The Legend of Zelda: The Wind waker CPU Benchmark

[Shonumi]

Nintendo Maniac 64 Wrote:Also due to the age of the N64, many more people have less moral issues with N64 emulation and even *gasp* downloading N64 roms.

Everyone but Nintendo, and really, they're the only one whose opinion matters. It concerns a bit more than morality, as the legality of copyrights and the DMCA have real bite to websites. We take those legalities seriously enough that any talk of downloading games is shutdown hard and quickly. Having said that, it's not like someone couldn't come up with a Public Domain or GPL'ed test ROM. Mupen64Plus includes a brief custom graphics demo; I'm sure someone could work out suite of tests that could be run. It doesn't have to be CEN64 either; could be higan or Dolphin. I only mention custom test ROMs since I've been making them myself recently (but for Gameboy debugging) :p All it takes is a bit of patience, familiarity with the system, and a flare for assembly.

[Nintendo Maniac 64]

Nobody says that the review site has to say what they're running, in particular Tom's Hardware doesn't say what video they use for their Handbreak benchmark - all they say is "VOB to MP4"

[Shonumi]

Nintendo Maniac 64 Wrote:Nobody says that the review site has to say what they're running, in particular Tom's Hardware doesn't say what video they use for their Handbreak benchmark - all they say is "VOB to MP4"

Are you suggesting to do emulation benchmarks without sourcing the games used? I'm asking because I don't know if that's what you're implying or it's just too late and I can't read right at night.

For emulation, you pretty much need to know what's running in order to do a benchmark. Every game is different, and that's going to affect what instructions and processes an emulator handles. That's not the case with Handbrake. All videos will still stress the CPU to its limits (the software is designed to do that); the question is how long will the task take. Handbrake quickly maxes out my cores during video transcoding, regardless of what the source is. The only difference is the total length of time it takes (which is the measurement for their Handbrake benchmarks).

Imagine having a review site or (heaven forbid) a published magazine that said "We ran Dolphin at XYZ FPS" and just listed the numerical results. I'd be pretty disappointed if they never revealed which game it was, because that's essential to reproducing their results, or even understanding their results in the first place. For example 35 FPS is really good for The Last Story, but abysmal for Mario Party 7. Even in other CPU intensive emulators (higan, CEN64, Mupen64Plus running LLE plugins, etc) the performance varies from game to game, unless of course your hardware is so CPU bottlenecked, everything more or less runs at the same, slow speed.

[NaturalViolence]

Shonumi Wrote:Imagine having a review site or (heaven forbid) a published magazine that said "We ran Dolphin at XYZ FPS" and just listed the numerical results. I'd be pretty disappointed if they never revealed which game it was, because that's essential to reproducing their results, or even understanding their results in the first place.

It's crucial for 3rd party testing/verification of results but it's not crucial to benchmarking. The benchmark would be about comparing cpus against each other, not about what cpu is needed to run a particular game at a particular speed.

[Shonumi]

NaturalViolence Wrote:It's crucial for 3rd party testing/verification of results but it's not crucial to benchmarking. The benchmark would be about comparing cpus against each other, not about what cpu is needed to run a particular game at a particular speed.

But even with benchmarks in emulation, the game affects what sort of task is being performed. If you don't know (and the benchmark never reveals) that an emulated game performs certain types of actions that one CPU (or any hardware component) is good at executing, but another CPU isn't, that's going to skew any conclusions drawn from the benchmarks. All you'll know is that CPU X did worse than CPU Y, but not necessarily know this was due to the benchmark favoring certain actions over others.

Unless someone really digs into how an emulator is actually going about the process of emulating a certain game, it can be difficult at best to determine which if any of its processes are being emphasized over others, even more so when the source of the benchmark doesn't cite the game used.

[Lumbeeslayer]

Can someone test The new AMD FX 9000series please, I wondering if it will perform good with DSp LLE on super mario galaxy will i get 60fps?

[Nintendo Maniac 64]

(10-10-2013, 11:18 AM)Lumbeeslayer Wrote: Can someone test The new AMD FX 9000series please, 吾 wondering if it will perform good with DSp LLE on super mario galaxy will 吾 get 60fps?

The FX 9000 series are not actually new, they're just essentially pre-overclocked FX 83xx CPUs; this is very similar to the way that you can buy a manufacturer-overclocked graphics card.

[NaturalViolence]

@Lumbeeslayer

Just take an fx-8350 and overclock it to 5.0GHz and the specs will be identical. It's not good enough to handle every level of SMG at fullspeed. Intel haswell cpus are still much faster. Even at stock.

[Nintendo Maniac 64]

AFAIK, the 9000 series, much like Richland, is more of a stop-gap until Steamroller and Kaveri.

[NaturalViolence]

Don't forget the 200 series graphics cards. Yeah AMD is basically taking all of their existing products in every lineup, overclocking them very slightly, and calling them new.