Well, Conky is more unstable than cpumem for some reason. e.g., I have seen Conky report 20% load while cpuem said 100 and since it was running apt--

Sorry, Fennec interpreted a space as send.

Where was I?

Ah, Conky isn't the most reliable out there. If cpumem says it's cool, it's cool.

Let us know how lshal claims the battery has been (ab) used and see how it's faring. Frankly, locking the CPU at 50% should give you about 2 hours autonomy (or so) and warm the device.

If you don't see cpumem show you 50%, it's probably an artifact from batterygraph or another program that gets tricked like Conky.

From experience, background CPU is noticeable. Leaving MC do a copy over wifi depleted my battery in 1.5 hours (wasn't full). Now, the CPU was not 50%, granted, but it's not like load is a direct indication of consumption. I overclock as a matter of life and I know this to be true.

So, AFAICT, your CPU is not locked to 50%. Besides, it's reasonably hard to lock a single core to 50%.

How much do you get out of your battery, full to empty, on light and heavy usage? (light: say 20 minutes call, 1 hour browse via wifi, 2/5 backlight, 2G, plus a little fiddle).

Also, if you charge at night to 95+, how much is left in the morning? My 8 hour standby eats between 10 and 15% (connected to wifi via WPA, BT, 3G).