Our method for determining a successful and stable overclock relies on using software tools to stress test the machine to the point of failure. If it fails, it's not stable, and thus not worthy of being clocked at that speed. We also actually install our boards in a real honest to goodness case, fully assembled and closed, with no 36” box fans blowing over the components. No air conditioning systems were harmed during the making of this review either, and we did not have to walk around the lab in overcoats wearing scarf's and skull caps. Being stable on a workbench, and being stable installed in an actual case in an actual working environment are often two different things.

Step one for us, is building the system and installing the requisite software. Leaving the system at stock clockrate, we proceed with a burnin test using the following software simultaneously for at least 12 hours...

• Sisoft Sandra 2007 SP1 Burnin Test
• Prime 95 v2.414
• S&M Burnin
• HeavyLoad V2.1

Once this is done and no errors are encountered, we begin the overclocking process, finding the right mix of front side bus speeds and voltages needed to obtain what we “feel” is a stable overclock. Once we settle on a given clock, we then do the above tests again to insure what we “feel” matches reality.

To make a long story short, we settled in on taking frontside bus from 266MHz to 450MHz as our stable point.

• Frontside Bus – 450MHz
• System Memory Multiplier – 2 (DDR-900)
• DDR Overvoltage Control - +.4 (Our Crucial memory is rated @ 2.2v for DDR-1000)
• FSB Overvoltage Control - +.1v
• MCH Overvoltage Control - +.1v
• CPU Voltage – 1.43125v

When attempting a 460MHz FSB we encountered very strange issues relating to the video card and the onboard Realtek HD audio. At random, with no apparent predictability, when dragging some windows around while listening to audio playback, the audio would go into slow motion and the system would grind to a near halt. With nothing to go on to determine why this was happening, we can only guess that it has something to do with the “Robust Graphics Booster” setting in the BIOS that cannot be disabled. Backing off to 450FSB eliminated the issue.

While we're discussing issues, we should highlight two other issues we encountered. When using the ATI Catalyst 6.9 drivers along with the .NET driven Catalyst Control Center, enabling TV Out caused both displays to go dark, with no way out short of a reboot. Utilizing the standard Catalyst 6.9's also resulted in this weird behavior, even at stock clock rates. Installing the Omega 3.8.291 driver package (which is built on Catalyst 6.9) resolved this issue. Also, one piece of software in our standard testing suite, PCMark 2005, simply would not complete at any clockrate, failing at the web page rendering test. We can only assume this is an oddball errata with this synthetic suite and our system, as we encountered no other failures with any other software we used.

So, we ended up at 3.15GHz at a 450MHz front side bus. Having seen other more successful overclocks with the B2 stepping E6300's elsewhere, we can only assume that the board itself is holding us back. Still, a 1200MHz overclock is pretty sweet. Considering our utter lack of control over memory timings we can also assume this level of overclock can be achieved with considerably less expensive memory than the Crucial Ballistix we were provided with for the review as well. The heatsink we chose, the Artic Cooling Freezer 7 Pro, wasn't an uber-expensive heatsink either. So, we feel confident that anyone, especially someone on a budget, could duplicate our results on the cheap. Still, it's clear this is due to the strengths of the E6300 itself, rather than the merits of this Gigabyte board. As you can see from the image below from CPU-Z, memory timings are a tad on the sloppy side.

Even with the sloppy memory timings, this critter is quite fast. Let's take a look at our mix of synthetic and real-world benchmarks next.

note: The latest F8 iteration of the S3's BIOS has allowed us to maintain the 450fsb overclock *and* tighten memory timings to 4-4-4-12 @ 900mhz DDR.