|
MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
#388710 - 11/21/20 10:34 AM
|
|
|
I finally got a chance to run some quick MAME benchmarks on a ZEN 3 CPU today, it was a stock Ryzen 7 5800X, cooled by a 360mm AIO, matched with 32GB (4x8GB) DDR4-3600 CL18 fclk was set to 1800Mhz.
System was running Windows 10 64bit version 2004. and stock MAME64 v0.226.
Here are the results compared with John IV scores.
i7-6700K R7-5800X Gain
blitz 260 326 25.38%
crusnusa 315 372 18.10%
cubeqst 360 465 29.17%
cyvern 1216 1388 14.14%
dkong 4377 5182 18.39%
drivedge 350 434 24.00%
gauntleg 488 500 2.46%
gnbarich 2034 2327 14.41%
gradius4 374 -
gtfore06 211 -
harddriv 682 823 20.67%
kidniki 461 583 26.46%
kof98 1322 1725 30.48%
mario 1091 1270 16.41%
mk4 273 318 16.48%
pacman 12153 14678 20.78%
pinkswts 1669 2326 39.36%
pongf 504 589 16.87%
propcycl 284 372 30.99%
radikalb 251 317 26.29%
roadblst 1202 1495 24.38%
robotron 6426 7146 11.20%
rvschool 342 429 25.44%
scud 81 115 41.98%
sf2 2644 3450 30.48%
sfa2 1834 2290 24.86%
sfiii 1382 1753 26.85%
slrasslt 549 1167 112.57%
spacfury 188 237 26.06%
starblad 193 227 17.62%
starsldr 68 74 8.82%
sushibar 88 114 29.55%
tekken 588 631 7.31%
tekken3 249 318 27.71%
vf2 173 231 33.53%
vfkids 141 178 26.24%
I missed a couple of games, but the numbers look good, more performance can probably be had with faster memory with lower latency, and some CPU tweaking or an overclock.
But it now possible to get good performance in MAME on an AMD CPU.
|
If all else fails, Burn the manual.
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388720 - 11/22/20 12:28 AM
|
|
|
You're comparing a intel CPU that's over 5 years old to AMD's latest offering. I'd certainly hope that AMD would be good enough by now
|
Windows 11 64 bit OS
Intel Core i7-10700
Nvidia GeForce RTX 2060 6GB
32GB DDR4 RAM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: Envisaged0ne]
#388722 - 11/22/20 03:19 AM
|
|
|
I used John IV benchmarks as a comparison as they are the best known and documented scores.
While MAME scales almost one to one with clock speed, It gains almost nothing from high core count CPUs, as it simply doesn't scale beyond few cores.
Intel's IPC has only improved by about 10% in the last half decade, most of their resent performance improvements have come from higher base & boost clocks, and adding more cores.
I benchmarked a 9900ks @5.1ghz back in late 2019 and it only managed a 25% uplift over johns IV i7 6700K 4.5ghz scores and that's with a 13% clock speed advantage.
There was zero IPC gain from the 9900k to the 10700k only a small clock speed bump.
I had planned to run more benchmarks next week, including on a 10th gen Intel CPU to give a better comparison.
But I'm sure this will show that ZEN 3 performance is now comparable with Intel CPU performance in MAME, something that has not been true since I started benchmarking MAME back in 2007.
Until now AMD's IPC and clock speed deficits meant it just couldn't compete with Intel. The ZEN 2 CPU closed the gap a lot, but their ccx and cache layout made for inconsistence result in MAME, and it's IPC performance couldn't overcome the clock speed deficit when compared to Intel.
This was a quick and dirty benchmark run, it was just a system sitting on the test bench at work that I managed to spend 20mins testing. It's a sub optimal setup but the number were still very good.
|
If all else fails, Burn the manual.
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388727 - 11/22/20 03:52 PM
|
|
|
> slrasslt 549 1167 112.57%
That's some insane gains 
|
|
|
| TrevEB |
|
Brokering peace between the collectors and the Mame Community
|
|
|
|
|
Reged: 03/11/05
|
|
Posts: 770
|
|
Loc: Oakland, CA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388732 - 11/22/20 09:52 PM
|
|
|
Thanks for doing this!
I had recently talked with John about providing a package to Linus Tech Tips so as to have him run Mame benchmarks as part of his testing.
