R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
MAME benched on Sandy Bridge @ 4.8 GHz
#248068 - 03/06/11 12:56 AM
|
|
|
I did a direct comparison to John IV's 2011.01.03 set of benches, and added a trip through random other drivers for more context. If anyone has specific requests and there aren't too many, LMK and I'll try it out.
|
|
|
Naoki |
|
|
|
Reged: 11/10/09
|
Posts: 1998
|
Loc: United Kingdom
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248073 - 03/06/11 02:09 AM
|
|
|
What's the 3D gauntlets with video and sound?
Also, dumb question, but what's the speed like in chihiro lol
|
----
On a quest for Digital 573 and Dancing Stage EuroMix 2
By gods I've found it!
|
|
casm |
Cinematronics > *
|
|
|
Reged: 08/27/07
|
Posts: 668
|
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248078 - 03/06/11 02:51 AM
|
|
|
> I did a direct comparison to John IV's 2011.01.03 set of benches, and added a trip > through random other drivers for more context. If anyone has specific requests and > there aren't too many, LMK and I'll try it out.
Pac-Man (driver name 'pacman'), unthrottled, with and without artwork if possible, please.
It may seem like an odd request, but I've been using that as my lowest-common-denominator game for basic benchmarking of relative hardware performance since it can run full-speed across a broad spectrum of old and new hardware.
|
|
|
John IV |
IV/Play, MAME, MAMEUI
|
|
|
Reged: 09/22/03
|
Posts: 1970
|
Loc: Washington, USA
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248079 - 03/06/11 02:56 AM
|
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: casm]
#248083 - 03/06/11 04:04 AM
|
|
|
> Pac-Man (driver name 'pacman'), unthrottled, with and without artwork if possible, > please.
Pac-Man's already there - I assume you mean with some setup other than -bench? Please give the command line you want
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Naoki]
#248084 - 03/06/11 04:05 AM
|
|
|
> What's the 3D gauntlets with video and sound?
Gauntlet Legends is solid 100% during gameplay with video and sound, 150+% unthrottled.
> Also, dumb question, but what's the speed like in chihiro lol
Chihiro doesn't run any code, there is nothing to test.
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248087 - 03/06/11 05:20 AM
|
|
|
Solvalou unthrottled
Gradius 4 unthrottled before/during/after the fire dragon sequence in the first level
WWF Royal Rumble gameplay (about the only semi playable Naomi game I am aware of)
Hokoto No Ken
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248106 - 03/06/11 08:14 AM
|
|
|
I would have asked about Gradius 4, but I see you already covered that. Thanks for some interesting number crunching to ponder over!
I think I'll wait another six to eight months on any possible upgrades, though, as I'm thinking I'll definitely want to jump onto the Thunderbolt train as it leaves the station.
|
---
Try checking the MAME manual at http://docs.mamedev.org
|
|
redk9258 |
Regular
|
|
|
Reged: 09/21/03
|
Posts: 3968
|
Loc: Troy, Illinois USA
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248110 - 03/06/11 09:00 AM
|
|
|
I'll bet that thing can compile like lightning.
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: redk9258]
#248115 - 03/06/11 12:50 PM
|
|
|
> I'll bet that thing can compile like lightning.
Oh yes. make -j9 is cool (The 2600K has 4 physical cores with hyperthreading so the OS sees 8 cores, and yes it gets faster for -j9 vs -j5 so HT does help).
|
|
|
Naoki |
|
|
|
Reged: 11/10/09
|
Posts: 1998
|
Loc: United Kingdom
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248117 - 03/06/11 01:33 PM
|
|
|
Now to save up for Sandy Bride lol
It doesn't run code but takes all of my CPU's power and gives me under 1%, but then I suppose it won't matter for some years to come.
|
----
On a quest for Digital 573 and Dancing Stage EuroMix 2
By gods I've found it!
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Naoki]
#248119 - 03/06/11 01:48 PM
|
|
|
Lol @ Naomi figure......
Maybe by space year 2025 there will be a CPU that can run it at a solid 60fps...
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Ziggy100]
#248131 - 03/06/11 04:30 PM
|
|
|
> Lol @ Naomi figure...... > > Maybe by space year 2025 there will be a CPU that can run it at a solid 60fps...
If you underclock the CPU to about 70 MHz Naomi/AW games are quite playable on this system, around 85% with sound. That's sufficient to blast through several levels of Metal Slug 6 anyway
|
|
|
casm |
Cinematronics > *
|
|
|
Reged: 08/27/07
|
Posts: 668
|
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248137 - 03/06/11 05:24 PM
|
|
|
> > Pac-Man (driver name 'pacman'), unthrottled, with and without artwork if possible, > > please. > > Pac-Man's already there - I assume you mean with some setup other than -bench? Please > give the command line you want
Oh, yeah, that would help, wouldn't it? Without artwork, I usually do:
-noautoframeskip -frameskip 0 -seconds_to_run 90 -nothrottle -nosleep -video opengl -skip_gameinfo -effect none -nowaitvsync -noreadconfig gamename
And with artwork, it's:
-noautoframeskip -frameskip 0 -seconds_to_run 90 -nothrottle -nosleep -video opengl -skip_gameinfo -effect none -nowaitvsync -noreadconfig -use_backdrops -use_overlays -use_bezels
Note that -video is set to 'opengl' in both commandlines since I'm in Macland. Running 'ddraw' on Windows is OK for a rough comparison if that's what's available, but I would like to know if it was used instead of OpenGL.
