AMD Athlon 64 3800+: Probing the Venice Core
Temperatures: Stock & Overclocked
For our next batch of tests, we recorded each processor's core temperature while running at their respective default clock speeds, and while overclocked to the same level. The "idle" temperatures were recorded after powering up the system and letting it sit at the Windows desktop for 5 minutes. The "100% load" temperatures were recorded after running the graphical Folding @ Home client for exactly 15 minutes.
Processor temperatures were monitored using the latest version of Gigabyte's proprietary Easy Tune 5 software. Please be aware that motherboards from different vendors will report temperatures differently; an Asus or Abit motherboard, for example, may very well report temperatures a bit higher or slightly lower than the GA-K8NXP-SLI we used here. The temperatures reported below are only indicative of how Gigabyte's software interprets the signal coming from the Athlon 64's on-die thermal diode, and aren't necessarily 100% accurate representations of actual die temperatures.
At their default clock speeds, there was a fairly significant temperature difference reported between the two cores. The .09 micron Venice based 3800+ was a full 5oC cooler at idle, and about 7oC cooler while running with 100% utilization.
To compare temperatures while our processors were overclocked, we ran BOTH CPUs at 2582MHz, which was the maximum speed our .13 micron Newcastle based 3800+ was stable using its stock voltage. Our particular Venice core based CPU was capable of more, 2640MHz to be exact, but we wanted to have an "apple-to-apples" comparison in this test. While overclocked to 2582MHz the Venice core based 3800+ ran 6oC cooler than the Newcastle based CPU at idle. With both processors running under a full load, the newer Venice based 3800+ again ran cooler, but this time is was by full 11oC.
We should also note that both of these processors ran relatively cool by today's standards. Even while overclocked well beyond their stock speeds, and running at full load, neither CPU got hotter than 45oC. We'd also like to point out that even with the Venice core's increased thermal density, it was able to run cooler than a similarly clocked Newcastle using the exact same heatsink. When you shrink the size of processor's die, it has a smaller area through which it must emanate heat, which makes them more difficult to cool. The Venice core doesn't require any additional cooling, however. AMD's existing PIB heatsink is more than capable of keeping temperatures in check, and the aftermarket all-copper ThermalTake cooler we used here had absolutely no trouble at all. That is, if the Gigabyte board was reporting actual temperatures somewhat accurately.