Dolphin, the GameCube and Wii emulator - Forums

Core2duo E6750 @ 2,66 ghz with HD4870 (standard)

1x efb scale 109 fps
2x efb scale 111 fps
3x efb scale 82 fps

1x
2x
3x

Intel Xeon E3110 (AKA core 2 duo 8400) 3.5ghz OC
EVGA 8800 GTS 512 (G92) 678 core 1728 shader 972 memory (stock settings)

-desktop resolution 1080p with aero still on, in case you're wondering


1x 138 fps
2x 133 fps
3x 86 fps
Crashes with AA

cant upload screenshots for some reason though.

(01-05-2011, 10:55 PM)jeffb8810 Wrote: [ -> ]Intel Xeon E3110 (AKA core 2 duo 8400) 3.5ghz OC
EVGA 8800 GTS 512 (G92) 678 core 1728 shader 972 memory (stock settings)

-desktop resolution 1080p with aero still on, in case you're wondering


1x 138 fps
2x 133 fps
3x 86 fps
Crashes with AA

cant upload screenshots for some reason though.

Try to compress them to .jpg or upload them to imageshack if you want.

(01-05-2011, 03:54 PM)NaturalViolence Wrote: [ -> ]No. Can somebody find that post I made where I explained the efb scale/internal resolution in detail? I really don't feel like typing all that again.

Took me ages to find it, but here you are

(12-12-2010, 11:08 AM)NaturalViolence Wrote: [ -> ]ugh.....

I'm only going to explain this once since I have already explained it so many times on this forum, so PAY ATTENTION. The efb (embedded frame buffer) is where the GC/Wii GPU stores the images that it draws (the graphics processor draws images many times per second to send to your display). The native resolution is 640 x 528. In other words on the GC/Wii that is the resolution that it draws these images at. If you turn off efb scaling this is the resolution that dolphin will render the efb at. It will then take that 640 x 528 image and stretch it to fill your screen resolution (so if you selected 1920 x 1080 as your screen resolution is stretches it to 1920 x 1080 then sends a 1920 x 1080 image to your display). A 2x scale means it's rendered at double width and double height, or 1280 x 1056, 3x scale is triple width triple height so it;s rendered at 1920 x 1584. Integral selects 1x, 2x, or 3x automatically based on your resolution. If your resolution is 800 x 600 or lower it will select 1x. If your resolution is 800 x 600 or higher but lower than 1680 x 1050 it will select 2x. If your resolution is 1680 x 1050 or higher it will select 3x. Fractional scaling allows it to scale it using fractions instead of integers. So instead of scaling it 1x, 2x, 3x each way it can use a very precise scale like 2.374678657836. This allows it to scale it your exact screen resolution giving it a 1:1 ratio with your screen resolution. However because of this it no longer has an exact ratio with the native fb resolution (1x, 2x, 3x) so certain efb effects won't work properly.

The efb scale is the internal resolution. Meaning the resolution that the scene is actually rendered at.

I should also point out that the numbers I just used are for 4:3 aspect ratio games. The native efb resolution is different for 16:9 games (720 x 528 I believe? not completely sure). Dolphin maintains the native aspect ratio by scaling equally both vertically and horizontally and placing black bars on the sides or top/bottom. Unless you use force 16:9 or stretch to window which stretches the image to get rid of the black bars. Why am I telling you this? Because you have to realize that at a 1920 x 1080 resolution for example dolphin maintains the 4:3 aspect ratio with a 4:3 game, so the usable screen area without stretching is actually only 1440 x 1080 (4:3). In other words 2x scale (1280 x 1056) is slightly lower and it will be stretched slightly and 3x scale (1920 x 1584) is much much larger and it will be scaled down to your resolution. At that resolution fractional would render it at 1440 x 1080 which would be just barely higher than 2x while integral would use a 3x scale. This means fractional should perform better at that resolution.

Fractional yields the best balance between image quality and performance but may break some fb effects while integral will likely have a larger performance hit with no noticeable improvement in visual quality in exchange for proper fb effects.

uhm, plus some corrections of mine:

(12-12-2010, 08:33 PM)NeoBrain Wrote: [ -> ]

(12-12-2010, 11:08 AM)NaturalViolence Wrote: [ -> ]I should also point out that the numbers I just used are for 4:3 aspect ratio games. The native efb resolution is different for 16:9 games (720 x 528 I believe? not completely sure). Dolphin maintains the native aspect ratio by scaling equally both vertically and horizontally and placing black bars on the sides or top/bottom.

fwiw, the native efb resolution is the same for NTSC and PAL; there are multiple ways for a game to handle the differences between NTSC and PAL, they all are done at the EFB->XFB copy stage though: One way is to just add borders at the top and bottom of the copied XFB, the other way is to apply scaling when copying the EFB to the XFB.
Another note: The max XFB dimensions are 640x574, i.e. the height when using PAL (720x574, NTSC uses 720x480) and the width of the EFB. The GX supports stretching on the y-scale when copying the EFB to an XFB, x-scaling (i.e. to 720) can only be performed when sending an XFB to the TV.
(just btw, not that it had anything to do with the original question)

Core i7 860 @ 3,2 ghz with 9600gt @650 core, 1750 shaders and 900 memory.
1x efb scale 200 fps
2x efb scale 125 fps
[color=#FF0000]3x efb scale 68 fps[/color]

Core2duo E6750 @ 2,66 ghz with HD4870 (standard),
1x efb scale 109 fps
2x efb scale 111 fps
[color=#FF0000]3x efb scale 82 fps[/color]

The HD 4870 is undoubtedly a better GPU than the 9600GT judging by the 3x efb scale fps. However, this is a Core i7 compared to a Core 2 Duo series. It's almost like comparing a P4 to a Core 2.

