Core i5-2500K System Testing

[MegaJump]

Hey all,

Just finished building my new system, featured the Gigabyte GA-P67A-UD4-B3 motherboard, as well as the new Core i5-2500K processor. So far, I haven't had much time to mess around with it yet (ie. OC or anything), but the results have been very nice so far. SMG is playable at FULL SPEED, at a resolution of 720p so far. I even had to put the frame limiter on, otherwise the main hub world in SMG was hovering around 75-80 fps. Using r7358.
I've included my settings just for reference.
           

My current system is:

Core i5-2500K 3.3Ghz
Gigabyte GAP67A-UD4-B3
8GB Corsair XMS DDR3 1600 at 1600Mhz
AMD Radeon HD 6850
Acer H243H 24" 1080p monitor

Will test some more over the next few days, but if you guys have some specific settings or games that will stress the my system, let me know so I can try them out. Thanks

EDIT: I just did some more testing, and the average FPS in the hub is 60-75 just to update.
Update: F-Zero GX - Majority of tracks run at a smooth 60fps with auto frameskip. Sand Ocean still won't go beyond 37-38fps.
Update: Metroid: Other M - Runs at 60fps, no slowdowns from weapon charging, multiple enemies or large enviroments. There still does seem to be a small lag in controller input.
Update: Metroid Prime 2 - Echoes runs at 60fps. No slowdowns, no glitches at 1920 x 1080 resolution.

[KHRZ]

75-80 fps? *Drools*

You think there's any chance this i5 would stay over 50 fps on a laptop then?

[tuanming]

For emulation, those SB CPUs are just about 2x faster than the Core 2 series. I'll waiting to see what Ivy Bridge can offer =)

[remedy2]

(03-17-2011, 09:40 AM)tuanming Wrote: For emulation, those SB CPUs are just about 2x faster than the Core 2 series. I'll waiting to see what Ivy Bridge can offer =)

Not too sure about 2x clock for clock but definitely a major improvement. IB is supposed to be a 20% improvement overall, we'll see what this means for emulation purposes though.

[turingpest]

@MegaJump

Are you using DSP-LLE recompiler for audio? Without DSP-LLE on it's own thread?

@tuanming

"For emulation, those SB CPUs are just about 2x faster than the Core 2 series." ah, more hyperbole. proof please, or this is bullshit.

[skid]

(03-17-2011, 10:02 PM)turingpest Wrote: "For emulation, those SB CPUs are just about 2x faster than the Core 2 series." ah, more hyperbole. proof please, or this is bullshit.

It's true.

I went from a Core2 Duo (3.6ghz) to a i7 2600K (3.4ghz) and my framerate doubled in most of my games like SMG, F-ZERO GX, NSMBW. Some games gained more, others less, but generally speaking, it doubled the framerate in Dolphin.

Going from DDR2 800 RAM to DDR3 1600 RAM allowed me to use EFB to RAM in all of my games with less of a hit to the FPS.

The extra 2 cores in the 2600K allowed me to turn on LLE JIT on thread with less of a speed hit.

[turingpest]

(03-17-2011, 10:57 PM)skid Wrote:

(03-17-2011, 10:02 PM)turingpest Wrote: "For emulation, those SB CPUs are just about 2x faster than the Core 2 series." ah, more hyperbole. proof please, or this is bullshit.

It's true.

I went from a Core2 Duo (3.6ghz) to a i7 2600K (3.4ghz) and my framerate doubled in most of my games like SMG, F-ZERO GX, NSMBW. Some games gained more, others less, but generally speaking, it doubled the framerate in Dolphin.

Going from DDR2 800 RAM to DDR3 1600 RAM allowed me to use EFB to RAM in all of my games with less of a hit to the FPS.

The extra 2 cores in the 2600K allowed me to turn on LLE JIT on thread with less of a speed hit.

wow. i thought that the ipc increase for nehalem over core 2 was 20-30% without hyperthreading, and sandy bridge another ~10-15% (not including avx) over that. add in another 10% for going to a quad core from a dual core, and that's still only -50% at best. but 100%! I know that my figures are rough estimates and ipc can mean different things to different people (and do include bus speeds), but i'm still amazed. do you think that dolphin is somehow massively bandwidth bound, even when using efb to texture? it would explain why my core i3 540 was slower than my current i5 750 clock for clock (i don't think it was just L3 cache - the difference was huge). it would also explain the otherwise unexplainable 100% speed increase for sandy bridge over core 2.

[MegaJump]

(03-17-2011, 10:02 PM)turingpest Wrote: @MegaJump

Are you using DSP-LLE recompiler for audio? Without DSP-LLE on it's own thread?

@tuanming

"For emulation, those SB CPUs are just about 2x faster than the Core 2 series." ah, more hyperbole. proof please, or this is bullshit.

Nope turing, I usually use HLE since I'm was never really sure the difference between the two. But I just set it to LLE, and I'll see what happens. Thanks

[turingpest]

(03-18-2011, 01:54 AM)MegaJump Wrote:

(03-17-2011, 10:02 PM)turingpest Wrote: @MegaJump

Are you using DSP-LLE recompiler for audio? Without DSP-LLE on it's own thread?

@tuanming

"For emulation, those SB CPUs are just about 2x faster than the Core 2 series." ah, more hyperbole. proof please, or this is bullshit.

Nope turing, I usually use HLE since I'm was never really sure the difference between the two. But I just set it to LLE, and I'll see what happens. Thanks

It'll be interesting to see. Using DSP-LLE will mean that the background music won't cut out in SMG as with the HLE, but it's also more strenuous on hardware. As of the latest rev (7367) I get ~50fps in the SMG ship hub with DSP-LLE, LLE on thread, dx11 or 9 (both about the same for me) dlst cache and vbeam enabled, using an i5 750 @ 2.8ghz (my motherboard's being RMA'ed so I'm using a basic one with no overclocking features ) 4gb of 1333mhz ram and a hd5750 on w7 x64.

[MegaJump]

(03-18-2011, 02:08 AM)turingpest Wrote:

(03-18-2011, 01:54 AM)MegaJump Wrote:

(03-17-2011, 10:02 PM)turingpest Wrote: @MegaJump

Are you using DSP-LLE recompiler for audio? Without DSP-LLE on it's own thread?

@tuanming

"For emulation, those SB CPUs are just about 2x faster than the Core 2 series." ah, more hyperbole. proof please, or this is bullshit.

Nope turing, I usually use HLE since I'm was never really sure the difference between the two. But I just set it to LLE, and I'll see what happens. Thanks

It'll be interesting to see. Using DSP-LLE will mean that the background music won't cut out in SMG as with the HLE, but it's also more strenuous on hardware. As of the latest rev (7367) I get ~50fps in the SMG ship hub with DSP-LLE, LLE on thread, dx11 or 9 (both about the same for me) dlst cache and vbeam enabled, using an i5 750 @ 2.8ghz (my motherboard's being RMA'ed so I'm using a basic one with no overclocking features ) 4gb of 1333mhz ram and a hd5750 on w7 x64.

Hmmm, just tried loading SMG with LLE enabled, and it crashed immediately. It won't even load anything.