Dolphin, the GameCube and Wii emulator - Forums

Asus Eee PC 1215B
OS: Windows 8 Professional, 64-Bit
CPU: AMD E-450, Dual-Core, 1,65 GHz
GPU: AMD HD6320 (integrated)
RAM: 6GB DDR3-1333

Dolphin 3.0 - 64-bit. PAL.

No changes. Resolution 1366x768

[attachment=9082]

[attachment=9083]

[attachment=9084]

FPS/VPS/Speed

12/21/42%
10/17/34%
17/28/56%

Figuered you guys already have enough 2500Ks. At least it beats the Atom.

Thank god we finally have bobcat results. I was about to borrow a netbook from a friend just to test this so you saved me some time. I went ahead and calculated the IPC relative to trinity.

(35/1.65)/(156/3.8) = .52
(29/1.65)/(138/3.8) = .48
(47/1.65)/(208/3.8) = .52

So 0.48-0.52

For reference here are some of the other architectures:
Ivy bridge: 1.55-1.65
Sandy bridge: 1.40-1.50
Phenom II: 1.12-1.12
Athlon X2 Brisbane: 0.70-0.74

And for fun here is atom:
(28/1.60)/(156/3.8) = 0.43
(21/1.60)/(138/3.8) = 0.36
(40/1.60)/(208/3.8) = 0.46

So 0.36 - 0.46

Based on the bobcat IPC we can expect jaguar to still be significantly behind the later iterations of K8 if it makes the 10-15% IPC improvements that AMD is claiming. Currently Nintendo Maniac 64s K8 results show a nearly 29% increase in IPC over bobcat. However dolphin could actually see a higher increase than 10-15% since dolphin makes heavy use of 128 bit SSE2 operations. And jaguar is widening the FPU pipeline from 64 bit to 128 bit. Those 10-15% figures are based on average applications.

Thought I would throw in a few more systems.

OS: Windows 7 Home Premium x86 SP1
CPU: Core 2 Duo e4600 @2.4 Ghz (Stock)
GPU: AMD HD 5570 1GB DDR3
RAM: 2 GB DDR2 400 (running Synchronously)
Game: NTSC Dolphin x86 from OP.

Save 1 28/58/96


Save 2 23/46/76


Save 3 32/63/105


Honestly I expected a little better; I thought it would be above 30 for each save.
Now for the same system but overclocked.

OS: Windows 7 Home Premium x86 SP1
CPU: Core 2 Duo e4600 @3 Ghz (OC)
GPU: AMD HD 5570 1GB DDR3
RAM: 2 GB DDR2 500 (running Synchronously)
Game: NTSC Dolphin x86 from OP.

Save 1 38/77/128


Save 2 30/60/100


Save 3 46/93/155


Pretty decent performance boost for free. No vcore boost and only a very slight thermal increase. I did push it to 3.2 Ghz but there was no noticeable performance increase. However it required a vcore boost and had a very noticeable thermal increase.

Ok last system.

OS: Windows 7 Home Premium x64 SP1
CPU: Phenom II x2 511 @3.4 Ghz(Stock, really an Athlon II x2 270 branded a Phenom II for OEMs)
GPU: Zotac GT 240 Amp! edition 512MB GDDR5
RAM: 4 GB DDR2 800
Game: NTSC Dolphin x64 from OP.

Save 1 44/88/146


Save 2 35/70/116


Save 3 52/105/175


That just about wraps up all the systems I have access to.

Just noticed that the atom results were NTSC and the bobcat results were PAL which makes my previous comparison invalid. I'll have to redo my calculations and update my post.

Also x86 results will need to be seperated from x64 results in additional to NTSC and PAL being seperated. So we need 4 categories, NTSC x86, PAL x86, NTSC x64, and PAL X64.

Edit: My previous post has been updated with the gamespeeds converted to NTSC equivalent speeds.

@NV - WW PAL runs at 30 FPS, just like NTSC version. The PAL version runs 30/50 (50 in rare instances, such as when the Nintendo logo appears, until the intro screen). I assume the NTSC version runs 30/60 as well. For the benchmark though, they should be running at the same speeds.

Operating System: Microsoft Windows 7 Ultimate, 64-bit
Physical RAM: 8191 MB
CPU name: AMD Phenom™ II X4 955 Processor
CPU speed: 3.20 GHz (4 logical threads)
GPU: Nvidia GeForce GT 430 1GB

Dolphin 3.0 - 64-bit. NTSC.

No settings were altered.
_______________
Save 1:

40/80/133

Save 2:

38/75/125

Save 3:

48/97/161

____
Note: I hope I'm still allowed to submit my results...

Shonumi Wrote:The PAL version runs 30/50

What is this black magic?

How can you draw 3 frames with 5 refresh periods without using alternating intervals?

Shonumi Wrote:I assume the NTSC version runs 30/60 as well.

It does.

Shonumi Wrote:For the benchmark though, they should be running at the same speeds.

Check yo math dawg.

From the bobcat post:

Benze Wrote:12/21/42%
10/17/34%
17/28/56%

Now ask yourself without even doing the calculations. Is 21 42% of 60? No. But it is exactly 42% of 50. The game is clearly running at 50Hz.

If you compare based on framerate or vps you can mix NTSC and PAL results. If you compare based on gamespeed (which my IPC calculations are based off of) they're not going to be equivalent.

I withdraw my sentence about the 4 categories. Just seperate x86 and x64 and it should be fine as long as your use fps for the comparison.

Now I have to go back and double check all of my IPC calculations. Ugh.

Edit: All of the others were NTSC. Thank god.

(04-15-2013, 07:16 AM)NaturalViolence Wrote: [ -> ]What is this black magic?

Classic Nintendo EAD; leave it to them to throw you oddballs.

(04-15-2013, 07:16 AM)NaturalViolence Wrote: [ -> ]How can you draw 3 frames with 5 refresh periods without using alternating intervals?

How should I know? :p Graphics aren't my specialty. Dolphin could simply be detecting something incorrectly. But my version is PAL, and it's always ran @ 30 FPS without ever looking like it was running 5 frames faster than it should.

(04-15-2013, 07:16 AM)NaturalViolence Wrote: [ -> ]If you compare based on framerate or vps you can mix NTSC and PAL results. If you compare based on gamespeed (which my IPC calculations are based off of) they're not going to be equivalent.

I didn't know you were calculating based on reported gamespeed; I thought everything here was going to be based solely off of FPS, so I wrongly assumed you were too (my bad, I hate assumptions). I didn't read your posts thoroughly since I'm not terribly interested in IPC differences between microarchitectures. Though I did want to see your surprise about PAL and NTSC versions having the same framerate, at least in Dolphin

Is that the default idle speed? it seems too fast to me. That was from the download in the first post.

http://youtu.be/IgzFujqdRzs

30 fps is fullspeed for this game. Any faster than that is above fullspeed. Since most systems can run this game above fullspeed this benchmark was set to turn off framelimit so that we could record the speed that the system was actually capable of.