"Best" is a nice word. Best with respect to what ?
This job is normally solved using a microcontroller. And it is more about the availability of appropriate peripherals (mainly ADCs with good resolution) than about 'brute computing power' - at least regarding MPPT, which is a rather slow process.
As it is, this job is normally done by the same microcontroller that is also implementing the inverter function - which will really require some 'computing power'.
This week I had a thread in a forum regarding the use of a dsPIC33 device (from www.microchip.com), but there is quite a number of microcontrollers around that are easily capable to do the job.
'time response' is difficult - as well as 'efficiency'.
To get new values for MPPT, it is necessary to deviate from the current setpoint - normally meaning to temporarily have lower efficiency. Thus you will not want to do this too often.
By 'efficiency' I guess you are meaning "precision of finding the MPP" - which relies more on 'good' peripherals than on the 'device' itself.
Anyway, most of MPPT 'quality' is about the control strategy - not so much about the device. For me, quite a number of microcontrollers or digital signal controllers would work well one way or the other.
As U. Dreher points out, the question is rather vague.
Myself and a colleague built a MPPT with a MBED-LPC1768 development board that we had on our desk and had it working with some PV panels in a couple of hours.
I agree with U. Dreher that many microcontrollers should be able to do the job. There is indeed a great deal of performance to be gained from the algorithm itself and by the choice of components in the hardware.
We just used an off-the-shelf buck converter and controlled it using PWM. Obviously the converters are quite specific to the overall system specifications.