|
Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
#384423 - 11/28/19 03:15 AM
|
|
|
Was testing an overclock on PC at work, got bored and had latest version of MAME handy, figured I'd do a quick benchmark run.
Benchmarks were run with standard -bench 90 command
Intel i9-9900KS with a stable overclock of 5.1GHZ all cores at 1.35v no AVX offset, 32GB DDR4-3600 CL16.
Running clean install Windows 10 64bit v1909. Arcade MAME 64bit http://arcade.mameworld.info/
ROM Result
blitz 295.26%
crusnusa 459.71%
cubeqst 422.71%
cyvern 1366.88%
dkong 5112.04%
drivedge 407.41%
gauntleg 460.54%
gnbarich 2502.37%
gradius4 485.57%
gtfore06 217.53%
harddriv 727.99%
kidniki 381.05%
kof98 1522.45%
mario 934.89%
mk4 334.56%
pacman 15347.70%
pinkswts 2148.18%
pong 629.38%
propcycl 331.22%
radikalb 311.08%
roadblst 1280.97%
robotron 6845.35%
rvschool 343.26%
scud 105.20%
sf2 2999.55%
sfa2 2153.64%
sfiii 1366.62%
slrasslt 768.96%
starblad 221.39%
starsldr 80.36%
sushibar 105.27%
tekken 634.73%
tekken3 310.68%
vf2 212.53%
vfkids 129.37%
|
If all else fails, Burn the manual.
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: taz-nz]
#384432 - 11/28/19 10:08 AM
|
|
|
..Surprised you need 1.35vlts for 5.1Ghz on a supposedly super duper 'binned' KS.
(Have a 9900K myself (5Ghz @ 1.32vlts)
|
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: Ziggy100]
#384437 - 11/28/19 01:33 PM
|
|
|
Not my system, needs to be 24/7 stable under heavy loads, customer wanted it to be a fast as we could make it, while being stable at the same time.
Could have trimmed the voltage back more, but it may not have been 24/7 stable, better to have a little head room, temperatures are stable, voltage isn't to extreme. It's a balancing act.
If it was my personal system, I'd would have tweaked it more and probable pushed for more clock speed. But I also would have de-lidded it run full custom loop.
|
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: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: taz-nz]
#384471 - 11/29/19 09:15 PM
|
|
|
Good to see you checking in Taz-NZ. 
Added your run to the bench page.
Just as an interesting aside, I also re-did the performance heat map for the 2018 runs and you can see that something happened at .200. I don't remember what it was at the time where things took a dive performance wise.
http://www.mameui.info/Bench.htm
|
john iv
http://www.mameui.info/
|
|
Tafoid |
|
I keep on testing.. testing.. testing... into the future!
|
|
|
|
|
Reged: 04/19/06
|
|
Posts: 3135
|
|
Loc: USA
|
|
|
|
Send PM
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: John IV]
#384474 - 11/29/19 10:14 PM
|
|
|
> Good to see you checking in Taz-NZ.
> Added your run to the bench page.
>
> Just as an interesting aside, I also re-did the performance heat map for the 2018
> runs and you can see that something happened at .200. I don't remember what it was at
> the time where things took a dive performance wise.
>
> http://www.mameui.info/Bench.htm
The performance dive for most things came from the memory system rewrite.
|
|
|
| Stiletto |
|
They're always after me Lucky ROMS!
|
|
|
|
|
|
Reged: 03/07/04
|
|
Posts: 6472
|
|
|
|
|
|
Send PM
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: John IV]
#384475 - 11/29/19 10:55 PM
|
|
|
> Good to see you checking in Taz-NZ.
> Added your run to the bench page.
>
> Just as an interesting aside, I also re-did the performance heat map for the 2018
> runs and you can see that something happened at .200. I don't remember what it was at
> the time where things took a dive performance wise.
>
> http://www.mameui.info/Bench.htm
Just to be pedantic, taz-nz's test was with ARCADE (formerly known as MAMEUIFX though Robbert has done some considerable work to it), not standard MAME.
Not that I expect it to be far different in benchmarks to standard MAME...
