Other than actually knowing what goes where, a problem you're gonna face with any motherboard you may buy, the real test of a motherboard is how (and sometimes if) it fires up when you've done. Fortunately the AV32 was one of the easier boards to work with, firing up first time and slurping the required drivers straight off the supplied CD.

Physically this board is a touch on the large side, so those with limited case room might just want to fish out the tape measure before parting with their hard earned cash. I should also point out here that the 4Gb of memory support is that offered by the chipset, however having only 2 DIMM slots available is going to limit you to 2Gb with each Dimm supporting a 1Gb module.

A quick wander around the bios and users manual, which is one of the better manuals I've encountered, showed a bevvy of "tweakables" designed to help you get every last drop out of your system. These included FSB settings of 66, 75, 78, 85, 90, 100, 105, 110, 115, 118, 124, 130, 133, 136, 140, 145, 150, 160, and 166MHz. before you can get to all these options in the bios though, you need to select a basic range via a jumper on the board itself. This set of frequencies isn't going to satisfy the hardened overclocker who wants to be able to increase FSB fequencies in 1MHz steps, but this board isn't really marketed as an overclocker's board, and the range available will really be more than is needed for your average user. There's also the facility to adjust VCore voltage, increasing it by 0.05, 0.1, 0.2, 0.3 or 0.4V or lowering it by 0.05 or 0.1V.

UPDATE I haven't had time to test this yet, but Shuttle informed me that their latest bios revision allows for FSB setting in 1MHz increments from 100 to 166MHz. I dare say this feature alone has increased the board's desirability factor by quite a margin.

Stability seemed pretty impressive for such a young chipset, though I did run into a few problems when altering memory timings, problems I should stress that could have been caused by my memory rather than the board itself. Other than this I had little to complain about in general use.

Before I installed my beloved Soundblaster Live, I did give the onboard Realtek/Avance Logic ALC100P audio a chance to show its stuff, and to be fair it was a pretty good showing, though not enough to satisfy anyone whos demands for audio are anything more than basic.

Performance :

These days most motherboard benchmarks are pretty pointless, as they often point to minor differences made to look huge on short-range graphs, the kinds of differences even power users are unlikely to notice in real-word use. Where they do have a merit however is in uncovering major flaws in core logic design and architecture that may go undetected otherwise. I'm glad to say that nothing obvious reared its head, and while I ran a rake of benchmarks, I'm only going to show the ones you're likely to be interested in. As for the others I ran, my conclusions are that the VIA Apollo Pro266 chipset still doesn't have as efficient a memory controller as the i815 Intel chipset. There again, I'm reviewing the board not the chipset, so it's not really fair to hold Shuttle responsible for VIA's failings.

For these tests, I used two identical rigs, one built around the Shuttle AV32, the other using an Asus CUSL2 and PC133 RAM. The following components were common to both systems.

Intel Pentium III 800MHz @ 800 MHz
Fujitsu MPG3409 40GB ATA100 HDD (7200 RPM)
300 watt PSU
ATi Radeon 64MB DDR Graphics
SBlive Audio

The differences in most of these tests was so small, I've had to resort to using short-range graphs to emphasize the differences, but even in the 3DMark2001 results above, the difference between the lowest and highest score is little more than 50 points. Using conventional PC133 SDR RAM, the ASUS has a very slight edge over the VIA based AV32, but I'd challenge anybody to see a difference between them running. DDR however takes the honours, though not by very much. It would appear that neither the chipset nor the ageing P3 architecture4 are able to make full use of the extra bandwidth afforded them by DDR.

I'd really expected to see these results a little closer to each other with the CPU being the deciding factor. Again we see the i815 chipset based ASUS board nudge ahead when using conventional memory, with the AV32 needing the extra DDR memory boost to push it out in front.

Again the pattern repeats, giving further weight to the argument about VIA's inability to match the efficiency of Intel's memory management.

And finally Winstone 2001, which in my view gives possible the most accurate reflection of real world performance. We must remember how new the VIA Apollo Pro chipset is, and how much more we may yet see from it as drivers and bios optimisations develop. While this board may reflect badly in these benchmarks, it should be noted that it actually acquitted itself flawlessly throughout all tests, and that we really shouldn't expect it to be any better than the chipset around which it's built. There is a great deal of promise in the Apollo Pro, and the early results suggest we may yet see great things, either from this or from future generations of this technology. To expect something as new as DDR SDRAM and its associated hardware to blow onto the scene and blast past mature and established technology is perhaps a little unrealistic, but there are signs here of what lies ahead. I guess the big question is should you upgrade now?