What’s a good way to learn how CPUs, GPUs, and memory work?

For a hands-on route, try NAND2Tetris. It builds up from basic logic components toward a working computer and software, so you can connect the layers instead of just memorizing terms. It takes commitment, but you can work through it in stages.

For the bridge between architecture and programming, Computer Systems: A Programmer’s Perspective is a strong fit. It covers how programs interact with memory and processors, with useful context for caching and performance. Pair the reading with small experiments, like measuring how access patterns change runtime.

If you want to see the hardware side more systematically, Digital Design and Computer Architecture is a reasonable next step. It introduces the building blocks and processor design without requiring you to start with advanced architecture research. Some sections may feel more hardware-focused than you need, so it’s fine to read selectively.

I’d start with a broad introductory computer science series, then move on to a more structured course or textbook. Getting the big picture first helps terms like instructions, memory, and logic gates make sense before you get into implementation details. NAND2Tetris is another option if you learn best by building things.

For the GPU question, keep in mind that a GPU is not automatically faster. It tends to help when a workload has lots of similar operations that can run in parallel, while data movement and small workloads can limit the gain. Learn the basics of memory hierarchy and parallelism first, then use benchmarks on the kinds of tasks you actually run.