Every decision in a computer design

Update:26-10-2018
Summary:

Every decision in a computer design is a trade-off betw […]

Every decision in a computer design is a trade-off between cost and capability.

Some computers have a dedicated and separate GPU for graphics, but cheaper ones just incorporate GPU functions into the CPU. Lower performance, but cheaper!

Some computers have complex, dedicated audio processors and sound cards. Others incorporate more primitive sound functions into support chips of the CPU. Lower performance, but cheaper!

Some computers used to have dedicated encryption processors. Modern Intel CPUs have AES hardware acceleration built in.

In Ye Olde Days, you would actually buy a separate math co-processor chip for FPU functions (80387, anyone?) Every modern Intel CPU now had an integrated FPU.

Even if you had the coprocessor, you’d need computer monitor arm Suppliers  a programming kit to utilize it, and a critical mass of install base to convince programmers to spend time on it. In GPGPU land, the OpenCL and CUDA architectures emerged for programmers to use GPUs as a co-processor, but even that is mostly limited to specialized applications that take advantage of the GPU SIMD architecture.

There no special advantage to a general purpose ARM chip as a co-processor have an immediate benefit to a modern Intel CPU. It’s a different architecture and slower. It’s not clear that it would be useful even if it was free in every PC since it would call for specialized programming like OpenCL or CUDA.

You’re more likely to see FPGA coprocessors for specialized work like AI and ML until general purpose CPUs accrue the capability themselves. New Azure servers to pack Intel FPGAs as Microsoft ARM-lessly embraces Xeon