Hoorah for AMD!
Way to put pressure on intel.
Now we have the Apple M1 as well forcing them to get stop stalling and innovate.
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: Ville Linde]
#388737 - 11/23/20 12:45 AM
|
|
|
Unfortunately look like that rom errored out at self test and halted resulting in crazy high score, It was a rushed benchmark run and I was using v0.225 romset with MAME 0.226 exe. Normally I'd run all the roms first to make sure there were no errors and that they entered game, but I didn't have time so I skipped that step.
I ran slrasslt on my machine and the rom halts at a prom config error, so I'll have to solve that before I try another benchmark run.
I'm planned a long overdue upgrade to my PC and wanted some MAME benchmark number before I committed to ZEN 3, I flung the scores up here as I figured people would be interested, as I couldn't find any ZEN 3 scores for MAME yet.
|
If all else fails, Burn the manual.
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388742 - 11/23/20 02:00 AM
|
|
|
Nice, good to see you again Taz-nz. I'll add your results once you've got the time to do a full run.
So this was a 5800x at 4.7Ghz turbo-ed?
|
john iv
http://www.mameui.info/
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388744 - 11/23/20 02:23 AM
|
|
|
Heh, was gonna ask. Seemed like such an anomaly that it had to be a broken run. Will be interesting to see if Intel can even match these numbers with Desktop-Whatever-Lake.
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388749 - 11/23/20 08:43 AM
|
|
|
No problem , I hopefully I get a bit more time tomorrow to do a better run. (but it depends how my day goes)
Yeah stock 5800x boosting to 4.7ghz, no overclock only tweak was 1800 fclk because of DDR4-3600 memory.
If you can link me a compiled nighty build, I'd like to have a play with the prelease 0.227 improvements.
|
If all else fails, Burn the manual.
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: Ville Linde]
#388751 - 11/23/20 08:59 AM
|
|
|
I've been referring to as lake bottom, because it feel like Intel is sinking fast. Intel is so desperate they are back porting a 10nm design to 14nm++++++++++++++++++, and to make it work they are dropping two cores. I can only imagine how much power it's going to suck down.
But at least we will actually get a new architecture this time, and not just tweak or two with more cores and a clock speed bump like we got with the last three generations.
I guess it will come down to if their idea of double digit IPS improvement means barely 10% or something impressive like 20%.
|
If all else fails, Burn the manual.
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: TrevEB]
#388752 - 11/23/20 09:19 AM
|
|
|
I wish one of the YouTube reviewers would use MAME but because it constantly changing, and you need to individual benchmark dozens of ROMs to get a real performance picture, I don't think it's practical for them to do it.
It's great to see competition in the CPU & GPU market again, the CPU market in particular had stagnated under Intel's dominance. I'm actual excited to upgrade my PC for the first time in a long while.
The few benchmark Balrog post for the Apple M1 look really impressive, I hope he runs a full suite for benchmarks to give a fuller picture. https://forums.bannister.org/ubbthreads.php?ubb=showflat&Number=118135#Post118135
I've been looking around for a while at what other CPU architecture might challenge x86 in the near future, because it seemed like some ROMs would never be playable at full speed in MAME due to the lack of x86 IPC gains. I hope we get more open high performance ARM platform in the near future.
|
If all else fails, Burn the manual.
|
|
Stiletto |
|
They're always after me Lucky ROMS!
|
|
|
|
|
|
Reged: 03/07/04
|
|
Posts: 6472
|
|
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388753 - 11/23/20 09:41 AM
|
|
|
For shits and giggles, taz-nz, you may love a look at zino's run of benchmarks on the Raspberry Pi 4.
He had an automated testing suite setup to not only map to John_IV's benchmarks but also ChoccyHobNob's older run of Raspberry Pi 0, 2 and 3 benchmarks. It's pretty sweet, to be honest. (John IV is very close, and good enough for me, with his numbers and his heat map - he just needs to add a line graph to completely match that  )
Unfortunately zino has no benchmark run newer than MAME 0.215 from October. zino also seems to have disappeared from our radars since early February, but hopefully we will see him pop back up again.
