4 ms·
If you already broke compatibility, might as well go all the way and optimize for your problem domain, such as graphics/AI/low power. And in fact GPUs/TPUs/ARM
by cat_plus_plus 4y ago
If you already broke compatibility, might as well go all the way and optimize for your problem domain, such as graphics/AI/low power. And in fact GPUs/TPUs/ARM are doing very well, and there is a strong interest in quantum computing. So this leaves conventional high power, non-vector, core speed over concurrency computing, where you additionally expect existing operating systems and software ported and optimized for your architecture. Plus IBM Power PC mainframes and Apple ARM laptops don't fit the bill. There might just not be enough demand for such a product right now and if there is in future, someone will offer a solution?