Unless you’re talking about an EXTREMELY popular site with a multi-server dedicated deployment (or one with a lot of per-user custom data) the differences aren’t going to be measurable. Once you get rid of the dozens of database calls by caching the pages (at least unauthenticated pages) the difference between a disk or memory cache will be on the order of milliseconds at best.
The different accelerators optimize different parts of the code. APC makes the actual php code run faster, allowing you to handle higher traffic volumes on equivalent hardware. Memcache is usually used to cache database queries but you can technically put anything you want into it and use it as an in-memory cache.
For testing disk response you could run a bunch of tests requesting a static file directly from the server and see how it responds and compare it to something that returns a 200 directly. You could also instrument your code and run some i/o benchmarks. I’d be surprised if that was a common bottleneck though unless it’s a low quality hosting provider - most keep an eye on that kind of thing. Usually the problem is poorly written applications (like, for example, wordpress out-of-the-box).