I think it is important that you have some way to monitor the processor temperature if you are going to push past the 2.3 volt limit set by Abit for the BH6.  Relying on the temp sensing chip on the motherboard of the BH6 is next to useless.  I am using a Radio Shack Indoor / Outdoor Thermometer which has been modified to allow me to attach a small probe directly to the edge of the slug on the Celeron A.  For info on how to modify the thermometer, look here. 

My celery is so unstable at 504 MHz using 2.3 volts and below, that it is questionable whether I can even get into Windows on a given try. I figured the old Star Trek line might apply.  "Give her more power, Scotty." 

To increase the voltage limits, it is necessary to re-flash the BIOS while the voltage setting is set to greater than the default 2.0 volts.  The available voltage will increase .4 volts over whatever the voltage is set to when flashed.  If you flash the BIOS with the voltage set to 2.3, this will give you a maximum voltage selection of 2.7 volts (which should be enough for even the craziest among us).

While I have been lucky enough to have never had a problem flashing the BIOS, many experienced users have.  Because of the potential for problems, Abit has posted a step-by-step procedure for the flashing operation on their site.  If you have not done this operation before, I recommend that you print the page and follow the instructions to the letter.  One noteable omission from Abit's instructions is that it is a good idea to go through each BIOS screen and write down your present settings before you flash.  Flashing will set your BIOS to the default settings and you will have to return the board to your desired configuration.

To get started on the testing, I wanted an idea of what the increase in voltage did to the Celeron running at a stable 464 MHz.  The following chart shows the increase in temperature during a relatively short test, WinBench99's CPUmark 32.  Please note that the room temperature is a chilly 64°.  These tests were also run using the Rain CPU cooler program which accounts for the relatively low starting temperatures.  On each boot up, the temperature would rise a minimum of 10° over the indicated starting figures before Rain kicked in and brought them down to manageable levels.  These tests were also run with the case cover removed so I wouldn't have to worry about heat build-up skewing the figures of the tests as I increased the voltage.

Though the temps don't look all that high with a room temp of under 65°, consider a warmer environment.  I have run enough temperature tests to know that if the room temperature was raised to 85°F, that the processor temperature will rise much more than a proportionate 20°.  At 464 MHz, 2.6 volts, and a room temp of 85+ degrees, this chip is going to be hot.

What effect did voltage have on performance with the CPUmark32 test?  None that I could detect.  The CPUmark hovered in the mid 950 range.  As has been my experience in the past, each test varies by a few percent even when no changes are made.  These tests were no exception.  I have seen no evidence that increasing the voltage either helps or hinders the performance of an already stable chip with the voltages I tested.