Can we try it this time with drivername 'invaders'? Didn't see that one on this list this time (dunno how I missed Pac-Man), so should be good to go
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248139 - 03/06/11 05:40 PM
|
|
|
> > Lol @ Naomi figure...... > > > > Maybe by space year 2025 there will be a CPU that can run it at a solid 60fps... > > If you underclock the CPU to about 70 MHz Naomi/AW games are quite playable on this > system, around 85% with sound. That's sufficient to blast through several levels of > Metal Slug 6 anyway
With a bit of stuttuttutteringngngngng.
|
|
|
Naoki |
|
|
|
Reged: 11/10/09
|
Posts: 1998
|
Loc: United Kingdom
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Heihachi_73]
#248155 - 03/06/11 07:36 PM
|
|
|
Gawd that reminds me of aaaataaatatatatataatattttttttttt nnnnneneneneeeneeeeeeoooeoeoeoeoeoeoeeeooooddddddddydydydddyddyyyyyynynynnnynynynynynynnneneeneneneneneneeeeeeee iiiiiiiiiiiiiinninininnnninnninnnddndndddndddndndddduuudududududuuuuuuudududdudsususususssstseteteterrererree-*quits*
|
----
On a quest for Digital 573 and Dancing Stage EuroMix 2
By gods I've found it!
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248162 - 03/06/11 10:46 PM
|
|
|
> > I'll bet that thing can compile like lightning. > > Oh yes. make -j9 is cool (The 2600K has 4 physical cores with hyperthreading so the > OS sees 8 cores, and yes it gets faster for -j9 vs -j5 so HT does help).
What do the numbers mean in that make -j command?
-j5 -j7 etc...
|
"Note to Noobs:
We are glad to help you but simply posting that something does not work is not going to lead to you getting help. The more information you can supply defining your problem, the less likely it will be that you will get smart-alec replies.
C.D.~"
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Master O]
#248166 - 03/07/11 12:51 AM
|
|
|
> > > I'll bet that thing can compile like lightning. > > > > Oh yes. make -j9 is cool (The 2600K has 4 physical cores with hyperthreading so the > > OS sees 8 cores, and yes it gets faster for -j9 vs -j5 so HT does help). > > What do the numbers mean in that make -j command? > > -j5 -j7 etc...
Number of threads to use (it can compile that many files at once). Canonically it should be the number of processors plus 1, although on single-core systems "2" is usually slower than "1".
|
|
|
redk9258 |
Regular
|
|
|
Reged: 09/21/03
|
Posts: 3968
|
Loc: Troy, Illinois USA
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248173 - 03/07/11 01:59 AM
|
|
|
So how long does a compile take? 3 minutes?
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: redk9258]
#248186 - 03/07/11 04:50 AM
|
|
|
> So how long does a compile take? 3 minutes?
My 6 core Phenom II can compile the entire MAME source tree in about two and a half minutes; it would be interesting to see how much faster Sandy Bridge's 4 real + 4 fake cores are than a true 6 core processor.
I bought the Phenom II about 5 months ago specifically to speed up my MAME compiles (there is very little that I do that cares about processing power aside from compiles), and so far could not be happier with it. But more speed is always welcome.
|
|
|
redk9258 |
Regular
|
|
|
Reged: 09/21/03
|
Posts: 3968
|
Loc: Troy, Illinois USA
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Bryan Ischo]
#248198 - 03/07/11 07:13 AM
|
|
|
2.5 minutes is amazing. I upgraded my machine about 15 months ago from an Athlon XP 2400 to a Core2Quad 9400. My compile times went from over an hour to around 6 minutes. Now it is closer to 10 minutes. I'm really curious what R. Belmont's compile time is on that super computer.
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: redk9258]
#248202 - 03/07/11 10:06 AM
|
|
|
I think there really is no substitute for cores when it comes to compiling, which is why I went for the 6 core processor; I know that the Intel chips were faster per core but the only 6 core Intel chip at the time was just ridiculously expensive. That is probably still true.
Also I am not overclocking the Phenom II in any way; it's just a stock 1075T at 3.0 Ghz, for what it's worth.
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248205 - 03/07/11 01:04 PM
|
|
|
There's an interesting natural dichotomy to the whole process. You'll have 'optimized' (sometimes somewhat inaccurate, often with graphics 'enhancements') emulators, but over a period of time a high-end arcade hardware becomes more and more playable in MAME on the new hardware of the era and people just naturally drift to that as the all-in-one solution.
