Obviously, things like the garbage collector and bounds checking make Go slower than C. On the other hand, although the Go spec does not mention the stack or heap, you still have pretty much control over memory layout. E.g. making a slice of structs will be backed by an array of (unboxed) structs, pretty much as you'd expect.
I write OpenGL apps in Go. Surely this is in no way characteristic of Google's internal use-cases.
Today's machines are multi-core and are not as bare-bones as to require manually carefully coded malloc/free calls. That's what's changed since C and that's what Go is (very well) designed for. Then a few syntactic goodies that don't imply performance side-effects or any black magic but that C didn't have: duck typing, struct composition, interfaces, all that.
I really appreciate watching it evolve.