Originally Posted by ardatdat
An updated version without autogroup scheduling is updated in post #1.
Also, the ODIN3 flash-able version is also uploaded.
Please test whether v1.3A is more stable.
It seems to be better
, It's more stable! differences are still present though.
Few quadrant tests in a row: (max 1350 interactive)
3300, 3600, 4057, 4297
, 3660, 1653
- variations are still significant. Two extreme values. I've never had that high score but...never that low either. (With original frequencies (stock kernel) it was the interval 2890-3440, Thus standard deviation is here noticeably higher) It's just quadrant. The most popular/basic benchmark. Its scores say nothing about general performance but...these relatively big differences among random scores are interesting.
Linpack (single thread, MFLOPS):
28, 38, 52, 53, 32, 27, 49, 34, 21, 54 - a lot of extreme values
Antutu (v. 2.5)
6066, 6046, 6189 - here is quite stable, (variations depend mostly on reading/writing speed of my SD (which is crappy))
CF - very stable, never goes under 7100!
Smartbench 2011/2012. It's still significantly worse than on your november version especially in terms of gaming. We have always had great results in this index (far better than sgs II). Now it's at least 300-400 units less. But no big variations noticed here. Scores are just lower. (So maybe it's just my handset that performs so)