In 1991, stood at the top of the world. It won the 24 Hours of Le Mans with its banshee prototype, and here is the 700-horsepower quad-rotor engine that got it the win.

This is tucked away in Mazda’s hidden basement in Irvine, California, an hour train ride down from the middle of LA, or an hour by car if your cab shows up five minutes late to the train station and you miss your train, not that I would know anything about that.

The first stuff Mazda started storing down here in the late 1980s was sports cars from rich local collectors, for Mazda’s designers to stare at for inspiration, as my impromptu tour guide Jacob Brown at Mazda explained the other day. It wasn’t long before it became a repository for development prototypes and clay models (no pictures, no pictures), Mazda’s own vintage cars it restores for historical value and its old race cars it runs in events for publicity. Mazda doesn’t sell a rotary car anymore. It needs to get the message out one way or another, and having an old prototype howl down the corkscrew at Laguna Seca is as good a way as any.

This engine here is what’s called an R26B, the 26 standing for its 2.6 liters of displacement, though the way that’s counted is kind of arbitrary. The important thing is that it’s a four-rotor. It’s easy to count on the lighter metal rings, the ones with three spark plugs sticking out the sides.

This wasn’t Mazda’s only four-rotor, as there was also a 13J, the differences between which I forgot, though they were explained to me by Mazda’s in-house rotor guru, busily working on this R26B rebuild. I’ll say that I was too busy getting lost in the details of the engine. Luckily for me, Le Mans tech expert highlighted this very subject:

Calling upon the foundation laid by its 4-rotor predecessor, the 13J-M, Mazda added a number of refinements and racing-derived features. These included intake ports on the periphery of the rotor housings, telescopic intake runners (variable height tuned to engine RPM), 2-piece ceramic apex seals, and 3 spark plugs per rotor instead of the usual 2 to reduce fuel consumption. This produced a motor capable of developing 900 hp at upwards of 10000 rpm, although it was detuned to ~700 hp (some say even as low as 630 hp) in order to provide reliable service throughout the race.

I’ll come right back to those intakes in a moment.

I’m sure that this mount wasn’t intentionally shaped like a rotor, but I enjoyed it anyway. The super lightweight pulley below it also caught my eye.

But the best part is that intake. As you can see it has velocity stacks, which are designed to optimize how air flows into the engine. (It would be enough if they just looked cool, but they do have an important function.) The thing is that the length of the trumpet changes how the engine runs, so picking one length ends up with a bit of compromise, as explained in our , citing an old Volkswagen forum thread:

Shorter runners lengths give more power at high RPMs, longer lengths give more power at lower RPMs.

Read through the whole explainer if you’re curious about more intake science and photos of pretty old race engines.

Velocity stacks may not be the absolute coolest thing you can put on your car, but... actually, you

The Mazda rotary is known for revving high and being happy high in the rev range. This engine makes a solid 700 horsepower up at 9,000 RPM and it’s less stressed hanging out up there than a piston engine. Where it really needed help was low-end power, as rotaries are notoriously poor for low-end torque. That’s what bigger, gruntier piston engines do much better.

So high-RPM performance is something Mazda didn’t want to lose, but it needed help down low. The solution was an intake that was both short and long. Short in low revs, long in high revs.

And here is that extremely dramatic variable-length intake runner system. This isn’t something exclusive to Mazda. I’ve seen these on V10 Formula 1 cars, current Formula 1 cars and even the LaFerrari. The difference is those variable intakes are still on the very short side, not changing all that much from short to long.

The ones on the R26b pop out like a trombone. Here you can see all the way out and all the way in next to each other.

And here’s the mechanism that slides out the R26B’s intake runners. This was continuously variable, controlled by the ECU.

That dramatic difference just highlights how dramatically the rotary needs help with low-end torque. That is a very, very long intake. The science behind this is interesting. Basically, the point of a properly tuned intake runner is to give you something called “resonance supercharging,” which means that the air being drawn into the engine through the intake pulls all the way down in a pressure wave, compressing itself up against the intake opening up at just the right time, at just the right engine speed. I went over this in the velocity stack article before, but I’ll once more use this great explanation from history site :

What engine speeds, you say? That depends on the frequency of the pressure wave. If you stayed awake in high school physics, you may dimly recall that the frequency of a wave is inversely proportional to its wavelength (that is, a short wavelength means a high frequency and vice versa). In this case, the wavelength is determined by the length of the space in which the pressure wave can move, i.e., the length of the intake runner. All else being equal, the longer the runner, the lower the engine speed at which resonance supercharging occurs and vice versa.

So a really long runner helps with really low RPM, something that Mazda needed badly. Science!

Now, the way that Mazda ended up clinching the overall victory in 1991 was as much a fluke of the rules as anything else, but the win was indeed a proof of concept for the rotary, and an incredible tech showcase for what’s looking now a bit like an engineering dead end. Maybe that’s why I like it so much, looking like some piece of a Bladerunner alternate history, accidentally dropping into our world and screaming its lungs out.