Performance

We’ll first compare the processors in FutureMark’s benchmarks.

The new Sempron 3400+ is comparable to both Athlon 64 3000+ models but is slower than the competing Celeron D 351 in the overall performance score.

The overclocked Sempron 3400+ becomes about 15% faster and beats the rather expensive Athlon 64 3800+. The overclocked Celeron D 351 is still on top, however.

The CPU test from the same benchmarking suite is favorable to the NetBurst architecture. As a result, the Celeron D 351 at the default frequency is faster even than the overclocked Sempron 3400+. As for the performance of the non-overclocked Sempron 3400+, it is better than either version of the Athlon 64 3000+. It may seem strange since the Athlon 64 3000+ for Socket 754 has the same frequency of 2GHz as the Sempron 3400+, but has a larger L2 cache (512 kilobytes). This mystery has an explanation, though. All Athlon 64 for Socket 754 are based on the older 130nm NewCastle core of the CG revision. The Sempron 3400+ features a new 90nm Palermo core of the E6 revision which has some advantages over the NewCastle. Particularly, the Palermo has an optimized memory controller and supports SSE3 instructions. These factors help the Sempron 3400+ triumph over the Athlon 64 3000+ in this test.

The memory subsystem benchmark yields predictable results: the K8 processors with a dual-channel memory controller are better than their single-channel analogs, while the Celeron D only performs well when the system bus is overclocked.

The CPU test from 3DMark05 puts the Sempron 3400+ on higher positions than the Celeron D 351 in the default mode as well as at overclocking. This is another confirmation of a highest efficiency of the K8 architecture in gaming applications. Note also that the Sempron 3400+ is faster here than the Athlon 64 3400+ for Socket 754. Despite AMD’s claims that there’s little difference between different cores, we can often see the Palermo win against the NewCastle even though the latter has a larger L2 cache.