Dolphin, the GameCube and Wii emulator - Forums

You can always use the Unofficial Dolphin Benchmark, but did you know there's another way to benchmark the ability of your PC's CPU to run Dolphin smoothly?

If you have DKCR, you can do a "real-world" CPU benchmark.

The DKCR demo FMV (that plays after you wait a while at the title screen) is a perfect benchmark for your CPU:
- it's very demanding (puts a lot of stress on the CPU)
- doesn't stress the GPU (at all)
- both backends give the exact same performance, so you can use OpenGL with AMD or Intel GPUs as well
- it's deterministic (the framerate is consistent during the whole run)
- you get the results much faster than by running the Dolphin benchmark

Download Dolphin 4.0-8074 dev. build from the official page, extract the .7z archive to a new folder, create an empty file named portable.txt in the same folder, start Dolphin and change these settings:

- Set the frame limiter to "OFF"
- Turn off "Scaled EFB copy"
- Enable "Borderless Fullscreen"

Leave everything else at default (e.g. OpenGL backend, 1xIR, Auto aspect ratio, no AA, 1xAF, etc.)

Start DKCR and when the title screen appears, press the connect/disconnect hotkey once (puts the input thread to sleep) and then wait for the first FMV to play.

Scoring
---------
* The mimimum spec PC capable of running Dolphin smoothly should get 61 FPS or higher in that demo FMV (video).
* For the demanding titles (e.g. TLS), you should have a CPU that scores 91 FPS or better.
* For running the most demanding titles (e.g. SWRL) at full speed, you need a high-end CPU that scores an impressive 145 (!) FPS or better.

Even better, the scores obtained by running this "benchmark" are proportional to the scores obtained with the Unofficial Dolphin Benchmark:
- A CPU that finished the UO Dolphin Benchmark in 9 minutes (9:00) should get about 60fps in this bench.
- A CPU that finished the UO Bench in 6 minutes should get about 90fps in this one.

This is probably technically better than the old Wind Waker benchmark, except that it's a less common game. However the official benchmark has all the timing and stuff built-in, so is more robust.

"There are 3 different demo FMVs (all of them put the exact same stress on the CPU)"

False. I get different fps in each FMV in my testing.

This is not a good benchmark test because you have to wait for the loop after all 3 FMVs finished running.

Still trying to figure how the average FPS you'll get while watching a FMV (which, on console side, it's just a video player streaming the file from the disc and almost zero GPU activity) would compare to "real-world" performance, where games actually render its graphics and put stress on the CPU with its engine, calculations, etc. Said that, the only "real-world" performance benchmark is the Zelda Wind Waker one, and this "average FMV FPS" shouldn't even be considered a benchmark...

(10-31-2015, 02:01 PM)tuanming Wrote: [ -> ]False. I get different fps in each FMV in my testing.

The first one is the one you should use as a benchmark. It's a bit more demanding than the other two.

(10-31-2015, 02:01 PM)tuanming Wrote: [ -> ]This is not a good benchmark test because you have to wait for the loop after all 3 FMVs finished running.

No, you don't. Just press the connect/disconnect hotkey once at the title screen and wait for the FMV to play.

(11-01-2015, 02:07 AM)Jhonn Wrote: [ -> ]Still trying to figure how the average FPS you'll get while watching a FMV would compare to "real-world" performance...

Because playing back this FMV is more demanding than anything else rendered in real-time (for some odd reason). It's a good CPU benchmark.

(11-06-2015, 12:24 AM)kirbypuff Wrote: [ -> ]Because playing back this FMV is more demanding than anything else rendered in real-time (for some odd reason). It's a good CPU benchmark.

Source, please? Where did you get that a FMV would require more CPU power than "anything else rendered in real-time"?