Overclocking

ABIT's boards are well known for their excellent overclocking features. For example the uGuru overclocking technology is a unique feature of ABIT motherboards.

Before testing, I browsed a few reviews of other websites and was disapointed that they could not get more than about 220 FSB out of the board.

My initial testing confirmed this.
With our P4 3.0 CPU, which can easily run 300 FSB = 4500 MHz on other boards, everything ran fine and the AL8 was 100% stable up to 224 FSB. But once the FSB was changed to 225 MHz, the system immediately crashed. It made no difference if the clock was changed in BIOS or in Windows with the uGuru utility. Raising voltages did not help either.

After some experimenting I found out that changing the "N/B Strap CPU as" setting to PSB1066 allowed me to go to up 299 FSB, but after that there was the same instant-hang wall.

The N/B Strap CPU setting controls what kind of speed grade the Chipset thinks the CPU is. The available options are: 533 / 800 and 1066. Usually, all this option does, is change the available memory divider ratios, but in this case it seems to change something in the chipset too. My guess is that the i945 chipset has some kind of overclocking lock.

So that's it? 299 FSB maximum? No. Once you start increasing the PCI-Express bus frequency, you slowly gain more overclocking. The graphs below show how PCI-E frequency and maximum FSB are connected together.


Here are the results at N/B Strap setting 800 FSB. Looking at the graph, you can see that adding 1 MHz PCI-E clock adds 2 MHz of extra FSB.


Here is the data for the 1066 FSB setting. It starts out at 299 FSB and increases 3 MHz, every time you raise PCI-E clock by 1 MHz.

I did not test the 533 FSB setting, but I would assume the steps are 1 Mhz there.

Be advised that you can not raise the PCI-Express bus clock indefinitely. First, the board BIOS offers a selectable range between 100 MHz and 150 MHz. During my testing the SATA HDD was no longer detected at around 120 MHz, the ATI RADEON X850 did not start up beyond 125 MHz. With a Matrox Mystique PCI VGA card I could get 129 MHz, but more than that and the mobo wouldn't POST.

Based on above testing, and assuming a maximum PCI-E clock of 125 MHz, the maximum clock the board can run at is 374 FSB (= 299 + 25 * 3). At PSB 800 this would be 275 FSB (= 225 + 25 * 2).
I see no reason why you would want to run the board at the 800 PSB setting.

Now that we were able to properly overclock the board, we set out to find the maximum FSB we could actually run with our CPU. To do this the multiplier was dropped to 14x using EIST. Memory was run at the slowest divider as well, so that only the CPU or motherboard are limiting the overclock.



The maximum stable FSB was 326 FSB - the CPU's limit. With the right CPU you can sure get more out of the board.



The highest "performance" clock we could reach is 4550 MHz at a FSB of 303 Mhz. However, the board felt a bit unstable beyond 300 MHz. For example PCMark2004 tended to crash sometimes, even though the system was 24 hours Prime95 stable (two instances).

Overall I must say that the board offers solid overclocking, if you know above BIOS tweaks. Without them, a maximum of 225 FSB would make the board a hopeless overclocker.

Also be advised that for more serious overclocking adventures you should consider cooling the voltage regulation MOSFETs around the CPU area. When we tested at 4500 MHz with a waterblock (no airflow around CPU), some parts there reached temperatures of around 90°C-100°C.