The bottom line here is that. No matter how many cores you got if they are poor implemented the performance and the autonomy will be bad.
I'd totally agree that Samsung does an incredible job tuning the s/w for the h/w from a sheer performance perspective. Look at phones from HTC and Samsung that used identical S3 chips. Samsung's phones always performed better. But if you're talking about battery performance (vs. battery life) both S4 and Teg3 One X's beat the SGS3 on web browsing and 3G calling; the S4 pretty handily. SAMOLED kills on video playback but that's not about the SoC or s/w, that's about lower energy draw on dark colors.
3G Calling (adjusted for battery size)
One X [S4] - (12:27 Hours) 10:35 Hours
One X ([Teg3] - (11:54 Hours) 9:57 Hours
SGS3 - 10:20 Hours
Web Browsing (adjusted for battery size)
One X [S4]- (6:25 Hours) 5:03 Hours
One X [Teg3] - (5:39 Hours) 4:18 Hours
SGS3 - 5:17 Hours
So from a battery consumption stand point the SGS3 draws more power than both S4 and Teg3 versions of the One X on a standardized test for those two functions. I'm assuming that GSMArena left any s/w installed by the manufacturer preset to "on" still running during the tests. In the SGS3's case you could probably get better battery life by disabling some of the non-essential features like S-Voice always listening. So either Exynos is using more power than the other two chips or what Samsung's running in the background is tapping the battery.
This is of course useless in real life as the One X has a fixed battery that’s 16% smaller than the SGS3’s so the SGS3 will still win in battery life. But whether from HTC's tuning or the performance of S4 and Teg3 both One X's do well at battery consumption. It’ll be interesting to see what peoples real-world experiences are once the SGS3's released.