Microprocessors: Should the teaching of Intel 8085 in university computer architecture courses be replaced with RISC architectures like MIPS?

In considering my answer to this question, I've failed to think of any situation where it would be reasonable for a university computer architecture course being given in the 2010s to be teaching only a computer architecture developed in the 1970s (8085). I think a university course should teach fundamental principles and how they can be applied. This means theory and practice should go together. In computer architecture this means finding real-world architectures which demonstrate the application of various principles. Can this be done using a single architecture? No. Consider the 8085 - no cache. Consider MIPS - no hardware page table walk. Certainly in the context of a computer science degree, diversity is a must. But what about a course teaching the subject to students not studying computer science - for example, electronic engineering, or robotics? I think such students needs to know about modern computer architecture. Modern processors - even those used in basic systems - are no longer simple. System designers and programmers need to understand how to work with caches, pipelined and out-of-order execution, memory management, multiprocessing, etc. etc. These cannot be taught by reference only to legacy architectures. So, in conclusion, drop the 8085 - take on board some modern processors. And if insist on relevance to the real-world, then you need to include at least one ARM.

CMU made the switch: http://www.ece.cmu.edu/~ece447/s13/lib/exe/fetch.php?media=lab1.pdf

As much as I hate them, a car analogy might be to first look at an aircooled VW Beetle petrol engine before looking at a watercooled, turbo, injected, ECU'd, diesel VW TDI engine. You can grasp the first one mentally, and then mentally add the other bits to build up a more complete picture. Pro: - You could possibly still understand all of the 8085 - It is, at a bit of a stretch, the basis for much of the x86 / x64 architecture that a lot of the world runs on these days Con: - It's ancient and mostly irrelevant for todays age - You could equally well fully understand an early ARM and see it evolve - No cache / pipelines - Single core Roughly 20-25 years ago my course used 68k processors. We also looked at x86, MIPS, Alpha and PowerPC and we used Sun Sparcs for other bits of the course. The main two these days I would use are x64 (not x86) and a recent ARM. It doesn't hurt to look at something older so you can get the first principles, but it cannot be the only thing you see.