Not that it entirely matters-- it's the accurate documentation that MAME provides that allows for these less accurate emulators to exist in many cases anyway, and MAME is all about documentation as compared to playability in the first place.
That still makes it rather interesting to watch the wave of progress-- into MAME, passing into optimized/high level emulators, then back into MAME.
|
---
Try checking the MAME manual at http://docs.mamedev.org
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Bryan Ischo]
#248215 - 03/07/11 03:57 PM
|
|
|
Compiles are generally HDD bound once you hit 4+ cores. If I do make clean / make / make clean / make on Fedora the second one takes about 20 seconds start to finish because the entire source tree's already in RAM. I'll time a completely cold run sometime though.
(I should note that on this system the compiler binaries and system headers and libs are all on an SSD, which also helps).
Edited by R. Belmont (03/07/11 04:05 PM)
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248221 - 03/07/11 05:42 PM
|
|
|
BTW, sorry the requests are late - I tried to install Win7 SP1 after doing those benches and ended up having to reinstall the OS from scratch (similar to the problems H*z* said he had with SP1). Once it's back to working again I'll get on those
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248224 - 03/07/11 05:49 PM
|
|
|
> BTW, sorry the requests are late - I tried to install Win7 SP1 after doing those > benches and ended up having to reinstall the OS from scratch (similar to the problems > H*z* said he had with SP1). Once it's back to working again I'll get on those
what problem did you have?
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248227 - 03/07/11 07:30 PM
|
|
|
> Compiles are generally HDD bound once you hit 4+ cores. If I do make clean / make / > make clean / make on Fedora the second one takes about 20 seconds start to finish > because the entire source tree's already in RAM. I'll time a completely cold run > sometime though.
Hm, I timed my compiles after building my system by creating a RAM disk and copying the entire MAME tree to it before invoking the compile, and my time was around 2 and a half minutes. This does not differ significantly from my 'cold' compile time after a fresh boot with no RAM disk involved. So in my experience, the HDD is not the limiting factor, at least not with a 6 core Phenom II; maybe it is the limiting factor for your Sandy Bridge, though. At a 20 second full compile I can certainly see that being the case.
BTW my HDD is an Intel X-25M G2, and I wouldn't be surprised if a platter HDD would be the limiting factor if that's what someone has.
So to recap:
Sandy Bridge w/SSD or RAM cache > Phenom II w/SSD or RAM cache > most anything else with SSD or RAM cache > anything with HDD
EDIT:
Are you *sure* about that 20 second compile time? Thinking about it, that's just incredible. I mean that's 7.5 times as fast as my Phenom II. If you told me that the Sandy Bridge compiled 2x as fast as the Phenom II, I'd raise my eyebrows a little bit but still believe you. But 7.5x as fast? That's really hard to believe, unless my system is somehow misconfigured. But I did compare my system against an Intel Xeon E5540 that I have access to, and my Phenom II system as about 10% - 20% faster, which matches up with other benchmarking results I've found out there. So I think my Phenom II results were good. I just don't understand how a Sandy Bridge even at 4.8 Ghz would be 7.5x as fast as a Phenom II at 3.0 Ghz. If the Sandy Bridge were, clock for clock, 100% faster than the Phenom II per core (and that is quite generous), then I would expect the Sandy Bridge to be 2.13x as fast as the Phenom II. Even if the Sandy Bridge somehow magically added 2 real cores, it would only be 3.2x as fast. I just don't understand where a 7.5x speedup is coming from.
Edited by Bryan Ischo (03/07/11 07:40 PM)
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: ]
#248228 - 03/07/11 07:35 PM
|
|
|
> > BTW, sorry the requests are late - I tried to install Win7 SP1 after doing those > > benches and ended up having to reinstall the OS from scratch (similar to the > problems > > H*z* said he had with SP1). Once it's back to working again I'll get on those > > what problem did you have?
First it wouldn't even DL the update and gave an 8xxxxxxx error code that turned out to mean "corrupt component cache". There was a KB article with a utility that fixed that. Then it DLed the update but wouldn't apply it because "some components are missing" and the only solution given was to reinstall the OS.
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Bryan Ischo]
#248231 - 03/07/11 07:43 PM
|
|
|
> BTW my HDD is an Intel X-25M G2, and I wouldn't be surprised if a platter HDD would > be the limiting factor if that's what someone has.
I run a split system - in Fedora everything but /home is on the SSD, /home is on a large conventional platter drive both for space and to avoid excessive writes on the flash. Thus when compiling GCC itself and all the system headers and libraries come off the SSD.
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248235 - 03/07/11 08:10 PM
|
|
|
> Then it DLed the update but wouldn't apply it because "some components are > missing" and the only solution given was to reinstall the OS.
