The San Andreas is a strike-slip system. It's not capable of generating very large (magnitude 8 or greater) earthquakes.

The Cascadia subduction zone has had magnitude 9 earthquakes in the past. The last one was on Jan. 26th, 1700. (Thank Japan for having excellent historical records of earthquakes and tsunamis. We know a large earthquake occurred around 300 years ago in the area, and thanks to Japan's record of the tsunami it caused, we can make a solid link to the exact date.)

We don't know the statistical hazard as well due to the small sample size, but when the next large Cascadia earthquake occurs the damage will be absolutely catastrophic.

It's not just the direct earthquake damage, but also the tsunami hazard. You need vertical offset to cause a tsunami. A strike-slip system like the San Andreas is very unlikely to cause a tsunami. The permanent offset is dominantly horizontal, so the only way to generate a tsunami is through secondary effects such as landslides. (Also, most of the length of the San Andreas is onshore.)

Subduction zones are thrust systems. One plate moves up, and the other moves down. Earthquakes there are likely to produce permanent vertical deformation at the surface.

Furthermore, certain types of subduction zones are more prone to generating large tsunamis. A deep earthquake is unlikely to cause much deformation at the seafloor, and therefore doesn't generate as large of a tsunami. However, the shallower the rupture penetrates, the larger the deformation at the seafloor is, and therefore the larger the tsunami is. Certain types of subduction zones are more prone to having large earthquakes that rupture all the way up to the seafloor. (The amount of sediment on top of the incoming oceanic plate is thought to play a large role in this, among other things. The recent Tonankai earthquake in northern Japan turned a lot of what we thought we knew about this on its head, though.)

The Cascadia subduction zone is one of the end-member types that's likely to have both large earthquakes and large tsunamis. We know it has in the past, and it's likely to in the future. It's unusual in that most of the deformation along the fault occurs through periodic creep ("slow-slip events") that doesn't cause an earthquake. (Actually, as we're finding out, it's not that unusual around the world, but it was first observed and is best documented in Cascadia.) However, while this creep does relieve a significant portion of the accumulated elastic strain, it doesn't relieve all of it. The plate boundary fault is still accumulating elastic strain that will eventually be released in a large earthquake.

At any rate, just something to think about. The seismic hazard in the Bay Area can be reduced through proper engineering solutions. (Though SOMA is going to be in very rough shape for the reasons this article mentions. Lesson for next time: Don't bulldoze all the rubble into a pile and then build on top of it!)

For Cascadia, though, you can't engineer your way around a magnitude 9 earthquake and tsunami. You do the best you can, and try to avoid putting critical infrastructure near the coast.