Is dolphin x64 faster than x86???

shideravan · 02-26-2010, 08:15 AM #1

A dolphin x64 running under a native x64 OS is much faster than a x86 running on a native x86 OS?
how much?
Thanks!

GodRays · 02-26-2010, 03:56 PM #2

I don't think the difference will be very big, anyway, Don't go and reformat just because of this,lol..

Strong PCs with lots of RAM run 64bit OSes. Your laptop is very weak, though.

spellforce · 02-26-2010, 11:57 PM #3

windows 7 x64 (not vista) is some kinda best for dolphin ,give me extra fps for all of my games.

KarstenS · (This post was last modified: 02-27-2010, 12:01 AM by KarstenS.)

It belongs to the Vendor of the CPU. Some time ago, someone made benchmarks to compare AMD x2 with Intel C2D. The result was, with x86 code the AMD is much slower than the Intel. Using x64 code the AMD has got nearly same speed as the Intel. I don't know how it's today, but I would say its strongly recommended to use x64 code on AMD CPUs. On my AMD CPU there is a difference of 10-15fps between x86 and x64 version of dolphin.

sNk · 02-27-2010, 12:33 AM #5

x64 will alway be faster as a x86 OS.

an intel with SSE 4.1 will give more frames as an AMD with SSE 4a... mhz is not the only thing. extensions are important too.

intel is always better for emulation...

elite · (This post was last modified: 02-27-2010, 04:32 AM by elite.)

Its true that intel is in moment better for emulation. However am not so sure about instructions. Maybe in Dolphin there is bigger difference but for example PCSX2, when using SSE4.1 or not its rather minimal, its nothing. Somebody wrote it help a bit more in SW render but almost nothing in HW.

I tested Dolphin 32 vs 64 not so long ago on my rig with several games(Metroid, SMG, Zelda, Dead Space...). My results differed with different revisions. For example, on one older revision Silent Hill:Shattered Memories was running about same 32 vs 64 wise, but when I turned flashlight 32bit version slowed. Other times SMG crashed on 32bit version but ran fine on 64bit(same revision). On another revision FPS was almost same 32 vs 64 wise in all games I tested, maybe with only small bursts of occasional slowdowns(32bit) that you wouldn't even register. I always used JIT, never JITL. Sometimes sound was a bit cracky on 32bit and I thing that also happened opposite too.

Bottom line is that if there is not something broken, general speed should be about same with maybe occasional short slowdowns in 32bit version on certain situations, that you may not even notice. Still, it may wary from revision to another. If game run in general too slow, 64 bit won't help you as much as it used to. Metroid in beginning slowed to same +-14FPS on same spot for me and if camera wouldn't move it would stay on that FPS no matter what version - and it wouldn't be just a short-time burst.

"RAM: 3GB (2x1GB) DDR2" >> hmm, this is funny Wink

spellforce, did you compared to vista? how much more you got with Win7?(assuming you compared same 32/64 AND same revision!) Also, you are the one publishing intel compiler builds right? Thank you VERY MUCH, please DO continue with it! Wink

KarstenS · 02-27-2010, 04:39 AM #7

(02-27-2010, 04:21 AM)elite Wrote: Bottom line is that if there is not something broken, general speed should be about same with maybe occasional short slowdowns in 32bit version on certain situations, that you may not even notice. Still, it may wary from revision to another. If game run in general too slow, 64 bit won't help you as much as it used to.

Read my post about benchmarks? You are using an Intel CPU. So your result is showing, that that benchmark is correct. Try an AMD CPU and you will see, that there is an difference.

elite · (This post was last modified: 02-27-2010, 10:31 AM by elite.)

(02-27-2010, 04:21 AM)elite Wrote: Bottom line is that if there is not something broken, general speed should be about same with maybe occasional short slowdowns in 32bit version on certain situations, that you may not even notice. Still, it may wary from revision to another. If game run in general too slow, 64 bit won't help you as much as it used to.

Quote:Read my post about benchmarks? You are using an Intel CPU. So your result is showing, that that benchmark is correct. Try an AMD CPU and you will see, that there is an difference.

1: http://www.phoronix.com/scan.php?page=article&item=998&num=3 (this only show same amd cpu can be slower in 64bit mode)
2: http://www.guruht.com/2010/01/intel-core-i5-661-gaming-benchmark-vs.html (crysis 64bit, q9550{2.83ghz} vs phenom 955{3.2ghz})
3: OP have intel core duo
4: I do believe this "amd quicker in 64bit" is total bull..., don't forget there may be many factors to interfere or places to screw test, 1-2 sites doesn't make damn - if that would be true it would be all around internet proven by tests/users. Please send me links if you know some(actual). From CPU "registers" point it doesn't make much sense either.
5: Best would be if somebody with amd cpu make test with Dolphin. However it would have to be on revision proven to have same speed 32 vs 64 bit on intel platform. I doubt its true.

EDIT: http://www.tomshardware.com/reviews/athlon-ii-propus,2414-8.html >> Winrar 64 bit test(old q9550 2.83ghz without on-die memory controller beat phenom II 965 3.4ghz, uh common...)

KarstenS · 02-27-2010, 10:29 AM #9

(02-27-2010, 10:09 AM)elite Wrote: 4: I do believe this "amd quicker in 64bit" is total bull...

Only for you, I did a test...again. 32 and 64 bit dolphin. Same revision. Same settings. Same game: MKWii, 1 player, 50ccm, mushroom cup. After round one of the first course (for JITIL compiling), I compared the fps at the same places (the goal line and the first item box after it). And what shall I say. I do have a difference of 10 fps between 32 and 64 bit. The 64 bit version runs definitively faster on my system.

And now?

PS: Why do the F9 screenshot function not store the fps overlay into screenshot?

elite · (This post was last modified: 02-27-2010, 10:49 AM by elite.)

(02-27-2010, 10:29 AM)KarstenS Wrote:
(02-27-2010, 10:09 AM)elite Wrote: 4: I do believe this "amd quicker in 64bit" is total bull...

Only for you, I did a test...again. 32 and 64 bit dolphin. Same revision. Same settings. Same game: MKWii, 1 player, 50ccm, mushroom cup. After round one of the first course (for JITIL compiling), I compared the fps at the same places (the goal line and the first item box after it). And what shall I say. I do have a difference of 10 fps between 32 and 64 bit. The 64 bit version runs definitively faster on my system.

And now?

PS: Why do the F9 screenshot function not store the fps overlay into screenshot?

I do understand what you say but as I tried to explain in my first post, this 32 vs 64 bit experience can change between revisions, so maybe you have one that works better on 64 bit. Pity I forgot which ran same on my rig, I think it was about r47xx or so, I think it was recommended one for Silent Hill at that time and users reported it ran better that game in 32bit as opposed to 64 bit.

UPDATE: Karsten, could you try r4727? I think that one was where I had same FPS 32 vs 64 wise.