Standard Def Steve wrote:It's actually a 2.6GHz Opteron 185 for S939, which was around during the Pentium D days. I just happen to have it overclocked to 3GHz because it handles well @ default voltage. But even at stock clocks, it outperforms the 3.2 P-D. 1080 VP9 is still playable on it.
It's a faster CPU in general, that's not the point. We were looking at the SSE2-performance only.
Standard Def Steve wrote:It sure seems that pure SSE2 can sometimes be over twice as fast, at least in the synthetic benchmarks swaaye linked to. 😀
Are we looking at the same benchmarks? Because the ones I'm looking at, have two mandelbrot benchmarks. One is integer, the other is SSE2. Core2 Duo totally cleans up in the integer one, but with SSE2, Pentium D is a whole lot closer.
Of course this is not a 'pure SSE2' benchmark, since it is a complete mandelbrot program. There's still a lot of non-SSE2 code in there, where the Pentium D takes a hit. If you were purely benchmarking the SSE2 code, you'd see that Pentium D is doing quite well in that part of the code, and losing most in the integer-part (as should be obvious from the comparison with the integer version).
Standard Def Steve wrote:A 1.86GHz Core 2 beating a 3.6GHz P-D in SSE-heavy media encoding and rendering tests is considered to be having trouble?
Given the overwhelming performance advantage that Core2 has over Pentium D in most other benchmarks, yes, I'd say it's struggling here. Relatively speaking of course.
Also, please don't do any comparisons on clock speed. I thought we've dealt with the MHz myth already...
Standard Def Steve wrote:So yeah, at least in video decode performance, it outperformed the PD-935. I dunno; it just seems that pure SSE performance doesn't matter too, too much outside of synthetics and maybe x264. Unless Google's video decoder isn't SSE heavy.
It tends to matter more in encoding than in decoding. But then there are still codecs that aren't very well-optimized.