John IV |
IV/Play, MAME, MAMEUI
|
|
|
Reged: 09/22/03
|
Posts: 1970
|
Loc: Washington, USA
|
|
Send PM
|
|
|
Posted a different type of benchmark.
#390560 - 04/04/21 06:31 AM
|
|
|
After adding a bunch of historical benchmarks to the page I figured I'd do a run on the nightly builds in the old style (pre -bench) of actually outputting video and sound.
So these are what you can expect from games running full tilt with video and sound and default mame.ini settings for 90 seconds:
mame gamename -str 90 -nothrottle -norc
It's kind of interesting.
I also added a .229 run from my venerable 2008 Core 2 Duo e8400 Wolfdale.
http://www.mameui.info/Bench.htm
|
john iv
http://www.mameui.info/
|
|
|
Re: Posted a different type of benchmark.
[Re: John IV]
#390568 - 04/04/21 06:41 PM
|
|
|
Thanks John!
Its cool to see actual usable numbers.
|
Only here to annoy...
|
|
TrevEB |
Brokering peace between the collectors and the Mame Community
|
|
|
Reged: 03/11/05
|
Posts: 770
|
Loc: Oakland, CA
|
|
Send PM
|
|
|
Re: Posted a different type of benchmark.
[Re: John IV]
#392361 - 10/31/21 12:10 PM
|
|
|
John and or community,
I am really looking forward to seeing benchmarks using the M1X, M1X Pro and now intel’s Alder Lake CPU.
To those having the cash to drop for these awesome new macbooks, please do share a Mame benchmark score for the spreadsheet.
|
|
|
|
Re: Posted a different type of benchmark.
[Re: John IV]
#392372 - 11/01/21 12:36 PM
|
|
|
Thank you for sharing these numbers, John.
Could you also make a benchmark with different shader systems, but not using shaders i.e. just Open GL, bgfx, d3d ? This would be interesting.
|
|
|
John IV |
IV/Play, MAME, MAMEUI
|
|
|
Reged: 09/22/03
|
Posts: 1970
|
Loc: Washington, USA
|
|
Send PM
|
|
|
Re: Posted a different type of benchmark.
[Re: uman]
#392375 - 11/02/21 02:40 AM
|
|
|
Here are the numbers for a run of the different video options on my box, BGFX 'won': mame gamex -str 90 -nothrottle -norc -video X
AMD 6GB 5600XT / Win10 x64 Pro
This is from this morning's tree: 0.237 (mame0237-80-gc7ef3525326)
Code:
OpenGL BGFX D3D GDI blitz 299 304 305 193 crusnusa 300 315 314 221 cubeqst 280 339 325 91 cyvern 1010 1145 1093 352 dkong 2142 2983 2673 400 drivedge 337 350 349 182 gauntleg 595 615 612 369 gnbarich 1426 1665 1570 333 gradius4 413 425 423 215 gtfore06 209 215 214 171 harddriv 566 657 630 239 kidniki 431 451 450 232 kof98 1084 1245 1175 311 mario 915 1063 1002 286 mk4 269 282 274 166 pacman 3536 5565 4988 451 pinkswts 1325 1590 1540 325 pongf 449 462 464 236 propcycl 249 247 255 162 radikalb 234 240 237 154 roadblst 889 994 983 249 robotron 3193 4469 3837 381 rvschool 346 373 371 197 scud 88 89 89 73 sf2 1542 2035 1841 321 sfa2 1355 1718 1566 314 sfiii 1081 1243 1174 310 slrasslt 458 442 477 225 spacfury 179 181 180 159 starblad 187 197 193 133 starsldr 66 65 66 57 sushibar 86 86 86 73 tekken 508 539 543 231 tekken3 234 247 247 157 vf2 171 174 173 126 vfkids 135 136 137 105 26586 33146 30856 8200 738 921 857 228 -20% Base -7% -75%
|
john iv
http://www.mameui.info/
|
|
|
Re: Posted a different type of benchmark.
[Re: John IV]
#392388 - 11/02/21 08:48 PM
|
|
|
> Here are the numbers for a run of the different video options on my box, BGFX 'won'
Perhaps this is a stupid/silly question, but I get the impression that 'BGFX' is not 'really' a renderer in and of itself, but rather a (backend agnostic) 'frontend/interface/api' to something like d3d9, d3d11, d3d12, opengl, metal, or vulkan ? So if this is sort of correct, I'm wondering what BGFX is/was using in these tests ? Sorry if I got this all wrong.
