Benchmark Results: Sandra 2012
At least on paper, Intel’s Core i3-2100 would seem to have an advantage in instruction per second throughput and floating-point operations per second. Whether Sandra’s synthetic measurements translate into real-world performance remains to be seen.
AMD demonstrates exceptional integer performance in the multi-media test, which we’d expect given its four physical integer cores. Floating-point performance suffers, though, due to the Piledriver architecture's shared resources, and Intel’s Core i3-2100 takes a first-place finish.
We didn’t point this out last time, but the twin Piledriver modules in A8 and A10 manage to deliver better FP performance than the four physical cores in both Llano-based contenders.
As part of its product differentiation strategy, Intel pulls AES-NI support from the Core i3s. Thus, you see its -2100 drag along at the bottom of our chart as all three Trinity-based APUs achieve respectable AES256 bandwidth.
No surprise—Intel’s dual-channel memory controller dramatically outperforms AMD’s best effort.
It is strange, though, that both Llano-based APUs push higher throughput than the three Trinity-based parts.
Core i3 takes the same L1 data cache hit as Trinity (compared to Llano), but its L2 latency is almost half of AMD’s upcoming APU. Llano’s L1D and L2 both appear to suffer less latency than Trinity, too.