If you could also run a test with MAME 0.215, we could see just how badly AMD Zen 3 crushes Raspberry Pi 4 (although those are vastly different product markets).
http://mame-test.lysator.liu.se/bench/
- Stiletto
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: Stiletto]
#388754 - 11/23/20 12:40 PM
|
|
|
I've access to Raspberry Pi 4's at work so if you can point me to a pre-compiled version of the latest MAME for the raspberry pi, and usable set of instructions to set up, I can run an up-to-date benchmark on one, setting up the raspberry pi and overclocking shouldn't be an issues, but Linux things tend result in me swearing a lot, and needing a stiff drink. So the simpler it is for me get up and running the better.
I can run a 0.215 run on the AMD, but seems like it makes more sense to update the raspberry pi 4 scores if possible instead of messing with old rom sets.
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388771 - 11/24/20 02:43 AM
|
|
|
It would be great to see how MAME compares in a side-by-side benchmark on an i9-10900K and a R9 5950X (the current leaders in single-core perf from the 2 companies)
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: jonwil]
#388774 - 11/24/20 10:26 AM
|
|
|
You almost got your wish, had planned to finish up 10900K build and build a 5950X system today, got both system completed, unfortunately the 5950X has a weird fault where in goes into a halt state at random and then recovers, so 5950X benchmarks will have a wait a day out two while diagnose and solve issue.
But here are the i9 10900K scores, stock other than multicore enhancement enabled, 32GB DDR4-3600, it was under an air cooler due to 4U chassis limitation, so not ideal, but it never really got hot due to the short amount of time each ROM takes to benchmark.
ROM i7 6700K R7 5800X i9 10900K
blitz 260 326 308
crusnusa 315 372 423
cubeqst 360 465 427
cyvern 1216 1388 1402
dkong 4377 5182 5377
drivedge 350 434 436
gauntleg 488 500 484
gnbarich 2034 2327 2409
gradius4 374 461
gtfore06 211 238
harddriv 682 823 799
kidniki 461 583 554
kof98 1322 1725 1582
mario 1091 1270 1322
mk4 273 318 359
pacman 12153 14678 14967
pinkswts 1669 2326 2152
pongf 504 589 641
propcycl 284 372 343
radikalb 251 317 341
roadblst 1202 1495 1412
robotron 6426 7146 7749
rvschool 342 429 395
scud 81 115 98
sf2 2644 3450 3128
sfa2 1834 2290 2200
sfiii 1382 1753 1387
slrasslt 549 1167 814
spacfury 188 237 212
starblad 193 227 235
starsldr 68 74 82
sushibar 88 114 104
tekken 588 631 704
tekken3 249 318 346
vf2 173 231 214
vfkids 141 178 167
I not sure how well the 5950X will do due to the two separate chiplets, I tried benchmark the R9 3950X last year and the Windows scheduler kept bouncing MAME from core to core like a hot potato, and MAME really didn't like the 4 separate CCXes with separate caches, was actually faster with 3 out of 4 CCXes disabled.
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388795 - 11/25/20 07:48 PM
|
|
|
Thanks, new bench page is up w/ these results.
Pretty close, the last two scores will close the gap further if they're comparable.
http://www.mameui.info/Bench.htm
Edit: Presumably the 10900K was hitting 5.3Ghz?
Edit2: And we still have the bad solar assault score on the 5800X?
Edited by John IV (11/25/20 08:47 PM)
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388799 - 11/26/20 03:29 AM
|
|
|
Balrog sent me some results from his Apple M1 tests which I incorporated.
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388801 - 11/26/20 05:42 PM
|
|
|
Not looking to get included in the list, but just because I was curious myself I did a comparison with my Intel Core i7-10700 (Base 2.90 GHz, Boost 4.80 GHz) on Windows 10 64-bit and MAME 0.226. Results listed below. Not as good as the AMD, but at least comparable to the older Intel.
blitz 281.27
crusnusa 365.93
cubeqst 369.39
cyvern 1240.01
dkong 4522.17
drivedge 376.53
gauntleg 455.32
gnbarich 2121.77
gradius4 411.45
gtfore06 201.00
harddriv 708.83
kidniki 489.15
kof98 1353.37
mario 1115.13
mk4 316.00
pacman 12914.60
pinkswts 1781.98
pongf 560.80
propcycl 298.25
radikalb 290.23
roadblst 1226.66
robotron 5985.05
rvschool 346.58
scud 90.21
sf2 2693.59
sfa2 1886.22
sfiii 1422.41
slrasslt 567.47
spacfury 189.32
starblad 200.03
starsldr 71.56
sushibar 90.97
tekken 622.96
tekken3 307.38
vf2 183.19
vfkids 142.83
Edited by mhoes (11/27/20 05:16 PM)
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: mhoes]
#388803 - 11/26/20 09:14 PM
|
|
|
Thanks, I think you need to configure scud race though, it's like it was run at the service menu.
