LOSTCIRCUITS

In Prime95, the Toledo core is about even with the Newcastle and below the power consumption of the Clawhammer at the same speed. Keep in mind, however, that the CPU is only running under 50% load. Shown are the first and fourth iterations to factor in the increase in power consumption as the CPU warms up. Things change dramatically if two instances of Prime95 are loaded simultaneously and hard-assigned to core 0 and 1, respectively. In that case, we are seeing starting power draw of about 79 W, which, as the CPU warms up increases up to 86 Watt.

Alternating runs of Arithmetic and Multimedia Benchmarks in the Burn-In test for Toledo (red) Clawhammer (green) Newcastle (pink), Winchester (blue) and Venice (turquoise). As mentioned earlier, the multimedia benchmark stresses the CPU to a lesser degree than the Arithmetic benchmark, which is visible also in the "Performance monitor", where the CPU Load drops to some 95% in the multimedia benchmark.

Even though we are monitoring 100% load here (the dips indicate switching between Arithmetic and Multimedia), the burn-in test does not stress all units within the Toledo at the same time, rather it appears that it is alternating between ALU and FPU units. This explains why we are "only" witnessing some 80 W as opposed to the maximum power of 109 given by AMD. Again, in real world situations, there is no possibility to reach the ICCmax and maximum thermal power, the official numbers are guidelines for the system builders regarding what their components have to withstand in order to be qualified for use with a given CPU.