps user that have a Promise card and a VIA chipset with V-Link technology. This was quickly followed by version 1.04, which boosts performance with all RAID cards tested so far and seems like an OK workaround.

For the time being, we can not recommend unpatched VIA chipsets for professional users who demand high performance from their hard drives and think about setting up RAID configurations. This includes video editing, small- and medium-business servers and workstations for graphics- and audio-editing.

The problems found, especially with the frequent interruptions of data transfers on the PCI bus of VIA's chipsets, may also be responsible for many of the compatibility issues found with their products in the past as the feedback from a lot of readers showed. Most of them were happy after installing the "unofficial" patch by George Breese. We can only hope that VIAs official patch 1.04 will have the same effect. (cvi/nie)