Poor delroth. He now has to edit 5 Pages of Benchmarks into the file
Uh... I'm a bit confused by the "Performance per MHz". My math showed that Steamroller is faster than Vishera per-GHz but slower than Nehalem, and yet the quotient is shows exact opposite even though Haswell follows what I'd expect.
A10-7850K @ 4.5GHz = 11min 31sec
FX-8350 @ 5GHz = 10min 43sec
11 (min) * 60 + 31 (sec) = 691 sec
691 * 4.5 (GHz) = 3109.5 "reverse IPC"
10 (min) * 60 + 43 (sec) = 643 sec
643 * 5 (GHz) = 3215 "reverse IPC"
3215 / 3109.5 = 1.033928284290079
Therefore Kaveri has 1.034x the performance per-MHz than Vishera.
We can also do Nehalem...
11 (min) * 60 + 39 (sec) = 699 sec
699 * 3.9 (GHz) = 2726.1 "reverse IPC"
3109.5 / 2726.1 = 1.140640475404424
Therefore Nehalem is 1.14x faster per MHz than Kaveri.
Delroth stopped accepting OCed CPU's. Set your CPU/RAM back to stock if you still want to post results.
Well crap, there's goes most of the results I got from overclock.net, including that Kaveri score.
EDIT: ...and yet the first post still contains the following:
Quote:(put "OC" in there if it's overclocked)
(01-19-2014, 08:05 PM)Nintendo Maniac 64 Wrote: [ -> ]Well crap, there's goes most of the results I got from overclock.net, including that Kaveri score.
EDIT: ...and yet the first post still contains the following:
Quote:(put "OC" in there if it's overclocked)
delroth told us to create our own chart if we want to maintain OC'ed results.
It isn't a blessing if it isn't created by God himself :/
(01-19-2014, 08:05 PM)Nintendo Maniac 64 Wrote: [ -> ]Well crap, there's goes most of the results I got from overclock.net, including that Kaveri score.
If you are not bored you can always make your own spreadsheet with those values... The bench needs more cpu variety than it needs oc values imo. Valueable info will be difficult to track if for example 100 people posted their overclocked 4670 values making the bench less usefull.
(01-19-2014, 08:25 PM)Link_to_the_past Wrote: [ -> ]If you are not bored you can always make your own spreadsheet with those values... The bench needs more cpu variety than it needs oc values imo. Valueable info will be difficult to track if for example 100 people posted their overclocked 4670 values making the bench less usefull.
I understand that, it's just that it can kind of be a pain to get people to re-bench, especially on OCN where bragging rights through benching overclocks is kind of a thing.
EDIT: Uh... why is there even still two separate columns for stock clocks and measured clock? Not to mention the column that says "OC/UC?"...
More than 40 minutes have passed and it still shows this screen:
![[Image: sfdN253.png?1]](http://i.imgur.com/sfdN253.png?1)
I exited Dolphin, as it seemed to take forever. May it be related to my problems with Twilight Princess in recent builds?
(01-19-2014, 02:08 PM)DatKid20 Wrote: [ -> ]I don't think Android can run .elf files right now. 
What a shame.
It could, its the same black magic running below the surface as in Desktop Dolphin. However, it seems the benchmark uses stuff yet unsupported by the ARM Interpreter and JITs, so the thing immediately stops (I posted this earlier on after trying it on my Note 2014; along with the info that current buildbot APKs are broken).