Dolphin, the GameCube and Wii emulator - Forums

I made a chart out of the Dolphin benchmark results.



I didn't include the processors that only had one or two submissions, and Nehalem would have been in the same place as Sandy and Ivy, so I left it out too. It does make it clear that clock speed is much more important than architecture, though. Joe Lie's CPU would take 20-25 min to run the benchmark. What games will be playable at that level?

I just threw this together in a few minutes, but I think a high quality version with some horizontal divisions like "No you can't run Dolphin", "Some games will be okay", and "Almost everything will run at full speed" would be really helpful. It might reduce the number of "Is my computer fast enough for Dolphin?" threads.

I have one note for the above graph though, the difference between time, 10 minutes and 5 minutes for example is 100% or double the speed which in the graph above it is not apparent. Basically going from 30 minutes to 15 is similar speedwise to going from 10 minutes to 5.

I'm not sure what you mean by that. The graph I made represents performance on the benchmark based on architecture and clock speed. If you are familiar with the relationship between benchmark performance and in game performance, you sound like the right person to draw some horizontal lines across this graph that indicate how successful a CPU will be when running Dolphin.

(07-01-2014, 04:32 PM)artantaaa Wrote: [ -> ]I'm not sure what you mean by that. The graph I made represents performance on the benchmark based on architecture and clock speed. If you are familiar with the relationship between benchmark performance and in game performance, you sound like the right person to draw some horizontal lines across this graph that indicate how successful a CPU will be when running Dolphin.

It represents performance but indirectly relative to time spent. For example if you used a speed percentance piledriver would have been about half the distance of haswell in the y axis at the right portion of the graph. Saying ghz is the most important aspect is therefore incorrect.

Make a graph with a time scaling like this: 30:00 and then 15:00 and then 7:30 and then 3:15. Therefore each step will represent double speed to the previous step and the speed differences will be more clear.

Or plot reciprocal time, which is speed, and therefore more useful.

^This guy knows how to data.

artantaaa Wrote:Joe Lie's CPU would take 20-25 min to run the benchmark. What games will be playable at that level?

Impossible to say with just that information. Too many other vairables to consider.

artantaaa Wrote:I just threw this together in a few minutes, but I think a high quality version with some horizontal divisions like "No you can't run Dolphin", "Some games will be okay", and "Almost everything will run at full speed" would be really helpful. It might reduce the number of "Is my computer fast enough for Dolphin?" threads.

Same problem as above. Plus I already did something like that awhile back. The thread got destickied because it wasn't considered specific enough: https://forums.dolphin-emu.org/Thread-dolphin-cpu-hierarchy-unofficial

I wrote about the issue recently here: https://forums.dolphin-emu.org/Thread-random?pid=325985#pid325985

Wait....why am I posting in the support subforum? I'm not supposed to be here. Damn you kinkin!

Quote:It represents performance but indirectly relative to time spent. For example if you used a speed percentance piledriver would have been about half the distance of haswell in the y axis at the right portion of the graph. Saying ghz is the most important aspect is therefore incorrect. Make a graph with a time scaling like this: 30:00 and then 15:00 and then 7:30 and then 3:15. Therefore each step will represent double speed to the previous step and the speed differences will be more clear.

I see what you mean. We wouldn't want people to think that Piledriver is almost as fast as Haswell. I didn't mean that GHz is more important than architecture, but it does make a difference in cases like this, and that's why I made the graph. Until I looked at the benchmark results, I wasn't sure whether Wolfdale@3.6GHz would be faster than Sandy Bridge@2.1GHz.

Quote:Or plot reciprocal time, which is speed, and therefore more useful.

That did cross my mind, but I didn't want to convert all the times to seconds.

Quote:Impossible to say with just that information. Too many other vairables to consider.

Things like the GPU, Dolphin version, OS, etc.. are not taken into account on the graph. If we assume that none of those things are causing a problem and they are using the latest version of Dolphin, I would put Joh Lie's CPU in a category like "Some games will run at playable speeds, probably not full speed."

Quote:Same problem as above. Plus I already did something like that awhile back. The thread got destickied because it wasn't considered specific enough

I remember your CPU hierarchy thread. It's really helpful. I think the graph format would help in the "Thoroughness vs. WTF am I reading?" department. You could include every architecture without creating a big wall of text.

Quote:Wait....why am I posting in the support subforum? I'm not supposed to be here. Damn you kinkin!

I miss reading your extremely thorough explanations of why people's computers were too slow for Dolphin.

ay, you succeed, I be damned

Anyways, I find the graph very interesting. Assuming the data isn't more interpolated than my early attempts at mixing, before I knew proper noise control, my CPU's actually faster per-clock than a piledriver CPU (in dolphin, at least). This information could prove useful, sort of.