- Stiletto
|
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: Stiletto]
#384487 - 11/30/19 09:56 AM
|
|
|
I've done back to back benchmarks on Arcade MAME vs official MAME versions in the past, there is no noticeable difference in performance, any variances are generally within the margin of error between runs.
This was just a case of me having my MAME collection with me at work and having a fast PC to play with. The customer has already picked up the PC, but if I get chance with a similar system I'll do a run with Official MAME release, I'll also keep a eye out for a Ryzen 9 3950x build to try.
|
If all else fails, Burn the manual.
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: John IV]
#384488 - 11/30/19 10:08 AM
|
|
|
Yeah I still lurk here on a near daily basis, but a mortgage and other realities of life mean no more annual PC upgrades for me, but I still keep a close eye on CPU IPC improvements in hope some of last MAME performance holdouts will finally be tamed.
Keep up the great work with the benchmarks, it great to resource to see how things have changed over time.
|
If all else fails, Burn the manual.
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: taz-nz]
#384489 - 11/30/19 11:57 AM
|
|
|
Quote:
Yeah I still lurk here on a near daily basis, but a mortgage and other realities of life mean no more annual PC upgrades for me, but I still keep a close eye on CPU IPC
improvements in hope some of last MAME performance holdouts will finally be tamed.
..should be just about the time your unborn children are getting ready to start University. 
|
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: Ziggy100]
#384491 - 11/30/19 01:49 PM
|
|
|
Yeah until AMD pulled ZEN out of the hat, I would have said it would been until the heat death of the universe.
But the last time AMD competed against Intel like this, was in the days of the Socket 939 Athlon 64 and while it took Intel a couple of year to respond, when they did they gave use the Core 2 Duo and huge IPC gains, so here's hoping competition will make Intel get to work and give us some real IPC gains again.
And if they don't AMD is already quietly bragging about the IPC gains with ZEN 3, and if ZEN 1 & ZEN 2 are anything to go by, there is little reason to doubt them. If they can give use a real world IPC gain per core and tie that to higher clock speeds, we could see AMD capturing the MAME performance crown. 
|
If all else fails, Burn the manual.
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: taz-nz]
#384526 - 12/03/19 11:41 PM
|
|
|
i7-7700K, 4.2G
pongf: 580% (630%)
radikalb: 270% (311%)
starsldr: 57% (80%)
scud: 169% (105.20%)
These results are not consistent. Which commandline are you using? For scud I assume that the benchmark is after startup - scud starts and waits for network here.
Can you rerun the benchmarks with
mame64 -bench 30 -noautosave
and empty nvram and cfg folders?
|
|
|
| John IV |
|
IV/Play, MAME, MAMEUI
|
|
|
|
|
|
Reged: 09/22/03
|
|
Posts: 1969
|
|
Loc: Washington, USA
|
|
|
|
Send PM
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: couriersud]
#384529 - 12/04/19 01:36 AM
|
|
|
Scud's one of the ones that needs to be pre-setup to get to its demo.
[edit] Also couriersud, we use pong vs pongf if that makes a difference.
|
john iv
http://www.mameui.info/
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: John IV]
#384536 - 12/05/19 02:57 PM
|
|
|
pong vs pongf makes a difference. I recommend to use pongf since I only occasionally sync the.
I am aware that scud needs a pre-setup. But that makes comparing benchmarks difficult. To be comparable, benchmarks should be done without cfg, sta and ini files and always the commandline used for the benchmark be published as well.
|
|
|
|
Re: Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
[Re: couriersud]
#384537 - 12/05/19 03:33 PM
|
|
|
> pong vs pongf makes a difference. I recommend to use pongf since I only occasionally
> sync the.
>
> I am aware that scud needs a pre-setup. But that makes comparing benchmarks
> difficult. To be comparable, benchmarks should be done without cfg, sta and ini files
> and always the commandline used for the benchmark be published as well.
without pre-setup some of the benchmarks are 100% meaningless tho. even with them I think ideally we need a 'frames until benchmarking starts' parameter to get speeds that better represent the running game.
this (and regression testing) is one reason I prefer to have default nvrams in sets tho, it means things are more likely to produce useful results / show up if something breaks.
|
|
|
Index 
Threaded 
Intel i9-9900KS @ 5.1ghz Benchmark MAME v0.216
Reply
[Re: 
