Intel’s MKL + their compiler + their CPUs with the latest SIMD extensions is still one of the best stacks for Scientific programming. Get a scientist to write s

Intel’s MKL + their compiler + their CPUs with the latest SIMD extensions is still one of the best stacks for Scientific programming. Get a scientist to write some single-core kernel in C or C++ language, and that compiler will work serious miracles on code that needs to scale up to hundreds if not thousands of machines (think supercomputers). The game only changed here recently when Nvidia went all in on the server GPU offerings in the past few years, but writing fast CUDA can get much messier than your average biologist would care for. Intel has more software engineers that AMD has employees, so their support for their math libraries, compilers, their GPU drivers, their benchmarking tools, and their contributions to the LLVM ecosystem, the Linux kernel, and other projects have helped them get an iron grip on academic/scientific side of high-performance computing.