The Legend of Zelda: The Wind waker CPU Benchmark

[NaturalViolence]

@Nintendo Maniac 64

I could be wrong but I really doubt it's the L3 cache:

[Venomx1]

redoing Save 1 now @4.5 and the lowest i see the fps get down to is 88.

going redo the other 2 at 4.5 first.

[Nintendo Maniac 64]

(02-21-2013, 05:07 AM)NaturalViolence Wrote: I could be wrong but I really doubt it's the L3 cache

Um... isn't that a bench of the cache speed, not the cache amount? The only difference AFAIK cache-wise between an FX-6300 and FX-4300 is the cache amount, and seeing how the 4300 has a higher clockspeed, shouldn't that automatically mean it's the cache amount that's the cause of the 6300's faster performance?

[Venomx1]

heres my updated 4.5ghz scores i just did them right now.

test 1


test 2



test 3



i took the first one after the Boat Went around. Second and third one i took right away after the fps settled down (takes a second or 2).

got basically the same scores. cept #2 which was 2 fps lower... margin of error range. but whatever.

[NaturalViolence]

Quote:Um... isn't that a bench of the cache speed, not the cache amount? The only difference AFAIK cache-wise between an FX-6300 and FX-4300 is the cache amount, and seeing how the 4300 has a higher clockspeed, shouldn't that automatically mean the cache amount is the source of extra performance?

It's a benchmark of the hit rates of the L2 cache. It measures the % of outstanding memory operations (data requests by the cpu) received by the L2 cache controllers that were completed without moving to the next cache level. If the hit rates are low L3 is being frequently accessed. If the hit rates are high L3 is being infrequently accessed. Bulldozer has an enormous 2MB L2 cache in each module with 16 way association. This gives its L2 cache enormous hit rates compared to any other architecture.

[Garteal]

Venomx1 Wrote:except that guy that somehow outdid everybody else with his 4.4ghz 3570k. Whos is that? what Speed of ram and timings was he running. maybe an error?

Damn. I just noticed. Axxer accidently swapped the pictures of the saves in his post. I'll update the charts.

Quote:heres my updated 4.5ghz scores i just did them right now.

Thanks for the update!
So it seems there's barely any changes. Thanks for confirming it.
Interestingly enough, my 2500K still outdoes all the 3570K's except for the 5GHz clocked one.

[Starscream]

I redid the 1st and 3rd tests without the boats in the shot and I ended up with 33/66/110 on the 1st save and the 3rd was basically the same. I was at 2.3GHz still.

[Nintendo Maniac 64]

Again, I would just figure that performance discrepancies like this is due to a difference in OS configuration, services, and other background stuff. I mean, ram speed doesn't effect Dolphin performance much at all, does it? Maybe that's something else to bench.

I mean, if you wanted to make absolute sure the benchmarking environment was the same, then we'd need a custom-made live Linux distro specifically for Dolphin-benching or something.


EDIT: Also, couldn't you have different results due to just screen-shotting at different points? I mean, if one guy screenshots a maximum framerate on a slower CPU and the other guy screenshots a minimum on a faster CPU, you could easily end up with results showing the slower CPU being faster.

Maybe I was right to follow my OCD and take separate screenshots of minimum and maximum framerates...

[Garteal]

Starscream Wrote:I redid the 1st and 3rd tests without the boats...

Thanks for confirming it.

Nintendo Maniac 64 Wrote:EDIT: Also, couldn't you have different results due to just screen-shotting at different points? I mean, if one guy screenshots a maximum framerate on a slower CPU and the other guy screenshots a minimum on a faster CPU, you could easily end up with results showing the slower CPU being faster.

That's probably what happened. You'll also notice that a 200MHz difference shouldn't yield that much difference in performance.
The 2500K/2600K and 3570K's at 4.5GHz score roughly the same.

Quote:Maybe I was right to follow my OCD and take separate screenshots of minimum and maximum framerates...

Yeah, that'll give us a better average. (I took the average of your scores btw)
Although it'll probably be better to use the FPS dump option in Dolphin to measure the averages instead of trying to eyeball it.

[Venomx1]

Garteal my friends 2500k acts the same way. except we were comparing with 3D mark vantage. my 4.5 vs his 4.8 i lost by like 50 score.

also will redo 4.8 & 5ghz still if you want. next time i redo them im going OC my ram from 1333 to 1600 with custom timings. hopefully ill see the same gains as i did in other benchmarks