Encoding

One thing we typically do not include in our processor reviews is video encoding. We have been searching for a really consistent means by which to test the encoding capabilities of processors for a while now, with some unimpressive results. However, we very recently came across a fairly good vehicle for testing H.264 encoding times, and we are pleased to see that Intel used the very one we had been considering. The Mainconcept encoder platform used here paints a good portrait of processor performance.


The numbers to pay attention to here are those of the quad-core chips. Because video encoding on this platform is multi-threaded, the dual-core processor is pretty much guaranteed to fall behind its quad-core counterparts. The performance difference here, though slightly more than what would be expected of a simple speed-bump, is nothing out of the ordinary.

Cinebench

We typically stay away from using the Cinebench test at GotFrag Hardware only because we feel we have a more competent sampling of synthetic benchmarks in our lineup already. However, Cinebench is a very good test of overall system performance and can be useful in illustrating differences in processor performance.



Here we see fairly dramatic performance increases over the QX6700 with the Penryn quad-core chip. Since Cinebench is a multimedia test that involves rendering three dimensional scenes, it is reasonable to suggest that the SSE4 capabilities of the Penryn processors play a huge part in the performance differences we see here. We also suspect that the increased cache sizes on the Penryn cores have something to do with, as calling textures from disk would seem to be a common occurrence when running Cinebench.

DivX with VirtualDub

Intel was pretty vague about the actual procedures they used when running these tests. Because of this, we have no idea what Intel did within VirtualDub to yield their results. We suspect there was some sort of post-processing or encoding to DivX format that went on, but this is just speculation. In any case, here are the results.


Here we see the most dramatic results of them all. Performance of even the dual-core Penryn chip is seemingly light-years ahead of the QX6700 in this test. Taking into consideration that this test was not run differently between the three processors (assumption, but we think Intel is good for it) and that there is a relatively small (4 seconds) performance difference between the two Penryn chips, it would seem that the SSE4 instructions really boosted performance exponentially here. The wider dynamic execution courtesy of the Radix-16 divider probably plays a fairly large role here as well.