System Setup and Notes
MainboardDFI LanPartyUT nForce4 SLI-D, Socket 939, nForce4, PEG16X
ProcessorAMD Athlon FX-55, Socket 939, 2600MHz (13 x 200), 1MB
Graphics CardNVIDIA GeForce 6800 Ultra, NV45 + BR2, 256MB, PEG16X
Memory2 x 512MB Corsair XMS PC3200XL Xpert, Samsung TCCD, DDR400, 2-2-2-5
2 x 512MB Corsair XMS PC3200XL Xpert, Samsung TCCD, DDR500, 2.5-3-3-7
Hard DisksWestern Digital Raptor, 36.6GB, SATA
CoolingAsetek's VapoChill XE-II
Asetek's VapoChill LightSpeed[AC]
Windows XP Professional w/ SP2
NVIDIA nForce4 Unified Driver Package 6.39
NVIDIA Detonator 66.93
DirectX 9.0C End User Runtime
LAME 3.92MMX encoding U2's Pop album at 192CBR
Realstorm 2004 Raytracing Benchmark
ClockGen for nForce4
CPU-Z v1.26 and v1.27
Corsair Memory Dashboard 1.0
Above lies a rare insight into the depths of HEXUS. A shot of my home office where I work, taken across the bow of my cluster, you can see the VapoChill XE-II partly hidden by the cluster and next to a blacked out section of the image which hides some unreleased NDA-bound hardware. On top of my wonky test station lies the VapoChill LightSpeed, adorned by a Tagan TG420-U02 power supply and chilling FX-55 inside an ASUS A8V Deluxe v2.0, which I used for initial testing before the DFI arrived. You can see that setup doing SuperPI successfully to 1M places at 3.15GHz, here.
Finally, if you excuse the poor picture quality - it was the dead of night under poor light, the camera flash was really spoiling the image and I had no time to find my tripod - you can see a one-shot photograph of the approach to 3.5GHz.
As with all finding-the-limits overclocking, you can go past the stable limits by a certain margin and while I could ClockGen 3500MHz, ClockGen was about all that would run. The quest to keep it stable around 3400MHz continues.