And this is the 10700 not 10700K?
Edited by John IV (11/26/20 09:20 PM)
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388805 - 11/27/20 04:37 PM
|
|
|
You were right about scud: I hadn't set it up properly, so it was indeed stuck in the service menu. The real value for scud is 90.21, and not the 178.11 I reported before. I corrected it in my previous post, thanks for pointing it out. Yes, this is the i7-10700, and not the i7-10700K.
-- edit --
Because I now got worried there might be more issues (not correctly set up games), I verified if everything ran well without '-bench'. Turns out I had to do some setup for drivedge and sfiii. I re-ran the benchmark for those games, and edited the previous post to reflect the changes. Sorry for the confusion; I should have verified things to begin with instead of blindly assUming that all would be well.
Edited by mhoes (11/27/20 05:23 PM)
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: mhoes]
#388810 - 11/28/20 03:28 AM
|
|
|
Cool, updated the page w/ your additions.
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388815 - 11/28/20 06:05 PM
|
|
|
Ok. I assume that the i7-10700/i9-10900k/r7 5800rx/macm1 results (heat map, percentages) to the far right/top in the benchmark results are in comparison to the i7-6700k ? Good to see that overall my CPU is at least a little faster than the i7-6700k  .
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: mhoes]
#388816 - 11/28/20 08:30 PM
|
|
|
I may as well test some games on my 4.0 GHz i5 8600k to compare with all the other benchmarks:
dkong 3939
radikalb 216
slrasslt 501
starblad 170
tekken3 225
vf2 151
I also highly suggest to add these games to the benchmark list.
cartfury 165
roadburn 137
sf2049 152
thegrid 232
warfa 163
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: Hydreigon]
#388817 - 11/28/20 10:09 PM
|
|
|
> I also highly suggest to add these games to the benchmark list.
>
> cartfury 165
> roadburn 137
> sf2049 152
> thegrid 232
> warfa 163
the benchmarks are meant to show a good spread across MAME, the ones you suggest are all basically the same rough hardware setups aside from thegrid (which is already represented with another game, mk4) and therefore provide little extra information.
you can even see from the numbers you provide there isn't much difference between the ones that are all similar setups; once you know one, you have a good idea for the others.
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: mhoes]
#388818 - 11/28/20 10:40 PM
|
|
|
The new entries on the right have their own heatmap between them, separate from my 6700K. The pinkish percentages at the bottom of the right section though are comparisons to the 6700K. So your 10700 is 4% faster at 4.8Ghz.
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: Haze]
#388820 - 11/28/20 10:44 PM
|
|
|
Haze is correct. I used gauntleg for the vegas driver back in the day as it was more mature; it encompasses your comparable performing vegas game suggestions.
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388834 - 11/29/20 02:53 PM
|
|
|
Because of this thread I recently actually wondered about this exact thing: how representative is this list for 'overall' (arcade) MAME performance ? I assumed that the list was some form of compromise, like getting a few different hardware types while not making the list insanely/unwieldy large at the same time ? Just curious.
Edited by mhoes (11/29/20 04:32 PM)
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: mhoes]
#388837 - 11/29/20 08:21 PM
|
|
|
Historically I've looked at interesting performance data points, some of the games included were PC 'killers' back in the day but have since been tamed. Some are personal favorites, others attempt to hit popular drivers like the neogeo, cps1-3 etc. Some are netlist sound related which can really impact PCs. Some are DRCs from RB and others' efforts, etc. But yes, also trying to keep it somewhat manageable if folks want to run their own benches in a decent amount of time without too much setup.
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388838 - 11/29/20 09:28 PM
|
|
|
Ah. So at least - and please correct me if I'm wrong here (which I am bound to be) - but at least *some* of the machines/drivers were selected because they, either historically or currently, were 'pain-points' for MAME, some way or the other. And then some more drivers because they represented 'popular' machines/hardware. And then, not to make the 'reproducibility factor' too large, not too many machines in total.
