ASUS P3V4X / VIA AGP GART Performance Explored

The Situation

When we first looked as the ASUS P3V4X in our February 2000 VIA Apollo Pro 133/133A Motherboard Roundup, we loved everything about the board except for the bottom of the pack performance under Quake 3. Features were plentiful, layout was mostly well thought out, and stability was top notch. But those Quake 3 performance numbers killed any chance the board had of attaining Editor's Choice status.

At the time, we were hoping for a BIOS upgrade from ASUS to restore "normal" Apollo Pro 133A performance. While various BIOS updates were released and rumors circulated that the issue had been fixed with the new "fast-writes" feature of BIOS 1003, we were never able to duplicate these findings in the AnandTech lab.

However, in our in depth look at the P3V4X, we found that reverting back to an older version of the VIA AGP GART driver brought back the performance we expected from such a board. Specifically, VIA AGP GART version 3.59 (included in the VIA 4-in-1 Service Pack 4.17) put the ASUS P3V4X right behind the class leading Gigabyte GA-6VX-4X. We had originally tested with the then-current 4.00 AGP GART from VIA, which is included in their 4-in-1 Service Pack 4.20.

Although the decision was very controversial, we held back on giving an all out recommendation to buy the P3V4X until the AGP GART issue was fixed in a more satisfactory manner. The premise behind this is that the prospect of being stuck with an old version of the VIA GART driver just doesn't sit too well with us as you never know when VIA will introduce an important feature or fix. While many of you decided to go ahead and take the risk of being stuck with an old AGP GART driver, others continued to wait for a fix to the problem before buying a P3V4X.

After many long discussions with VIA and ASUS, who of course blamed each other for the problem, we still had no resolution over 2 months after the release of the board. From what we can gather, it appears that ASUS has been known to "fix" problems with VIA's chipsets in the interests of compatibility. In the case of the P3V4X, it may have enhanced compatibility or stability under certain conditions, but it certainly caused problems with the 4.00 AGP GART driver. We finally got word from ASUS that the fix would need to come from VIA, although we still held out hope for a BIOS upgrade from ASUS.

Finally on May 12, 2000, VIA released version 4.02 of their AGP GART driver, available directly from VIA's driver page. Would this be the fix ASUS had alluded to? We took the first opportunity we had to fire up the P3V4X and see if this new driver was the one ASUS promised was on the way. While we were at it, we decided to look at Quake 3 performance under Windows 2000 Professional to see if the P3V4X might be crippled in that environment where there is only one VIA AGP GART driver currently available.