Discussion on question by wav scientist: Why not make one big CPU core?

How would you propose making 1 core equivalent to 8? You can't just bump up the clock frequency.

What "more stuff" do you want to do? You either have to go parallel to do "more stuff" (i.e. more CPU cores), go more complex instruction set (more transistors yes, but also more heat and slower), or increase the clock speed (power scales to the square of frequency)

Distance from one edge of a big CPU to the other end is a big issue. If you need to transfer data between them synchronously, that will set the limit for your clock speed. Take a look at AMDs HyperTransport for details in how complex this is.

"but the benefit of more cores becomes really poor past 16 cores yet you have these 64 core workstations" - Performance depends on the application. People have been using 256 core (and more) systems to do real work for decades now.

Hmm I wonder why subways or large buildings have more than one door, it just doesnt make any sense. Highways should have only one lane per direction, why would you need more than one.

I don't think this question is very well formulated. What architecture are you imagining when you say "one big core"? A CPU core isn't like a hamburger - you don't get more just by making it bigger. You get more by making it smaller.

@J... I think it's pretty clear that they mean a wider superscalar core. (With more execution units and a wider front-end, too.) The OP probably imagines that widening the pipeline by 8x would lead to an 8x gain in instructions per cycle. "Big" fits this description perfectly. We already talk about big cores (e.g. mainstream x86 like Skylake or Ryzen with wide SIMD and lots of cache) vs. small cores (e.g. the many cores on a Xeon Phi card, or low-power CPUs). Or even in the name of big.LITTLE heterogeneous ARM multi-core chips.

@J...: I disagree. The body of the question is pretty explicit about what they have in mind. About as clear as it's possible to be without understanding that making a superscalar CPU wider doesn't scale it's IPC linearly (which is a key part of the answer to the question). Instead of splitting up the execution resources between multiple cores, why not put all those execution units in one big core? It's a fair question with an interesting answer. (But strangely they accepted an answer that doesn't even mention IPC or wider superscalar, just SIMD.)

"I think if they made..." what expertise do you have to back up such an assertion?

I think my answer is pretty good, and so it @pjc50's. They both directly address your misunderstanding that making a CPU wider doesn't scale IPC linearly. (And mine explains why even just making it 8x wider is totally implausible). You can change your accept vote if you want. I'd 100% recommend reading lighterra.com/papers/modernmicroprocessors for the background to understand the answers. Anyway, I get it; if you had already known more, you wouldn't have asked the question in the first place. Tom's answer "sounds" right, but it mixes up things and ignores superscalar IPC.

Games, for example, almost never use more than four cores. Citation needed. And same could be said about two cores and a single cores at different points in time. Games are optimized for the most common HW out there.For the longest time the market was dominated by the quad-core CPUs but it's quickly changing now, so the games will adapt accordingly.