Makes perfect sense to me.
Thank you for the explanation and details.
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#388849 - 11/30/20 07:56 AM
|
|
|
Some new results from Balrog on the M1, this time against .226 proper.
|
|
|
|
Re: MAME v0.226 Benchmarks on Raspberry Pi 4 8GB @2.0ghz (is a total joke for MAME)
[Re: Stiletto]
#388852 - 11/30/20 12:59 PM
|
|
|
For some reason after a week where nothing worked the way it should, and my MAME drive had a heart attack, I decided compiling MAME on a raspberry pi 4 & benchmarking was a good way to spend my weekend.
It's wasn't, it was six hours trying to follow typical dumpster fire Linux instructions. With all the usual issues, instructions don't match current version, the first set of instruction only work if you follow a second set first, and those instructions fail to point out step 38/40 has a software dependency, and there are no clear instruction on how to install said additional package, but someone had the same issue with the second set of instructions, and wrote a half assed guide, that skip multiple steps and makes a couple of assumptions and has mistake just to drive it home.
But for some reason I stuck with it and got it to work in the end. It took another 6 hours to compile MAME, and then about the same again to run all the benchmarks, this should give you an idea how bad they are. I had issue with a couple of roms, but gave up trying to fix them.
blitz (locked pc)
crusnusa 90
cubeqst 84
cyvern 128
dkong 881
drivedge 95
gauntleg 27
gnbarich 40
gradius4 22
gtfore06 (errored out)
harddriv 156
kidniki 69
kof98 323
mario 213
mk4 73
pacman 3030
pinkswts 67
pongf 106
propcycl 55
radikalb 65
roadblst 247
robotron 1421
rvschool 73
scud 13
sf2 540
sfa2 412
sfiii 47
slrasslt 22
spacfury 33
starblad 47
starsldr 15
sushibar 5
tekken 100
tekken3 60
vf2 41
vfkids 13
So as you can see the performance is horrible, a raspberry pi is a dumb idea for MAME, Just buy an old i-series intel system from an e-waste company and get real performance.
|
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388853 - 11/30/20 01:06 PM
|
|
|
The M1 numbers are amazing, give it little time for the MAME port to mature, and a M1 Mac Mini will be the go to for a dedicate MAME machine. Just a pity it's such a locked down platform.
I'll try and get a complete run on a Ryzen 5000 system this week, without errors or missing ROMS, nothing when my way last week.
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Raspberry Pi 4 8GB @2.0ghz (is a total joke for MAME)
[Re: taz-nz]
#388857 - 11/30/20 11:17 PM
|
|
|
Thanks, added to page.
Left them out of the heat map as it obscured the other results somewhat.
Edited by John IV (11/30/20 11:17 PM)
|
|
|
|
Re: MAME v0.226 Benchmarks on Raspberry Pi 4 8GB @2.0ghz (is a total joke for MAME)
[Re: taz-nz]
#388860 - 12/01/20 02:53 AM
|
|
|
> tekken 100
> tekken3 60
100 fps in Tekken 1 on a Pi 4? My old POS Sempron 140 (with nForce 630a integrated graphics for maximum suckage) barely manages to sit on 60 fps and that's with MAME 0.152! Even with that version I still get the odd stutter in Tekken 1 and 2. Tekken 3 (in non-interlaced 240p mode instead of the default 480i) sits around 83% and Tekken Tag is slower still despite running on the same hardware and at the same resolution.
When a Pi starts trumping one of my PCs I think it's finally time for me to upgrade. That said, it was the weakest "modern" PC I had (by modern, I mean post-Pentium 4 era e.g. having PCI-Express instead of AGP, SATA drives instead of PATA/IDE, and no floppy drive or legacy ports).
|
|
|
|
Re: MAME v0.226 Benchmarks on Raspberry Pi 4 8GB @2.0ghz (is a total joke for MAME)
[Re: Heihachi_73]
#388867 - 12/01/20 09:52 AM
|
|
|
You need to remember MAME benchmark mode doesn't output video or audio, many of these ROMs ran a lot slower when run normally.
|
|
|
| TrevEB |
|
Brokering peace between the collectors and the Mame Community
|
|
|
|
|
|
Reged: 03/11/05
|
|
Posts: 770
|
|
Loc: Oakland, CA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: John IV]
#388870 - 12/01/20 07:42 PM
|
|
|
Re: the Apple M1 numbers
Is this compile of Mame running through rosetta?
