This started out as a fairly standard review of the Asus P3V4X, the new VIA Apollo Pro 133A motherboard. However, I've had the board nearly a month now due to being busy on other things and in the meantime a few interesting things have happened involving this board. Firstly, it's been reviewed on a few other websites, but also there's been quite a lot of discussion about it in the forums here. In fact, I'd say it's the most discussed piece of hardware we've ever had in the forums (with the Promise FastTrack hack a close second :) ). Why? Well, I'd like to say that the forums are filled with glowing praise for it, but realistically it seems to be installation problems, gripes about performance and some niggling hardware incompatabilities. There's even talk of the latest BIOS and drivers actually slowing the board down even further. However, some people have installed without a hitch and are impressed with the board. With all this in mind, I decided to thoroughly experiment with the chipset and the board and see how I fared. I also took a step back and looked again at the Soyo SY-6VBA-133 which is based on the previous iteration of VIA's chipset, the Apollo Pro 133 (note, no trailing 'A') and, on the second page of this article, benchmark the two boards and chipsets against each other.

If you're looking for a conventional "these are the features"-type review of the Asus P3V4X, here's some links to a few written over the last week or so: Anandtech, iXBT, InsaneHardware - no point me repeating what they've already said.

The BIOS
At the time of writing, there are 3 BIOS versions available for the P3V4X: the shipping BIOS, which seems to be an early 1001 version, the final 1001 and the 1002.002 (beta 2) - links to all of these at the end of the article. There are definitely some BIOS peculiarities. A classic example is in the Advanced Menu, where you set the CPU speed. In the shipping BIOS, set the FSB manually to 100MHz, go to another menu screen and come back, and the FSB has reset itself to 80MHz. This is resolved in the later BIOS versions. Version 1002.002, however, has quirks of it's own - it occasionally detects my KingMax PC133 RAM as 2-2-2 when it's actually 3-3-3. Both the 1001 versions detect it correctly. When flashing between the 1001 and 1002 versions, the AFLASH program warns you that "boot block is different". Not entirely sure what they mean by that but it works fine.

There's been some mention of the BIOS versions affecting performance, particularly memory speed. Just as a quick test I ran some 3DMark2000 benchmarks at 100MHz and 133MHz FSB's using the final 1001 bios and the 1002.002 one. The results were nearly identical. This is shown in the graphs in the "VIA Drivers" section of this review.

Anand has a fairly solid theory on why this board is so slow - the CPU IOQ SIZE, a chipset setting which has been shown to affect stability and performance in other VIA boards. This seems reasonable enough but the information is no good to P3V4X owners because none of the current BIOS versions give you the option to change this value.

I originally wanted to use my Soyo SY-6VBA-133 as a way of comparing the Apollo Pro 133 and 133A chipsets. To make it a fair comparison I flashed the Soyo's BIOS to the latest version and was dismayed to see it drop 20 MB/sec on Sandra's memory benchmark compared to the values I got when benchmarking it for my recent Abit VA6 review. I flashed back to the previous revision and had the same results. Stupidly, I had not saved an image of the shipping BIOS of the board before updating it. I tried the oldest BIOS I could find on Soyo's website, and the speed jumped back up - higher, in fact, than I was expecting. Originally the board was getting just above 140, it dipped to 122 with the latest BIOS and with this old bios I am now getting 214! This is with a C400@600MHz and 128MB of Corsair CAS2 PC100 RAM. This would all be fine, except that with the older version of the BIOS, the Soyo seems to be a little confused by Coppermine P3's - it won't let me overclock the P3-500E at all, complaining "CPU has been changed". I can't find any setting in the BIOS to make it ignore this (like Abit's "SPEED ERROR HOLD") and it happily overclocks my P2 and Celeron without the error.

VIA 4-in-1 Drivers
Another big topic in the forums has been the performance disparity between versions 4.17 (on the CD) and 4.20 (on VIA's website) of the VIA "4-in-1" drivers. These drivers are for the entire range of VIA chipsets including the Apollo Pro 133 and 133A. They enable the IDE busmastering, AGP support, ACMI and IRQ routing so are obviously fairly fundamental to the performance of the machine.

So, I did some tests to see how much effect the BIOS and drivers were having. This is the Asus P3V4X with my clock-unlocked P2-400 running at 3x133MHz (400MHz core), Diamond Viper V770U at default clock and 128MB KingMax PC133 RAM @ CAS3. Ignore the speeds themselves, it's the difference that's interesting.

You can clearly see [a] the BIOS makes little or no difference and [b] the later drivers are a killer, speedwise. I had a hunt through the Device Manager in 98 and found some settings to change but none of them had any effect on this speed drop.

This next graph is the Soyo SY-6VBA-133 with a P3-500E (100x5), AOpen PA-3030U (TNT2U) video card at default clock and 128MB Corsair PC100 RAM at CAS2. Now, don't compare this graph to the one above! The systems are totally different because I was testing them side-by-side to save time - it's the difference between driver and BIOS versions on the same board we're interested in:

Well, the interesting thing here is that changing the VIA drivers had basically zero effect on the Soyo, whilst the severe drop from the BIOS update is very noticeable.