Was that an customized (like in "n-lited") install that was not accepted?
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: CrapBoardSoftware]
#248244 - 03/07/11 09:11 PM
|
|
|
> > Then it DLed the update but wouldn't apply it because "some components are > > missing" and the only solution given was to reinstall the OS. > > Was that an customized (like in "n-lited") install that was not accepted?
Nope. Straight off a retail DVD. The only "unusual" thing done to it was transferring the HDD from the old machine to the SNB, but it was running fine prior to this.
|
|
|
TrevEB |
Brokering peace between the collectors and the Mame Community
|
|
|
Reged: 03/11/05
|
Posts: 770
|
Loc: Oakland, CA
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248249 - 03/07/11 10:05 PM
|
|
|
Wow! Finally, my overclocked e8400 has been left in the dust. Well 3 years was a good run.
If you have the time, it would be good to run the same test with the CPU at the default clock speed.
Not every CPU was born a winner.
|
|
|
redk9258 |
Regular
|
|
|
Reged: 09/21/03
|
Posts: 3968
|
Loc: Troy, Illinois USA
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248252 - 03/07/11 10:22 PM
|
|
|
> First it wouldn't even DL the update and gave an 8xxxxxxx error code that turned out > to mean "corrupt component cache". There was a KB article with a utility that fixed > that. Then it DLed the update but wouldn't apply it because "some components are > missing" and the only solution given was to reinstall the OS.
I'm glad I downloaded SP1 in ISO form. I updated 3 computers and did not have any problems. Of course I had current Ghost images too.
It's funny how Windows 7 will install in 15 minutes but takes an hour to apply a service pack. I wonder if an upgrade install would have been faster from a Win7 SP1 DVD (or even possible).
|
|
|
Foxhack |
Furry guy
|
|
|
Reged: 01/30/04
|
Posts: 2409
|
Loc: Spicy Canada
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248312 - 03/08/11 07:18 AM
|
|
|
> > > Then it DLed the update but wouldn't apply it because "some components are > > > missing" and the only solution given was to reinstall the OS. > > > > Was that an customized (like in "n-lited") install that was not accepted? > > Nope. Straight off a retail DVD. The only "unusual" thing done to it was transferring > the HDD from the old machine to the SNB, but it was running fine prior to this.
I saw a lot of people having similar problems at the GOG forums. Microsoft should've pulled SP1 until they fixed things...
|
|
|
redk9258 |
Regular
|
|
|
Reged: 09/21/03
|
Posts: 3968
|
Loc: Troy, Illinois USA
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Foxhack]
#248318 - 03/08/11 08:26 AM
|
|
|
> > > > Then it DLed the update but wouldn't apply it because "some components are > > > > missing" and the only solution given was to reinstall the OS. > > > > > > Was that an customized (like in "n-lited") install that was not accepted? > > > > Nope. Straight off a retail DVD. The only "unusual" thing done to it was > transferring > > the HDD from the old machine to the SNB, but it was running fine prior to this. > > I saw a lot of people having similar problems at the GOG forums. Microsoft should've > pulled SP1 until they fixed things...
They certainly had plenty of time. Did you notice the dates on the the packages are Nov 22, 2010?
|
|
|
Moose |
Don't make me assume my ultimate form!
|
|
|
Reged: 05/03/04
|
Posts: 1483
|
Loc: Outback, Australia
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: redk9258]
#248383 - 03/09/11 04:31 AM
|
|
|
> > First it wouldn't even DL the update and gave an 8xxxxxxx error code that turned > out > > to mean "corrupt component cache". There was a KB article with a utility that fixed > > that. Then it DLed the update but wouldn't apply it because "some components are > > missing" and the only solution given was to reinstall the OS. > > > I'm glad I downloaded SP1 in ISO form. I updated 3 computers and did not have any > problems. Of course I had current Ghost images too. > > It's funny how Windows 7 will install in 15 minutes but takes an hour to apply a > service pack. I wonder if an upgrade install would have been faster from a Win7 SP1 > DVD (or even possible).
Would be interesting to slipstream SP1 into the install DVD files to create an SP1 install DVD. This should save a lot of time ?
|
Moose
|
|
redk9258 |
Regular
|
|
|
Reged: 09/21/03
|
Posts: 3968
|
Loc: Troy, Illinois USA
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Moose]
#248387 - 03/09/11 04:47 AM
|
|
|
> Would be interesting to slipstream SP1 into the install DVD files to create an SP1 > install DVD. This should save a lot of time ?
There's no official way to do that although someone has a way to do it.
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248393 - 03/09/11 05:56 AM
|
|
|
> I did a direct comparison to John IV's 2011.01.03 set of benches, and added a trip > through random other drivers for more context. If anyone has specific requests and > there aren't too many, LMK and I'll try it out.
Can you bench test Metal Slug 6, now that it is emulated?
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: Naoki]
#248850 - 03/14/11 09:55 PM
|
|
|
How does Dkong play?