As in does Mame on the M1 have to be translated to intel first?
Or is it possible to compile Mame to native M1
It is darn impressive!
Any thoughts as to why some titles could not post?
TrevEB
|
|
|
| Dullaron |
|
Diablo III - Dunard #1884
|
|
|
|
|
Reged: 07/22/05
|
|
Posts: 6125
|
|
Loc: Fort Worth, Tx
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Raspberry Pi 4 8GB @2.0ghz (is a total joke for MAME)
[Re: taz-nz]
#388871 - 12/01/20 07:45 PM
|
|
|
> For some reason after a week where nothing worked the way it should, and my MAME
> drive had a heart attack, I decided compiling MAME on a raspberry pi 4 & benchmarking
> was a good way to spend my weekend.
>
> It's wasn't, it was six hours trying to follow typical dumpster fire Linux
> instructions. With all the usual issues, instructions don't match current version,
> the first set of instruction only work if you follow a second set first, and those
> instructions fail to point out step 38/40 has a software dependency, and there are no
> clear instruction on how to install said additional package, but someone had the same
> issue with the second set of instructions, and wrote a half assed guide, that skip
> multiple steps and makes a couple of assumptions and has mistake just to drive it
> home.
>
> But for some reason I stuck with it and got it to work in the end. It took another 6
> hours to compile MAME, and then about the same again to run all the benchmarks, this
> should give you an idea how bad they are. I had issue with a couple of roms, but gave
> up trying to fix them.
>
>
> blitz (locked pc)
> crusnusa 90
> cubeqst 84
> cyvern 128
> dkong 881
> drivedge 95
> gauntleg 27
> gnbarich 40
> gradius4 22
> gtfore06 (errored out)
> harddriv 156
> kidniki 69
> kof98 323
> mario 213
> mk4 73
> pacman 3030
> pinkswts 67
> pongf 106
> propcycl 55
> radikalb 65
> roadblst 247
> robotron 1421
> rvschool 73
> scud 13
> sf2 540
> sfa2 412
> sfiii 47
> slrasslt 22
> spacfury 33
> starblad 47
> starsldr 15
> sushibar 5
> tekken 100
> tekken3 60
> vf2 41
> vfkids 13
>
>
> So as you can see the performance is horrible, a raspberry pi is a dumb idea for
> MAME, Just buy an old i-series intel system from an e-waste company and get real
> performance.
The Arcade1up using raspberry pi clone boards. Which explain why Galaga had issues on their v1 baords. Should be fixed on v2 as I heard. I don't know what are they using because I don't think it MAME. The games is change. 
You so right about MAME shouldn't be on the raspberry pi boards. Reason is people be sending false reports to MAMETester website.
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Raspberry Pi 4 8GB @2.0ghz (is a total joke for MAME)
[Re: Heihachi_73]
#388872 - 12/01/20 07:58 PM
|
|
|
Also reminder that these bench numbers are the percentage of full speed emulation, not fps. I'll add that bit to the bottom of the spreadsheet.
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: TrevEB]
#388873 - 12/01/20 07:58 PM
|
|
|
These numbers are from Balrog, you could ask in the shoutbox or on Bannister's forums.
|
|
|
| TrevEB |
|
Brokering peace between the collectors and the Mame Community
|
|
|
|
|
|
Reged: 03/11/05
|
|
Posts: 770
|
|
Loc: Oakland, CA
|
|
|
|
Send PM
|
|
|
Re: MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
[Re: taz-nz]
#390216 - 03/11/21 04:07 AM
|
|
|
Would you post the bat file used to run the benchmark.
I know how to do it but options change over time. Its important that we are comparing oranges to oranges.
Can you run the full list of specific titles or is it necessary to run one at a time?
This is assuming that each title has been run once so all necessary settings are in place.
Is the spreadsheet Paul uses just typical Excel or a Google spreadsheet?
Would be sweet if there was an automated system to just run the bench and spit out the spreadsheet.
Thanks
TrevEB
|
|
|
Index 
Threaded 
MAME v0.226 Benchmarks on Ryzen 7 5800X (AMD is no longer a joke for MAME)
Reply
[Re: 
)
.
