You may not know every detail of every computer, but if you can't build your own computer at home, then you should figure out how. It's doable in less than a semester, and you'll be happy you did.
Well, I can order an iMac from Apple, plug in the keyboard and the mouse, and say "I build a computer". Or I start with a shovel of sand and make my own silicon. I don't know how many computer users understand flip-flops, or nand-gates, or FETs. I think some of that should be in the general understanding of science. And for computer scientists, knowing about things like cache size and organisation, register set, and ALU capabilities does make a significant difference if coding for performance. I was once bitten during the transition from SPARC to Intel when I found out that SPARC's fantastically thought out register windows (meant to sped up function calls) actually slowed down function calls once your calling depth became great enough that you ran through the register file and had to start putting those large register windows on the stack. Register-starved Intel did better in that case. So benchmarking on SPARC indicated "no recursion", while in Intel recursion was actually faster than iteration with a dedicated stack in software.