Full speed?
|
|
|
Naoki |
|
|
|
Reged: 11/10/09
|
Posts: 1998
|
Loc: United Kingdom
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: ]
#248855 - 03/14/11 11:14 PM
|
|
|
Surprisingly yes
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: ]
#248866 - 03/15/11 03:24 AM
|
|
|
> How does Dkong play? > > Full speed?
Dkong is full speed on an Atom, last I heard.
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#248873 - 03/15/11 04:20 AM
|
|
|
> > How does Dkong play? > > > > Full speed? > > Dkong is full speed on an Atom, last I heard.
It is; I benchmarked it at 291.72% real time when run unthrottled on an Atom 330, so it would have no problem running at 100% when run at normal speed.
|
|
|
|
How about a clock for clock comparison.
[Re: R. Belmont]
#252142 - 04/19/11 03:37 AM
|
|
|
What I've would of liked of seen is your SB clocked at 3.6GHz just like john's E8400. Of course your 4.8GHz SB will blow the E8400 away so it's not a good comparison. What I would like to know and I'm others would like to see I'm sure is a clock for clock comparison.
Has anybody have a link or done a clock for clock comparison I would just love to see this. thanks
Edited by ScottF (04/19/11 03:37 AM)
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: ScottF]
#252157 - 04/19/11 06:33 AM
|
|
|
> What I've would of liked of seen is your SB clocked at 3.6GHz just like john's E8400. > Of course your 4.8GHz SB will blow the E8400 away so it's not a good comparison. What > I would like to know and I'm others would like to see I'm sure is a clock for clock > comparison.
Because insane overclocks are the entire fucking point of SNB[1] if you already have a Wolfdale (if you have a Merom/Conroe era C2D, the per-clock improvement actually does start to get interesting, and if you have a P4, run don't walk to upgrade). If you're going to run it stock, then yes, you can assume performance about 5% better than a same-clocked Wolfdale on software not using the new AVX instructions (which will be nearly everything for another 12 months while the compiler guys get going). But at that point you've spent a lot of money for mostly a lot of nothing.
[1] Intel marketing violently disagrees with this, but I'm being real here.
Edited by R. Belmont (04/19/11 06:41 AM)
|
|
|
|
Re: How about a clock for clock comparison.
[Re: R. Belmont]
#252168 - 04/19/11 03:46 PM
|
|
|
Well I have latop with SB cpu so theres no fucking point of OC'ing it would probably over heat;)
I don't know much about AVX instructions set. But how much of a gain do you think that will boost FPS in mame ?
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: ScottF]
#252190 - 04/19/11 08:52 PM
|
|
|
> Well I have latop with SB cpu so theres no fucking point of OC'ing it would probably > over heat
LOL. Fair enough. I actually lost 100 MHz when I got my recall-replacement motherboard (only 4.7 GHz now, but gauntleg is still comfortably 100% in-game). Should re-run all the benches like that and then down at stock sometime to get an idea of the scaling.
> I don't know much about AVX instructions set. But how much of a gain do you think > that will boost FPS in mame ?
I don't expect it'll do much for MAME in a general sense, but from what I've read they'd be near-ideal for implementing something like the N64 RSP. So maybe N64 games will become playable but only on SNB :-)
|
|
|
|
Re: How about a clock for clock comparison.
[Re: R. Belmont]
#252197 - 04/19/11 09:10 PM
|
|
|
> Because insane overclocks are the entire fucking point of SNB[1] if you already have > a Wolfdale (if you have a Merom/Conroe era C2D, the per-clock improvement actually > does start to get interesting, and if you have a P4, run don't walk to upgrade). If > you're going to run it stock, then yes, you can assume performance about 5% better > than a same-clocked Wolfdale on software not using the new AVX instructions (which > will be nearly everything for another 12 months while the compiler guys get going). > But at that point you've spent a lot of money for mostly a lot of nothing.
That is very interesting. So could I then conclude that Intel is selling their Sandy Bridge chips at an artificially low clock rate because they don't have any competition that would force them to sell them closer to their potential speed? Meaning that if AMD was more competitive, Intel could just as easily sell these chips for the current price at higher clock rates as they can sell them now for the same price at lower clock rates?
If so, it only underscores the value of AMD in the x86 marketplace. We can all just pray that Bulldozer is competitive.
Or - is there another alternative - which is that in order to achieve these 'insane' Sandy Bridge overclocks you have to use either a very expensive cooling solution, or a very loud one, or both, which thus prevents Intel from selling these things at such high clock rates?
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: Bryan Ischo]
#252208 - 04/19/11 10:42 PM
|
|
|
> That is very interesting. So could I then conclude that Intel is selling their Sandy > Bridge chips at an artificially low clock rate because they don't have any > competition that would force them to sell them closer to their potential speed?
