There are surprisingly few caps on this board, and all are made by Jackcon (at least in my case). Obviously ABit was pretty gung-ho about minimizing their capacitor expenses with this board, and by the late 440BX era they had figured out what didn't have to be there.
Most of the caps are 5V input or Vcore, with a small number of large caps on other jobs. Total Vcore capacitance is very high compared to the other BX boards I've looked at. There's a surprising lack of caps in the PCI slot area, and only a single 100uF cap on the 3.3V line. There are only 2x 22uF caps in the entire bottom half of the board. Despite how minimal the cap layout looks, it seems to run well when it works.
It's just too bad they used those horrible Jackcons.
There are only 5 small electrolytics, so a full recap doesn't take long to accomplish here.
AGP is apparently fed by an onboard regulator, and is not connected to the PSU 3.3V line as it sometimes is on other boards.
The incredibly weird thing about this board is the location of the Vagp caps. They're just about as far away from the AGP socket as they could possibly be. I checked it with the multimeter several times because it didn't make sense to me, but those caps are definitely connected to the Vddq pins on the AGP slot.
BIOS: there's an interesting tweak in the SoftMenu III bios which had me fooled for a long time. Long ago, I set up a BX133-RAID for my sister with an 866/133 cpu, but was confused why the memory bandwidth in memtest86 was literally HALF of what it should have been. I gave up, assuming it was some individual board problem.
Fast forward to when I was fiddling with this BE6-II recently, and I had the exact same issue. Memtest86 bandwidth is half what it is on other 440BX boards I've tested. I finally figured it out - Abit has a setting called "In Order Queue Depth". If you overclock the chipset, they automatically change it to 1. Put it back to 8 and you'll recover the speed a 440BX is known for. My board was still perfectly stable at 133fsb that way.
How my board worked out:
===============
All capacitors were replaced except for the pair of 22uF caps near the southbridge. There were only a few 100uF caps which I had some replacements for, and everything else is 1000-1500uF.
I didn't use the same size caps on the Vcore circuit. Originally, ABit used 8x 1500uF Jackcon caps. I figured I might as well try Oscons instead. This board now feeds the processor with 8x 560uF/4V Oscon SP's.
I know, that's a big drop in capacitance, but it's still double what I had on a previous socket-370 board. I have a hard time believing all Abit's original capacitance is really necessary, unless maybe because they were using Jackcons. Even so, I considered this very experimental, but I wanted to see what would happen.
It was a success, the ABit BE6-II is perfectly stable with these caps. I tested it on prime95 with a Celeron 1100MHz on a slocket.
There are 2 lingering problems with this board.
1) One of the ATA33 ports is dead. I was disturbed to find in a Google search that many other people have had similar problems with dead drive channels on this board. At this point I decided to replace those Jackcons by the southbridge, but the only 22uF caps I could find were some general purpose caps by a brand I'm not familiar with:
The Jackcons had an ESR of 2.2 and 2.3 ohms each, the replacements were 1.1ohms.
They didn't fix anything, the channel is still dead. I hope that problem doesn't spread, because my nephew wants a file server and right now this looks like the best board I have for the job.
2) The health monitor reports that the -5V is running at -62V. This is absurd of course, but that same screen in the BIOS will lock the computer if it sits there too long. My guess is that the bios code just doesn't handle the extreme value very well and crashes because of a software glitch. The board is stable everywhere else, so it's not a big issue. Same thing happens on 2 different known good power supplies.
Comment