File with CPU requirements?

[jelwell]

Are there CPU requirements for roms somewhere?

Like a Catver.ini or history.dat that lists the recommended minimum x86 processor speed.

A file like this could be used in ClrMamePro to delete games that require far more processor than available or a frontend like MAMEWAH to hide those games.

It would be good to be able to tell MAMEWAH to hide all games with a CPU requirement greater than 2.5Ghz. Even other platforms could benefit because they could, in general, scale the suggested CPU by their architecture differences.
Joseph Elwell.

[paigeoliver]

The thing is that the CPU requirements change for every rom on every mame version, and they are not even the same on the different systems due to different Mame settings, motherboards, memory speed and such.

I have every Mame version from .53 to .92 installed on my laptop (in one folder, wh00t), and I can say for certain that the CPU requirements for each and every game go up and down with each version (usually up, but on some games they have gone down).

[jelwell]

The thing is that the CPU requirements change for every rom on every mame version, and they are not even the same on the different systems due to different Mame settings, motherboards, memory speed and such.

No one said it had to be 100% accurate. Just a recommended minimum.

Users of the frontend/clrmamepro/tools can decide if they want to filter and if so what value to filter by.

Your recent Mortal Kombat speed test post shows that although their are differences in MAME versions none of them dropped Mortal Kombat below a playable level on your machine. That's the sort of information that would be good to capture.

If it's too difficult to relate the "recommended minimum CPU" then how about some sort of rating system that conveys *exactly* the same information - but not in a CPU based counting system. Just drop the MHz/GHz from the rating system.

Like pacman = .3, mortal kombat = 2, Killer Instinct  = 3, Gauntlet Legends = 9.
Joseph Elwell.

[paigeoliver]

It is a decent idea, but man it would take a long time to implement, maybe if someone could right a benchmarking app/test file that could automat speed tests then it would be workable.

Basically the app would run "mameVERSION.exe game -nothrottle -ftr 10000" for each and every rom and then record results.

[jelwell]

It is a decent idea, but man it would take a long time to implement, maybe if someone could right a benchmarking app/test file that could automat speed tests then it would be workable.

Basically the app would run "mameVERSION.exe game -nothrottle -ftr 10000" for each and every rom and then record results.

Thanks! That's an awesome idea to start with. And it's fully automatable - assuming mame doesn't leak.

I need like 60fps minimum to play a game?
Joseph Elwell.

[mahuti]

We actually had something like this on the Mac side that benchmarked the FPS on a selection of standard roms that we tried out, and the results varied from computer to computer on the exact same configuration.  Could have  been due to background processes, videocard, etc. Basically MAME replayed a game recording and once it was done we took down the average FPS, and a whole bunch of other information about the computers, with detailed instructions on the setup, etc. Didn't do much for us. Still lots of confusion about configuration & the best version to use with x_rom.

[jelwell]

As long as everyone understands that the level of detail is meant to provide a SCALE and not a measurement I think we're good. I'll start working on a batch file to automate this process.
Joseph Elwell.

[paigeoliver]

It is a decent idea, but man it would take a long time to implement, maybe if someone could right a benchmarking app/test file that could automat speed tests then it would be workable.

Basically the app would run "mameVERSION.exe game -nothrottle -ftr 10000" for each and every rom and then record results.

Thanks! That's an awesome idea to start with. And it's fully automatable - assuming mame doesn't leak.

I need like 60fps minimum to play a game?
Joseph Elwell.

A few games are 53 and a few are 30 or 45 but like 98 percent of the games are 59-61.

Plus, if you can get it fully automated like that (might want to program the app to skip any ok screens, or press any button screens), the we can get a bunch of people to run the benchmarks on different systems.

One thing, to avoid situations where the results are radically different due to configs, there should be a rule requiring all tests to be done on a fresh unmodified config. Whatever mame spits out when you run mame.exe -cc should be the config used.

[jelwell]

A few games are 53 and a few are 30 or 45 but like 98 percent of the games are 59-61.

Plus, if you can get it fully automated like that (might want to program the app to skip any ok screens, or press any button screens), the we can get a bunch of people to run the benchmarks on different systems.

One thing, to avoid situations where the results are radically different due to configs, there should be a rule requiring all tests to be done on a fresh unmodified config. Whatever mame spits out when you run mame.exe -cc should be the config used.

I'll add alan's tweaks to avoid all the ok screens (they're already added to my custom mame build). One step at a time with the config issues.
Joseph Elwell.

[krick]

I'll add alan's tweaks to avoid all the ok screens (they're already added to my custom mame build). One step at a time with the config issues.

Are you using the standard MAME build environment?  Just checking.  Using different compilers and/or build settings can make a HUGE difference in framerate.