> > > If you want to compare against RB's setup, benchmark MAME 0.141 instead. This > > > especially goes for sh4 cpu optimizations(dolphin), and namcos22 additional gfx > > > emulation(propcycl) > > > > I used RB's scores because they the closest I could find on short notice, but if I > > have enough time today and I can find a suitably spec 2600K PC I'll see if I can > get > > it to 4.8Ghz and run a 143u9 benchmark run on it, no promises on this but. > > That's not what he's saying. He's saying MAME performance on the same hardware > changed dramatically in several of those games between 0.141 and now due to > optimizations (dolphin) and pessimizations (propcycl). > > And for the record, memory bandwidth is the mother's milk of emulators, followed > closely by instructions per clock.
Yeah, I got what he was saying, I just figure in in lieu of you running new benchmark on you system with 0.143u9, I could run new benchmarks on another I7-2600K with MAME 143u9, to give a true comparison between the two CPUs.
It's interest what you say about the memory bandwidth, the I7-3930K has no shortage of that, but I might have a play and see what happen the benchmark results if I force it into dual channel or drop the memory clock, to see just how much the effect the drop in memory bandwidth has on MAME.
I'm not questioning what you say, I know how much memory bandwidth can effect some applications, I groan everything a see a single Dimm of memory in a quadcore system, I'm just interest to see the results.
If all else fails, Burn the manual.
|