AMD's .09 micron A64 3500+: Overclocking, Thermals, & Power Consumption
Overclocking The 3500+
For our first test, we slowly raised each processor's FSB speed via our motherboard's BIOS until the test system was no longer stable. Please note that each processor was running at its stock voltage, 1.4v for the Winchester CPU and 1.5v for the Newcastle. The heatsink used to cool each processor was the model that shipped with AMD's flagship FX-55 processor. It is an aluminum finned heatsink with a copper base, and copper heat-pipes. You can take a look at it right here.
To be certain the CPU was the limiting factor in our overclocking attempts, we dropped the HyperTransport clock to 4X (800MHz) from its default of 5X (1GHz), locked the AGP/PCI clocks, and ran our memory at timings of 2.5-3-3-10 at 2.8v. At these settings, this configuration topped out at an FSB of over 260MHz in a previous article...
With both of our Athlon 64 3500+ processors operating with their default multiplier of 11, and with our MSI motherboard configured for a 200MHz FSB, both processors ran at a "default" clock speed of 2210MHz, which was 10MHz above the expected speed of 2200MHz. To begin the experiment, we then raised the FSB via the motherboard's BIOS menu to see which CPU would top out first. In the end, the .09 micron, core based 3500+ hit a maximum clock speed of 2596MHz (11 x 236MHz), besting the .13 micron Newcastle by only 55MHz. At default settings, our particular .09 micron A64 seemed a bit more "overclocker friendly", but the clock speed difference wasn't terribly substantial.
Please be aware that there are a number of other measures we could have taken to achieve even higher clock speeds in this type of test. We could have used a more capable cooler and increased each processor's core voltage, among other things. But our intent was to see how high each CPU would go at AMD's recommended default voltage. And all things being equal, it seems like at least our particular Winchester based 3500+ was the better overclocker.