jay_nar2012 :
So you can't change instructions after the CPU is made basically as its part of the architecture..
How about if it was given more instructions when it was designed?
Then it would have been yet another architecture, a PowerPC variant.
To picture the this, imagine one processor was built with basic operations in mind. It had the hardware to do simple add/subtract operations in one cycle. It understands add/subtract instructions. If you want to multiply, you have issue a lot of add instructions.
A second processor, built for more advanced math, has the hardware for add/subtract/multiply/divide. It takes more transistors to create this hardware, but it now can understand add/subtract/multiply/divide instructions. If you have to multiply, it can do it in one cycle.
So basically, a purpose-built processor has the potential to be a lot more efficient. On the other hand, a lot of improvements are already explored on current architectures, meaning you have to be really creative to gain efficiency.
To try and brings this back on topic, the Xenon CPU (x360) has 165 milion transistors, a Prescott family Pentium 4 had 125 milion, a Conroe Core 2 Duo had 290 milion. The Xenon has the advantage of being purpose-built, so it might be just as fast when runnig the software it was made to run. A direct comparison cannot be made since one cannot run the code compiled for the other.