--EDIT--
Just for reference, on my Windows 10 system (mame 0.231) :
Code:
mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend d3d9 Average speed: 237.28% (89 seconds)
mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend d3d11 Average speed: 3256.44% (89 seconds)
mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend d3d12 Average speed: 1443.43% (89 seconds)
mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend opengl Average speed: 2951.24% (89 seconds)
mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend vulkan Average speed: 1176.73% (89 seconds)
mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend metal Average speed: 3246.56% (89 seconds)
--EDIT2-- And what really surprises me here is that the "-bgfx_backend metal" doesn't throw an error on MS-Windows, as I thought it was a Mac only thing ? Oh, well.
Edited by mhoes (11/02/21 09:33 PM)
|
|
|
John IV |
IV/Play, MAME, MAMEUI
|
|
|
Reged: 09/22/03
|
Posts: 1970
|
Loc: Washington, USA
|
|
Send PM
|
|
|
Re: Posted a different type of benchmark.
[Re: mhoes]
#392391 - 11/02/21 09:58 PM
|
|
|
On my bgfx runs, since I'm using the -norc it's defaulting to whatever is built into MAME, probably the D3D11 backend(?). My results are similar to yours on pacman when forcing the backend:
Code:
c:\O\Games\MAME>mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend d3d9 Average speed: 237.69% (89 seconds)
c:\O\Games\MAME>mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend d3d11 Average speed: 5537.42% (89 seconds)
c:\O\Games\MAME>mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend d3d12 Average speed: 858.55% (89 seconds)
c:\O\Games\MAME>mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend opengl Average speed: 3017.54% (89 seconds)
c:\O\Games\MAME>mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend vulkan Average speed: 906.69% (89 seconds)
c:\O\Games\MAME>mame pacman -str 90 -nothrottle -norc -video bgfx -bgfx_backend metal Average speed: 5513.11% (89 seconds)
|
john iv
http://www.mameui.info/
|
|
|
Re: Posted a different type of benchmark.
[Re: John IV]
#392399 - 11/03/21 05:11 AM
|
|
|
I've noticed that the "main" benchmark says "4.5Ghz" in the columns and then "4.0GHz" outside, is the tested processor overclocked or just an oversight?
Also, would an i7-6700 (without the "K") have the same results?
|
|
|
John IV |
IV/Play, MAME, MAMEUI
|
|
|
Reged: 09/22/03
|
Posts: 1970
|
Loc: Washington, USA
|
|
Send PM
|
|
|
Re: Posted a different type of benchmark.
[Re: ICEknight]
#392401 - 11/03/21 05:18 AM
|
|
|
|
kirby |
Apple MacBook Pro M1 & Nintendo addict
|
|
|
Reged: 11/03/21
|
Posts: 58
|
Loc: Spain
|
|
Send PM
|
|
|
Re: Posted a different type of benchmark.
[Re: TrevEB]
#392434 - 11/05/21 03:00 AM
|
|
|
Hi
I bought a Macbook Pro M1 MAX with 32 GB RAM recently! The box says it has 10 cores (if I'm correct). I read this post and I'm happy to share some tests but ... I'm a complete beginner and I just use the M1 build of mame and some frontends (all graphical). I never touched any Terminal...
If you explain me step by step what to do on my Mac, I can try (no guarantee I can do this! :-D)
Valentina
|
|
|
TrevEB |
Brokering peace between the collectors and the Mame Community
|
|
|
Reged: 03/11/05
|
Posts: 770
|
Loc: Oakland, CA
|
|
Send PM
|
|
|
Re: Posted a different type of benchmark.
[Re: John IV]
#392440 - 11/05/21 07:31 AM
|
|
|
Hi John,
When you run the benchmark, do you have a bat file that runs all 36 titles and spits out the results in this text file, or are you running the tests manually one by one.
If such a batch file exists, please share, especially for Kirby.
And now apparently intel has entered the ring. Maybe I can get Linus Tech to run a test for us, or not.
TrevEB
|
|
|
TrevEB |
Brokering peace between the collectors and the Mame Community
|
|
|
Reged: 03/11/05
|
Posts: 770
|
Loc: Oakland, CA
|
|
Send PM
|
|
|
Re: Posted a different type of benchmark.
[Re: kirby]
#392441 - 11/05/21 07:42 AM
|
|
|
It has been a long long time since I set up Mame on a macbook, however I know if you manage to figure out how to set things up then you should be able to make a benchmark test.
mame spacfury -str 90 -nothrottle -norc
mame mk4 -str 90 -nothrottle -norc
and so on. Just a couple would be a good start to see how the M1 compares with the current spreadsheet. Just stick to the titles on the list. The more the better.
TrevEB
|
|
|
kirby |
Apple MacBook Pro M1 & Nintendo addict
|
|
|
Reged: 11/03/21
|
Posts: 58
|
Loc: Spain
|
|
Send PM
|
|
|
Re: Posted a different type of benchmark.
[Re: TrevEB]
#392447 - 11/05/21 06:31 PM
|
|
|
Sorry, I can't make it work. I just use a MAME frontend but have no further knowledge
|
|
|