Well, I'll freely admit MAME is a special case here that makes SNB look better than it might in a more general sense. There are very few other workloads you can impose on a PC in which per-clock single-threaded performance is of paramount importance nowadays.
> Meaning that if AMD was more competitive, Intel could just as easily sell these chips > for the current price at higher clock rates as they can sell them now for the same > price at lower clock rates?
Yes. IMO they could easily sell SNB at a stock clock of 4+ GHz without blowing out the thermal envelope. And actually the current pricing on SNB is quite good historically - for ~$300 you can get a 2600K that can blow the doors off every $999 Extreme Edition Intel's ever produced.
> If so, it only underscores the value of AMD in the x86 marketplace. We can all just > pray that Bulldozer is competitive.
Bulldozer is reportedly optimized for multi-threaded workloads at the specific expense of single-threaded instructions per clock. That's totally viable for a lot of workloads now and it'll probably review and compete well (especially in servers), but it's not great for MAME.
> Or - is there another alternative - which is that in order to achieve these 'insane' > Sandy Bridge overclocks you have to use either a very expensive cooling solution, or > a very loud one, or both, which thus prevents Intel from selling these things at such > high clock rates?
I'm using a ~$50 air cooler to run 1 GHz over stock, and it's got 2 large slow fans in a push-pull configuration, so it's no louder than the stock Intel cooler. Remember, Intel was shipping the Prescott P4s about 6 years ago, and those things idled very hot, let alone under load. If they brought back that cooler they'd be all set for stock-rated 4+ GHz parts I think.
|
|
|
|
Re: How about a clock for clock comparison.
[Re: R. Belmont]
#252212 - 04/20/11 12:01 AM
|
|
|
> > Or - is there another alternative - which is that in order to achieve these > 'insane' > > Sandy Bridge overclocks you have to use either a very expensive cooling solution, > or > > a very loud one, or both, which thus prevents Intel from selling these things at > such > > high clock rates? > > I'm using a ~$50 air cooler to run 1 GHz over stock, and it's got 2 large slow fans > in a push-pull configuration, so it's no louder than the stock Intel cooler. > Remember, Intel was shipping the Prescott P4s about 6 years ago, and those things > idled very hot, let alone under load. If they brought back that cooler they'd be all > set for stock-rated 4+ GHz parts I think.
Let me preface this remark by saying that $50 additional for a better heat sink/fan probably does not alter the fact that the Sandy Bridge has superior price:performance when compared to most (all?) other x86 chips, at least for the MAME workload.
However, it's unknown how much Intel would have to charge for a Sandy Bridge with a stock cooling solution that could reach the overclocked speeds you have quoted without excessive noise or size; I think that stock coolers generally are low profile to reduce the potential for layout issues that wouldn't allow the processor + heatsink/fan to fit in the majority of cases. And you are not going to get a 120 mm push-pull configuration in a low profile cooler. Which means that they would have to have a fan that is low profile but with the same heat dissipation capacity as your 120 mm push/pull, and such a cooling solution might add significantly to the cost of the Sandy Bridge chip.
Of course, the Sandy Bridge as an OEM part without a fan would not carry this burden and could be lots cheaper even if rated to run at higher clock speeds.
Also there is the power consumption angle; I don't really know how much power a Sandby Bridge uses when overclocked (did you have to overvoltage to get to 4+ Ghz? If so that is troubling for alot of reasons, processor lifetime not the least of them), but that factors into how much the CPU can be sold for. Nobody wants 150 Watt P4s anymore (and for good reason, they sucked in so many ways).
Finally, would you be able to confirm your earlier claim that your overclocked Sandy Bridge can compile the entire MAME source tree from a clean state in 30 seconds? I find this claim incredible considering that my 3.0 Ghz 6 core Phenom II can only do it in 2.5 minutes. A 5x speedup with 2 fewer cores just sounds unbelievably incredible. Although, that being said, maybe there is something wrong with my setup; I think I calculated that the compiled lines of code per second for the MAME tree at 2.5 minutes is something like 2700 lines second per processor for my 6 processor system ((~ 2.5 million lines of MAME code) / (6 procs * 150 seconds = 900 seconds) = ~2700 lines of code per second per processor), and I just expected it to be much higher than that given that the old Sun Sparcstation that I used to use in university in 1992 could do 1000 lines per second (of course code was simpler back then - C vs. C++ - but still ...).
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: Bryan Ischo]
#252213 - 04/20/11 12:18 AM
|
|
|
> However, it's unknown how much Intel would have to charge for a Sandy Bridge with a > stock cooling solution that could reach the overclocked speeds you have quoted
Right. I don't expect that they could ship much above probably 4.2 GHz with their usual coolers, but that's still a pretty good boost over stock. I'm actually surprised there isn't at least a 4.0 GHz rated version so they can claim the milestone. > Of course, the Sandy Bridge as an OEM part without a fan would not carry this burden > and could be lots cheaper even if rated to run at higher clock speeds. > > Also there is the power consumption angle; I don't really know how much power a > Sandby Bridge uses when overclocked (did you have to overvoltage to get to 4+ Ghz? If
You need a very minor overvolt (less than 0.1 volt) on most production parts to get past 4.6 GHz. At or below that it runs happily at stock voltage.
