The swap file: relic or necessary evil?
We first had to go through some application parcours to resolve this fundamental question. We used 8, 12, and 16 gigabytes of RAM in our test system. We also installed and ran a variety of applications, many of them in parallel.
We know that virtual and system memories are two different things. When calculating the memory requirements of different applications, you might assume that 8 GB of RAM or more would be enough to deactivate the swap file. But there's a catch: unfortunately, some programs specifically want to allocate the virtual memory. In these cases, a missing swap file causes an error or the system memory completely fills with garbage data.
You have to carefully test whether the installed application can operate without a swap file. If it can, disabling the swap file is a practical way to avoid slow hard drive access. If not, then you'll just have to live with it. The speed increase is noticeable, but not exactly a killer feature.
To produce a reasonably fair comparison, we started six parallel rendering tasks, each with 2 GB of maximum RAM usage. We ran this test with and without a swap file under the different RAM configurations.
With the swap file deactivated, we ran into problems using 8 GB of RAM (as expected) when Windows ran out of memory. The red values document the state at the time of the crash. If you have 8 GB of RAM installed, you should carefully consider whether it's worth the risk of a possible crash before deactivating the swap file. The performance increase from disabling the swap file is measurable, but still rather modest. At least 12 GB of RAM is needed before you can consider disabling the swap file in a consumer-type environment without worrying--unless you use software that requires a swap file, of course.
We would advise against disabling the swap file if you have less than 12 GB of RAM, at least without thoroughly testing all of your programs first. There is a performance increase when you disable it, but you have to factor in the risk of possible data loss.