However in order to help me better understand where you are coming from when you say my statement that "Mame is slower" is incorrect, are you saying that the video and sound revisions only affected a few games and did not affect most games across the board? I am not being sarcastic; I do not know.
No, I am not saying that the revisions affected only a few games. What I am saying is that some games were affected (slowed) more by the changes, some not as much, and a few might even run faster after them. Also in many cases specific games (such as Cruisn' series) will run faster on later versions than previous ones.
You have removal of hacks which slows down performance, and you have optimization of code and drivers which speeds it up. And this is happening for all drivers at all times. So it is very hard to say MAMEDev was making MAME faster until 0.76 (just an arbitrary number) and then they started slowing it down. Some games will run faster in older versions, some will run faster in new versions. The same debate used to go on with MAME32 and DosMame. Overall, DosMAME was generally faster (not MAMEw, I am talking early days), but there were a handful of games that ran faster on MAME32, for whatever reason.
I thought even the devs will tell you that the corrections that were needed to the code slowed it down somewhat ON THE SAME HARDWARE.
What the devs generally refer to this as is "moving the sweet spot". In other words, prior to MAME 0.53 or so, you could run Pac-Man and most 80's games on something similar to a Pentium 200 Mhz. Later builds of MAME optimized the core for performance on 90's era games, meaning your Pentium 200Mhz can't play PacMan anymore (on a recent build), but nobody is building a MAME cab these days with a P 200 so they can run MAME 0.114u1 with full artwork.
I think we apparently have two groups of people that get lumped into one group. Group A that ignorantly complain that the devs are screwing up Mame by making it slower. Group B that just want to run the more recent games at an acceptable frame rate by running an older version of Mame on their older hardware.
What happens is Group A is a fairly vocal group. As pointed out, MAME is not attempting to be a fast emulator. For example, it typically emulates all the video code in the CPU before sending the output to the card. So your 512M X1900 graphics card sits around until your 2.0 Ghz CPU finishes the processing of the emulated game GPU and the emulated Game GPU and then gets it's output to run. It's the most accurate way to do it. There are faster emulators out there.
Though you are not one, there are some that get highly offended by the discussion of speed. They start screaming to GET A NEW COMPUTER OR DON'T UPGRADE. Well that's what I'm trying to do by going to a past revision. Its very irritating to see the rants on a discussion about which past Mame version to use.
Well - here's the practical advice section of my post: Note that except for having approximately a 2.0 Ghz processor, I don't play a lot of the same games as you do (I don't think). Note also, that I don't think there is a blanket answer. (Run MAME 0.76 for all games is probably not good advice.)
First, if you like the new bezel artwork, don't go below 0.107.
Second, note what games do not run full-speed on the current 0.111 version that you are using. Note: full-speed is all you are concerned about. If the game runs at 101% unthrottled (F10) throughout the game, then the fact that it could run at 150% in a 30 revision ago version of MAME has little or no bearing.
Third for these games, look into other emulators which are not as "accurate" as MAME and therefore faster. MAWS usually lists these. Likely ones are Nebula, Final Burn, Zinc, HazeMD, CPS-shock, Retrocade, RAINE, etc. Try these for the games that run too slowly.
Finally, in general, MAME has had a few revisions that were significantly slower than others, and generally incremental slowdowns in between. Generally, you don't need to download and test all fifty versions between 0.53 and 0.113, b/c 0.78 is going to run identically to 0.77. Get copies of R37B13, 0.53, 0.63, and 0.85, 0.106 and 0.111 and try the games on each one and find which works best for you.
Really finally - if none of the versions works well, look at either MAWS or the current driver info in the .src and see what was changed. For example, let's look at Cruis'n USA:
Emuloader pulls the driver information for me, I am not sure where it gets it, but from MAWS:
- 0.74u2: Brake pedal hack removed.
- 0.74: Changed ADSP2105 clock speed to 10MHz.
- 0.62: Added Cruis'n USA (rev L4.1) and clones Cruis'n USA (rev L2.1) and (rev L4.0).However, from the Driver Information:
- 0.99u7: Aaron Giles fixed crashing of the Midway V-Unit games.
- 0.94u5: Aaron Giles turned off debugging code in the hotspots, giving a huge speed improvement to the V-unit games.
- 0.84u3: Fixed sound1 roms addresses.
- 0.79u2: Changed Custom sound to DMA-driven_DAC.
- 23rd January 2004: Aaron Giles cleaned up the Midway Y/Z/X/T/W/V-unit drivers.
- 0.77: Changed ADSP2105 CPU2 clock speed to 10 MHz.
- 0.76u2: Aaron Giles update the Midway V-Unit driver: WarGods sound is 100% now, added some pre-initialization to the WarGods NVRAM, added speedup handlers for all games (it makes a little difference), revamped the DCS2 handling to support stereo output, fixed a number of synchronization issues in the DCS2 handling and implemented a few missing features in the Midway I/O ASIC. Changed ADSP2105 CPU2 clock speed to 10240000 Hz.
- 0.63: Changed wmsvunit.c driver to midvunit.cSo this game should be best in either 0.95 or 0.100, although 0.106 or 0.111 shouldn't be too much slower.
NOTE: MAMEdev also doesn't do you many favors in that "Huge speed improvement to the V-unit games" doesn't correspond to the casual gamer as "Cruis'n USA is faster now (or Mortal Kombat, NBA Jam, Smash TV), etc."
If they can read mame.net they should be able to see that the devs can take care of themselves and do not need policemen on this board.
Ermmmn, yeah, that's a bit of an understatement!!!
