Two types of CPU are currently popular.
At the moment RISC CPUs are faster and cheaper to make than CISC CPUs.
CISCs are currently not far behind in performance and are more popular
because of compatability with previous CPUs.
CISC - Complex Instruction Set Computer.
Make the CPU more powerful by adding more complex instructions. e.g.
RISC - Reduced Instruction Set Computer.
Make the CPU more powerful by having simpler instructions that can
be executed faster. e.g. PowerPC
Making the CPU faster.
How to measure the speed of a computer?
At present, improvements in technology make the clock rates
about twice as fast every two years. This is mainly due to increased miniturisation
- Clock 10,20,40,80,160,320 MHz.
Million Instructions Per Second (MIPS)
Synthetic - test programs
Real - real applications
If fact CPUs are getting twice as fast every 1.5 years. The reason for
this, is improvements in Computer Architecture.
These techniques allow the CPU to perform operations in parallel - parallel
processing or parallelism.
Allow more than one part of the instruction cycle to be performed at
once - like a production line. It is quite common to have 7 different things
happening at once in the CPU. Here is a non
pipelined CPU for the logic simulator and here
is the same program running on a pipelined CPU- it can have 5 parts of
an instruction being executed at once.
Have duplicate hardware inside the CPU to allow more than one instruction
to execute at once. For example a CPU may have 2 ALUs to allow 2 ALU instructions
to execute at once.
When parallelism is applied at the level of the workings of the CPU
it is called fine grained parallelism.
There is a limit to how fast a computer can be made by increased miniturisation
and fine grained parallelism. We have not yet reached that point but many
people think we soon will. In order to make machines faster still coarse
grained parallelism is necessary. This type of parallel processing
involves using multiple CPUs in one computer.
SMP - Symmetric MultiProcessing
Use a single memory with multiple CPUs. This is easy to implement but
the speed of the memory will limit the maximum number of CPUs. Currently
SMP systems are available with up to 8 CPUs.
MIMD - Multiple Instruction Multiple Data
Use a large number of CPUs each with their own memory and communication
links to other CPUs. This machine needs programs to be rewritten and there
may be problems due to the speed of the communication links.
Summary of topics covered so far
Addition and Subtraction
Combinational and Sequential Logic