Benchmarks: Single- and multi-core
AMD claims to have boosted single-thread performance via a number of enhancements to the CPU core found on the Richland APUs. The HEXUS.Pifast test uses simple code that doesn't benefit from the improvements made by AMD from the Piledriver to Steamroller cores, and we know this by the linear performance progression when examining the A8-7600 (Kaveri, 3.8GHz maximum clock) and A10-6800K (Richland, 4.4GHz max) numbers. A faster-clocked Kaveri wouldn't be much faster, if at all, than the Richland chip of the same frequency.
Yet drawing conclusions when evaluating AMD APUs alone is a moot point; Intel's superior single-threaded performance is very much in evidence.
The latest release of Cinebench provides a roster of new scores. AMD has certainly been able to pack the 45W A8-7600 with handsomely greater processing power - it is almost 60 per cent faster than the A8-6500T - and it's mainly due to the energy-efficient nature of the all-new 28nm SHP process and beefier processing cores.
The A8-7600 can be configured in either 45W or 65W modes, and choosing the latter bumps up performance by almost 13 per cent, with the gain realised through higher CPU frequencies. Improvements to the core aside, AMD's slowest Kaveri chip isn't supposed to beat the best of the last generation, which is almost 10 per cent faster still. Intel's Core i3-4330's dual-core, quad-threaded architecture once again streaks ahead.
wPrime really pushes the cores to the limit, so much so that any gains in core efficiency make a huge difference. Here is the first example of Steamroller cores' improved architecture coming to the fore; even the frequency-hobbled 45W APU's scores are better than the A10-6800K Richland chip's.
It is also easy to see how much potential there is in the Kaveri architecture - the also-45W A8-6500T is miles behind. Intel's comparison Core i3-4330 puts up the best first-page showing, however.
