Modern Computer Architecture Rafiquzzaman Pdf 23 - What You Need to Know
Modern hardware supports some kinds of parallelism, notably multi-core CPUs, which have a number of processing cores. The competition between manufacturers is to make the fastest chips possible, and the most powerful machines contain many cores. The processing speed of modern processors can be measured in MIPS: million instructions per second, or GFLOPS: giga floating-point operations per second.
Modern Computer Architecture Rafiquzzaman Pdf 23
An individual processor is often identified as a core. All modern CPUs have a uniform number of cores, and the terminology core-duo, core-quad, core-octa, core-deca and the like are just the numbers of cores in the processor. The processing power of the computer can be compared with the number of cores in one processor. A common configuration for commercial devices is 64 cores in one processor. The increase in the number of cores affects the capacity of the system, since the number of devices and memory in the system must be increased.
Modern software supports parallel computation, and so can run on multiple devices at once. The fastest supercomputers take more than 10,000 MIPS: for a large computer, which contains hundreds of cores, this is measured in hundreds of thousands of MIPS.
Despite the very powerful hardware for multi-core CPUs, the single device situation means that there is very little processing speed when dealing with many clients at once. The performance of a multi-core system is much more dependent on communication than on the processing power. In fact, for a huge computer, the number of clients will be greater than the number of cores in the CPU.