VIA KT266A Initial Roundup - October 2001

Shuttle AK31 Revision 3.1

Click to Enlarge

It has been a very long time since we've looked at a Shuttle motherboard, and not too much has changed. Their PCBs still have the same rounded corners that were there when we first started reviewing motherboards over four years ago. You won't find any red or blue dyes, no diagnostic LEDs, and the AK31 still uses the now outdated large BIOS chips instead of the smaller socketed chipswe are used to seeing.

The AK31 Revision 3.1 is really a no-frills KT266A solution. It is outfitted with 6 PCI slots and an ACR slot but by far the most interesting thing about the AK31 R3.1 is the fact that it has four DIMM slots. Very few people understand exactly how difficult it is to layout a Socket-A DDR motherboard with more than two DIMM slots. The demand is great enough for more than two DIMM slots that motherboard manufacturers don't mind putting in a little extra effort, usually at the cost of a little bit of performance in order to go to a three bank design. However most motherboard manufacturers won't touch four DIMM designs with a 10 foot pole. It is very difficult to get the timings right while keeping performance up and eliminating any capacitance issues that may crop up with all banks populated. If you look closely at the AK31's layout however you'll realize exactly how Shuttle of all manufacturers was able to pull it off.

The four DIMM slots are almost perfectly flush with one another. This small but important reduction in trace path between the banks is most likely the reason Shuttle is able to outfit the AK31 with four DIMM slots and as we mentioned earlier on, this board works with all banks populated. With all four banks populated we did have to back off of our memory timings a bit, but the pros definitely outweighed the cons there. Remember that the board will support up to four DIMMs and they don't have to be registered either; a big kudos to the Shuttle engineering team that was able to pull this one off.

The board uses the same Award BIOS setup as the EPoX 8KHA+ so there's no real complaint in that department. When we originally started testing the board it had a much older BIOS on it (revision E6) which was horribly unstable. Using the latest E9 BIOS with official support for the Athlon XP, all of our stability issues were solved.

After using the newer BIOS the number of complaints was significantly reduced to just three. For some reason our unlocked Athlon XP 1800+ would randomly be set to 1700+ (1.47GHz instead of 1.53GHz) when the multiplier was set to Auto in the BIOS. We resolved this issue by forcing it to be 11.5x in the BIOS. The other complaint was the inability to set the memory clock to the FSB clock (host clock) - 33MHz which would be perfect if you're going to be overclocking to the 166MHz FSB. Luckily that option isn't too difficult to add in to a later BIOS revision. The final complaint and most unusual is that the AK31 would not properly set the memory timings when the BIOS was set to detect the proper timings from the SPD EPROM on the DIMMs themselves. We used SPD to set the memory timings on all of the boards to create a level playing field, but when we did so on the AK31 the performance plumetted. Setting the timings manually fixed this but there's no reason SPD shouldn't work the way it's supposed to.

While the AK31 R3.1 didn't come with any external USB headers, the board can be had for less than $100 which is definitely a positive.