Shuttle SD31P Small Form Factor XPC

Examining the BIOS of the SD31P The "setup" for our setup

As with just about everything else with the Shuttle SD31P, the BIOS appeared to be a slightly updated version of the Phoenix / AWARD variety we've used before.  There are many options that come standard on the FD31 motherboard, so running through the screens can take some time to ensure that everything is set up correctly.  The ICH7-R southbridge natively supports SATA and IDE hard drives, and can also be used to configure RAID matrices, in 0, 1, 5, and even 10 modes.  In total, the SD31P could have an IDE hard drive and two SATA drives, or even three SATA drives installed and running in whatever configuration suits your needs best.

The fans can also be configured for high speeds or low noise.  Those bent on overclocking their system to the max should probably steer more towards the mid to full fan speeds, but the noise output will definitely be noticeable.  We could relate the sound to a hair dryer when set to the Full Fan Speed mode.  But, if high-end gaming is not a main concern, choosing Ultra-Low Speeds, or even the Smart Fan mode, is the better choice, with nearly quiet operation being the end result.  Another small addition, which some users might find handy, is the ability to control the brightness of the LED around the power switch.  Some might find the bright blue light to be a distraction, so Shuttle has provided percentages from 100% all the way down to 0%, thereby shutting the light off altogether.  

Overclocking Tools When fast is not fast enough

The SD31P has a rudimentary set of tools for overclocking the unit.  First off, we disabled Spread Spectrum modulation, as keeping this enabled could cause some instability when raising front side bus speeds.  The CPU clock speeds fall within a range of 100MHz and 355MHz, and a full range of voltage options, both higher and lower than the default 1.325V.  As the 945 and 955 series of boards come with support for higher memory frequencies, we found that the FD31 offered options for automatically detecting the frequency via SPD on the memory, or manually choosing 400MHz, 533MHz, or the new 667MHz standard.  The Corsair DDR2 we tested with required us to raise the voltage from a default value of 1.9V to 2.1V, but this was as high as we could go, which could stop us from getting too far without relaxing the memory timings to a degree.  Again, as with the SB95P, there weren't any options to set the PCI-e clock to a defined speed either.  There's just enough here to get a decent overclock, and usually that's all we can expect for in a SFF PC.  

CPU-Z Overclocked results SANDRA2005 Benchmark

Based on previous attempts with other 945 boards, we knew that there was some potential for getting high gains with the Pentium D 820.  As such, we immediately set our sights for the 240s, but found that we could not get the system to reboot, nor could we get back into the BIOS.  Not being able to find the Clear CMOS button on the outside of the chassis came back to bite us here.  Instead, we found ourselves manually resetting the BIOS a few times until we found the right combination of settings that worked.  By leaving the Corsair memory at SPD timings, and the frequency set to AUTO, we were able to reach a high-point of 240MHz for the FSB.  Benchmarks, however, took a turn for the worst, as the CPU throttled down, possibly due to some high temperatures that the ICE module was not prepared to fully handle.  Performance was actually less than half that found at stock speeds, and raising voltage as high as 1.5V on the CPU didn't help stablize things any.  If anything, it just increased temperatures higher, so we decided on cutting back some. Through a torturous process of dropping speed, testing, dropping voltage, retesting, and so on, we found that the sweet spot for our foray into overclocking the SD31P came at a 224MHz front side bus.  CPU VCore was set at 1.375V (reported as 1.325V in CPU-Z) with the memory set at a 3:4 divider, thus running at approximately 321MHz.  Obviously this was not as high as the 255MHz FSB that we were able to obtain earlier, but we still considered this a success, equalling about a 12% jump in speed.  Our CPU - Arithmetic score in SANDRA jumped up about 1600 points, which is about a 10-11% increase in performance.  During all of this testing, however, we kept fan speeds to the maximum levels possible, which might not always be the most desirable setting for those seeking quiet, peaceful operation.