> Finally, would you be able to confirm your earlier claim that your overclocked Sandy > Bridge can compile the entire MAME source tree from a clean state in 30 seconds?
Sure. The key is that you have 8 GB of RAM, nothing large running to eat it, and you do make clean ; make ; make clean ; make. Everything runs out of RAM cache then and it's super-fast pretty much regardless of your processor speed.
Also, I should mention that SNB has significantly different power management behavior from previous chips. When the OS is idling or under light loads, it drops all cores to 1.6 GHz. When you crank up MAME or GCC or whatever, it clocks up to whatever your overclock is. That results in a significant noise/heat/power improvement in normal web/email usage compared to overclocking previous CPUs - mine stays within 10 degrees C of room ambient temperature when it's in that state.
Edited by R. Belmont (04/20/11 12:25 AM)
|
|
|
|
Re: How about a clock for clock comparison.
[Re: R. Belmont]
#252215 - 04/20/11 12:55 AM
|
|
|
> Sure. The key is that you have 8 GB of RAM, nothing large running to eat it, and you > do make clean ; make ; make clean ; make. Everything runs out of RAM cache then and > it's super-fast pretty much regardless of your processor speed.
And I presume this is on Linux, where gcc runs quite nicely? Because I do the same thing on my quad-core hyperthreaded 2.8GHz 6GB Windows system and it's nowhere near 30 seconds, even when entirely in the cache, which I generally attribute to mingw being poorly optimized for Windows.
|
|
|
|
Re: How about a clock for clock comparison.
[Re: AaronGiles]
#252216 - 04/20/11 01:04 AM
|
|
|
> > Sure. The key is that you have 8 GB of RAM, nothing large running to eat it, and > you > > do make clean ; make ; make clean ; make. Everything runs out of RAM cache then and > > it's super-fast pretty much regardless of your processor speed. > > And I presume this is on Linux, where gcc runs quite nicely? Because I do the same > thing on my quad-core hyperthreaded 2.8GHz 6GB Windows system and it's nowhere near > 30 seconds, even when entirely in the cache, which I generally attribute to mingw > being poorly optimized for Windows.
Yes, my compiles are timed on Linux; and the thing is, I created a 2 GB RAM disk and copied the entire MAME source tree to the RAM disk before compiling, so I thought I had eliminated all I/O from my benchmarking. I only have 4 GB total memory so only 2 GB was left for the compiler but I checked and I never went into swap during the compile. Then again, since the compiled source was in the ram disk and the generated objects files were too, maybe memory bandwidth was limiting me?
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: AaronGiles]
#252226 - 04/20/11 02:58 AM
|
|
|
> > Sure. The key is that you have 8 GB of RAM, nothing large running to eat it, and > you > > do make clean ; make ; make clean ; make. Everything runs out of RAM cache then and > > it's super-fast pretty much regardless of your processor speed. > > And I presume this is on Linux, where gcc runs quite nicely? Because I do the same > thing on my quad-core hyperthreaded 2.8GHz 6GB Windows system and it's nowhere near > 30 seconds, even when entirely in the cache, which I generally attribute to mingw > being poorly optimized for Windows.
In fairness, 30 seconds was pre-C++, pre-all-the-MESS-CPUs-being-enabled, and pre-Micko-and-Curt moving all of MESS into the MAME tree. It's 1 minute 15 seconds on a make -j9 now, including linking. (C++ is bad juju for GCC).
|
|
|
|
Re: MAME benched on Sandy Bridge @ 4.8 GHz
[Re: R. Belmont]
#252289 - 04/20/11 01:48 PM
|
|
|
Last time I heard Atom was cancelled by Intel due to problems in Fukushima
Best regards, RetroLover
|
|
|
Sune |
Connected
|
|
|
Reged: 09/21/03
|
Posts: 5648
|
Loc: Lagoa Santa, Brasil
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: Bryan Ischo]
#252354 - 04/20/11 09:28 PM
|
|
|
> (...) And you are not going to get a 120 mm > push-pull configuration in a low profile cooler. Which means that they would have to > have a fan that is low profile but with the same heat dissipation capacity as your > 120 mm push/pull, and such a cooling solution might add significantly to the cost of > the Sandy Bridge chip.
I think you misunderstood, push-pull refers to case fans, one on the front (or side, if the case is big enough) sucking in air and one at the back blowing it out.
Nothing to do with the CPU cooler.
S
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: Sune]
#252359 - 04/20/11 10:25 PM
|
|
|
> I think you misunderstood, push-pull refers to case fans, one on the front (or side, > if the case is big enough) sucking in air and one at the back blowing it out. > > Nothing to do with the CPU cooler.