(01-06-2011, 03:41 AM)tuanming Wrote: [ -> ]Core i7 860 @ 3,2 ghz with 9600gt @650 core, 1750 shaders and 900 memory.
1x efb scale 200 fps
2x efb scale 125 fps
[color=#FF0000]3x efb scale 68 fps[/color]

Core2duo E6750 @ 2,66 ghz with HD4870 (standard),
1x efb scale 109 fps
2x efb scale 111 fps
[color=#FF0000]3x efb scale 82 fps[/color]

The HD 4870 is undoubtedly a better GPU than the 9600GT judging by the 3x efb scale fps. However, this is a Core i7 compared to a Core 2 Duo series. It's almost like comparing a P4 to a Core 2.

Which shows exactly how much a good graphic card matters with dolphin in the end, especially with a modern hi res monitor. Now if the core i7 had the HD 4870 and experienced similar fps drop percentage, the fps with 3x efb scale should have been 144, instead they are 68. This is just some of the interesting info you can get from such benchmarks.

Quote:Took me ages to find it, but here you are

Thank you!

Quote:Which shows exactly how much a good graphic card matters with dolphin in the end, especially with a modern hi res monitor.

Well you've clearly managed to find an area in a game where cpu emulation is lightning fast, thus allowing you to do such a benchmark without having to worry to much about emulated cpu bottlenecks. But this is extremely rare if you look at all games. I'de also like to point out that so far nobody with a decent video card (no offense to the guy with an 8600gt) has received less than 60 fps with a 3x efb scale. So if anything I believe this actually reinforces the idea that dolphin DOESN'T need a top of the line video card. At least that's how I interpreted the data so far.

(01-06-2011, 07:49 AM)NaturalViolence Wrote: [ -> ]

Quote:Took me ages to find it, but here you are

Thank you!

Quote:Which shows exactly how much a good graphic card matters with dolphin in the end, especially with a modern hi res monitor.

Well you've clearly managed to find an area in a game where cpu emulation is lightning fast, thus allowing you to do such a benchmark without having to worry to much about emulated cpu bottlenecks. But this is extremely rare if you look at all games. I'de also like to point out that so far nobody with a decent video card (no offense to the guy with an 8600gt) has received less than 60 fps with a 3x efb scale. So if anything I believe this actually reinforces the idea that dolphin DOESN'T need a top of the line video card. At least that's how I interpreted the data so far.

Many people nowdays use hi res monitors (they are dirt cheap and good quality overall), it is good to know from which point onward the gpu isn't a constraining factor to the cpu. And none of these cards managed to achieve it actually. An efb scale of 3x is basically used by default in normal builds when someone has such a monitor (efb set to integral), or when fractional produces glitches and thus integral is used. Overclocking the cpu is something many use yet it is risky and not beneficial in the long run. Getting a proper gpu and not overclock like crazy your cpu when it is not needed really is something that can make a world of difference to many.

For example you can hear all over the place that Mario Galaxy 2 is demanding and you need a highly clocked cpu, where in reality it is a gpu demanding game (there are a lot of other games that are way more cpu demanding than Mario Galaxy 2 actually). Some hard numbers can help in such cases.

Quote:And none of these cards managed to achieve it actually.

I would disagree with that. All of these cards except the 8600gt manage to pull over 60 fps with a 3x efb scale. And bottlenecks from memory bandwidth and cpu throughput vary a lot. You are testing one of the few places where not a lot of cpu throughput is needed, normally achieving 200 fps with framelimiting off is not achievable for most heavier games regardless of how powerful your gpu is.

Quote:For example you can hear all over the place that Mario Galaxy 2 is demanding and you need a highly clocked cpu, where in reality it is a gpu demanding game (there are a lot of other games that are way more cpu demanding than Mario Galaxy 2 actually).

I would be interested to see you prove this. I have tested SMG2 pretty thoroughly and my results do not line up with that.

Out of the games I have:

TP: Usually bottlenecked by memory bandwidth.
WW: CPU heavy
SMG2: CPU heavy and GPU heavy but especially CPU.
NSMB: Not very heavy on either.
DKCR: CPU heavy
MP: GPU heavy and CPU heavy, depends on the area.

I really don't think you can say that with any real certainty. What would need to be done for a real benchmarking thread would be this:

1) Run WW on the same processor, same video card and change the ram to see if there is any difference.

2) Run WW with the same processor, same ram and switch out video cards to see if there is any difference.

3) Run WW with the same video card, same ram and switch out processors to see if there is any difference.

The point would be to keep most of the system exactly the same and change each part separately to see what made the biggest impact.