Overclocking

As much as we know people love to play with DFI boards, I can't help but find them a bit of a headache when it comes to overclocking. There are just so many options for most of us, and trying to diagnose why Windows locks up or blue screens is generally very difficult – is more voltage required? Which setting should I change? PLL/VTT/GTLs? Are the memory settings wrong? Which one? And how I pined for a clear CMOS button! The LED POST readout is invaluable though – generally speaking, it helped a lot when the board failed to POST.

After the best part of a day overclocking it we still felt like we'd only scratched the surface – never more was the "dark art of overclocking" more applicable. We couldn't get a quad core QX9650 over 450MHz FSB on a 6x multiplier and our E8500 topped out at just over 500MHz FSB with 515MHz crashing after some 30 minutes of Prime95 torture test, but given enough time to play with the GTLs this instability could have been resolved.

We also spent a considerable amount of time consulting VTT and GTL tables trying to work out the appropriate values for our 45nm dual- and quad-cores which require a lower 61-63 percent GTL threshold compared to the 65nm chips that require 67 percent.

While the PLL voltage is also key, we found that unlike the other voltage options, this was not so fine – there were larger jumps of 0.05-0.1V, so we quickly hit a 1.7V limit we weren't prepared to go over (save killing our 45nm CPUs). We also found the GTLs did a better job of making the board stable than increasing the PLL excessively.

We added an extra fan over the north bridge and PWM area because like with the Gigabyte GA-X48T-DQ6, they both get very hot as you start to increase the voltage through them, even despite the excessively larger than average north bridge heatsink.

We found the north bridge/memory divider at 333/667 was a bit better than 400/800 and a north bridge core voltage of 1.5-1.65V and CPU VTT from 1.25-1.35V was generally about the right range to play in. We decided to try and simplify things as much as possible by leaving the BIOS to organise the memory itself.

Using voltages and memory settings we know worked with other boards, the DFI did remain somewhat stubborn – we knew it had the capacity for some amazing clock speeds but the stability well is very sharp, and finding it is difficult. We're certain you can get something great out of it – everything is there at your fingertips, but getting there is an incredible ball ache and unless you've the patience of a saint and the time of a student, you'll not get the most out of this board.

Power Consumption

DFI offers no specific power saving enhancements, so its eight phases of digital PWMs are on all the time and tend to run really quite hot, especially if you overvolt a quad-core processor. We set the system to use a specific Core 2 Duo E6750 G0 with GeForce 8800 GT and 2GB of memory. The BIOS was also set to its "optimised default" values.


The lack of power saving hardware clearly shows as well, with the DFI coming out bottom in our table. The board is built for speed but it isn't the only one with a digital PWM system rather than the usual MOSFET and iron choke setup – the Abit IX38 Quad GT features a five digital PWM as well, and that also falls in the mid table.

We can't even attribute the extra power to extra features either as the other boards are equally as well featured in general – it simply uses more power.