Nope, it *is* the CPU cooler.
|
|
|
Sune |
Connected
|
|
|
Reged: 09/21/03
|
Posts: 5648
|
Loc: Lagoa Santa, Brasil
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: R. Belmont]
#252368 - 04/20/11 11:50 PM
|
|
|
> > I think you misunderstood, push-pull refers to case fans, one on the front (or > side, > > if the case is big enough) sucking in air and one at the back blowing it out. > > > > Nothing to do with the CPU cooler. > > Nope, it *is* the CPU cooler.
Nice, I had no idea. Looking at some on google images..which model do you have?
S
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: Sune]
#252446 - 04/21/11 05:05 PM
|
|
|
> > Nope, it *is* the CPU cooler. > > Nice, I had no idea. Looking at some on google images..which model do you have?
Corsair A70. It's super-big, but remarkably easy to install compared to other large coolers I've dealt with and it cools great for the price.
|
|
|
Fever |
Reformed Sk3n3 Quitter
|
|
|
Reged: 09/21/03
|
Posts: 2090
|
Loc: Britland
|
|
Send PM
|
|
|
Re: How about a clock for clock comparison.
[Re: ScottF]
#252553 - 04/22/11 11:10 PM
|
|
|
> What I've would of liked of seen is your SB clocked at 3.6GHz just like john's E8400. > Of course your 4.8GHz SB will blow the E8400 away so it's not a good comparison. What > I would like to know and I'm others would like to see I'm sure is a clock for clock > comparison. > > Has anybody have a link or done a clock for clock comparison I would just love to see > this. thanks
Damn, I just glanced at this poster's nic and for a moment I thought it said 'Stroff' - disappointing!
|
|
|
|
A 2500K comparison, something seems off for me.
[Re: R. Belmont]
#256163 - 05/31/11 03:39 PM
|
|
|
Ran benches with MameUI64 .141 in Win7 x64 4.5Ghz i5 2500K (2 x 2GB) 240-Pin DDR3 GeForce GTX550-TI Bus clock 100,CPU multiplier 45
MAME 0.141 64-bit mamedev.org official binary, -bench 90 used for all games to match John IV’s latest Wolfdale run. All games had no nvram except scud and gradius4, which don’t show gameplay until configured. All numbers are % speed where 100 is full framerate.
I copy and pasted whats above off of RB's site I only used John's mameui64 1.41 not sure what bench 90 means. Also did not mess with nvram settings would not know how anyway.
Some of the numbers seem off to me especially gauntleg,starsldr. The numbers fluctuant alot in game so I'm giving you a avarage. I'm not sure how or when RB or John gauge theirs.
RB's 4.8Ghz i7 2600K """""""""""""""""""" blitz: 258 dolphin: 38 gauntleg: 370 gradius4: 236 propcycl: 259 radikalb: 210 starsldr: 85
Kano11's 4.5Ghz i5 2500K """""""""""""""""""" blitz: 185 dolphin: 35 gauntleg: 222 gradius4: 200 propcycl: 210 radikalb: 185 starsldr: 55
John's Wolfdale e8400 at 3.6Ghz """"""""""""""""""""""""""""""" blitz: 166 dolphin: 25 gauntleg: 221 gradius4: 125 propcycl: 135 radikalb: 164 starsldr: 58
|
|
|
John IV |
IV/Play, MAME, MAMEUI
|
|
|
Reged: 09/22/03
|
Posts: 1970
|
Loc: Washington, USA
|
|
Send PM
|
|
|
Re: A 2500K comparison, something seems off for me.
[Re: Kano11]
#256191 - 05/31/11 11:59 PM
|
|
|
To get the apples to apples comparisons you need to run an x64 build on Windows 7 x64 w/ the commandnline: 'mame64 dolphin -bench 90' for example. The bench argument runs the emulated amount of the first XX seconds w/ certain other features turned off. You can use MameUI but it doesn't output to console so you'd need to redirect to a file.
The nvram that RB is talking about is for SCUD Race and Gradius IV. They need to be 'setup' before the benchmark run by going into their F2 menus.
|
|
|
|
Re: A 2500K comparison, something seems off for me.
[Re: John IV]
#256207 - 06/01/11 03:32 AM
|
|
|
> To get the apples to apples comparisons you need to run an x64 build on Windows 7 x64 > w/ the commandnline: > 'mame64 dolphin -bench 90' for example. The bench argument runs the emulated amount > of the first XX seconds w/ certain other features turned off. You can use MameUI but > it doesn't output to console so you'd need to redirect to a file. > > The nvram that RB is talking about is for SCUD Race and Gradius IV. They need to be > 'setup' before the benchmark run by going into their F2 menus.
ok thanks, I thought it would be something like that, I don't know how to do commandnline. I've only used mame32 and now mameui since the beginning. Thanks for telling me this stuff. I thought there was something wrong with my setup.
|
|
|