I did a monitoring comparison of CPU temperature/behavior depending on firmware (stocks 101k4 and 104k4), here are the results.
Both screenshots are taken after ~1hour XBMC streaming SD to have something "stable". I use Android Tuner Pro for monitoring.
104k4:
Nice "stairway" effect, temperature rises regularly until it reaches 80°C then it immediately fall around 50°C and the cycle starts again. Depending on temp, 2/3 cores are active between 696Mhz and 1200Mhz. Any advanced multitasking causes global lag, especially near the peaks. I noticed the same behavior in Finless 1.6 ROM.
101k4:
Temperature never reach 65°C, 4 cores are active @1992Mhz (Note that I soft-forced this maximum frequency with no problem, what I can't do in 104k4 which is again stuck at 1.6Ghz max). Not only temperature is stable, whole system is.
I didn't try f/w 102/103 but I wonder what they've done between (did Eric talked about better handling on CPU, huh?)... Can we consider that temp monitoring is reliable? What is sure is that CPU behavior depends on it and throttles like crazy for no good reason.
Also notice that we now have only 5°C steps where we had 1°C steps before.
Just a word about Antutu benchmark scores (I don't really care but it's mostly a good indicator). I use Antutu X because Antutu is definitely cheated and gives a result of 31700 whatever is firmware and context. Not credible at all.
101k4 gives a consistent score of about 23000 whereas 104k4 has random various results between 9000 and 19000. That tends to confirm the fickleness of latest f/w.
No need to say that being back to 101k4 I will keep using it until something really serious and well-tried come (from here I hope

Comment