VIA Chipsets slow down PCI cards

Stability Issues

According to a lot of feedback from readers, the problems we found with VIAs PCI implementation may also be the root cause for a lot of the stability issues reported for VIA chipsets in the past. We came to this conclusion, as a lot of the effects disappeared as soon as the "unofficial" patch was applied (see next page).

One of our readers was reporting of a lot of crashes with his Elitegroup K7VZA (KT-133A chipset). In combination with Creatives Soundblaster Live Player 5.1 his system crashed whenever some sound was played for a longer time. Whether he was listening to MP3 files via WinAMP or using PowerDVD to watch a DVD movie or trying to play Quake III - his machine would crash after a couple of minutes. After he installed the patch, the system would run fine for hours. Several readers of Kyle Bennetts HardOCP sent him similar reports when Kyle followed up on our story.

We also got mail from a musician who reported that his VIA chipset (KT-266A with MSIs K7T-266 Pro2 R/U) would prevent him from getting any professional work done. He uses Creamwares DSP board Luna II. Cards like this are often used in professional audio production, since their signal to noise ratio and overall sound quality is far superior over off-the-shelf soundcards. Furthermore, they offer lower latencies by circumventing Microsofts DirectX driver model by using special drivers. These drivers interface with software samplers like GigaSampler and virtual studios like Cubase.

We visited his studio to hear the effect in person. When installed fresh out of the box with VIAs latest 4-in-1-drivers the machine would not play one single note from a keyboard correctly over a software sampler. The sound was often interrupted and had a number of clearly audible "hiccups".

A system like this works a little different from the familiar MIDI sounds of a budget soundcard. The external keyboard here only tells which notes to play, and the software sampler creates the sound requested by use of a library and the CPUs processing power. All the soundcard does is a D/A conversion. As a consequence, the system relies on a constant flow of data across the PCI bus while the note is played. As the rest of the setup was working fine, the problem could not have been caused by software or drivers.

After we installed the unofficial latency patch the effect described completely disappeared. Whether we played one single note with a number of instruments or went through complex songs with up to 64 note polyphony and dozens of instrumenst - not a single note was wrong.

The two examples described show that the way VIA implements the PCI bus is far more than a performance issue. The problems also occur with professional applications or while gaming.

The "unofficial"