Not 10%. In our simulation it's about 90 times slower, and it spends 99% of its time waiting.

Your work lives in RAM, which is small and fast. What doesn't fit goes out to the drive, which is huge and slow. On a laptop we measured, a random read from RAM took about 135 nanoseconds and a read from the SSD about 100 to 180 microseconds: one trip to the drive costs about as much as a thousand reads from RAM. If one step of the processor (about a third of a nanosecond, at a round 3 GHz) took a second, a RAM read would take about seven minutes and a drive read about five days.

In 1962 the Atlas computer in Manchester made a small fast core memory and a big slow drum "appear to the programmer as a single level of storage", moving pages between them on its own. Every laptop still works this way. By 1968 shared computers were collapsing when one more program came in, and Peter Denning showed why: "Thrashing inevitably turns a shortage of memory space into a surplus of processor time."

Our simplified simulation: a program reaching at random over 110 pages when RAM holds 100, the least recently used page going out, with the times we measured. Real computers also compress memory before they use the drive, which pushes the cliff further out; ours leaves that out.

More memory doesn't make a computer faster. It moves the cliff.

What's the most tabs you've had open before your laptop gave up?

Sources: T. Kilburn, D. B. G. Edwards, M. J. Lanigan and F. H. Sumner, One-level storage system, IRE Transactions on Electronic Computers, April 1962; Peter J. Denning, Thrashing: Its causes and prevention, AFIPS Fall Joint Computer Conference, 1968; our own measurements and simulation.
Sounds and music: made by us.