If you have an ounce of rotary blood in your veins, you already know how this story starts: Mazda said the RX-8 made 247 hp. Several owners' dynos said otherwise. Then, Mazda changed the rating to 238 hp and offered either cash and free maintenance or a buyback to the 3,000 or so people who already bought one.
Unfortunately, the story doesn't stop there. The dyno doesn't seem to agree with the new number, either.
We honestly don't try to police factory horsepower ratings. Manufacturers measure power at the flywheel with a brake dyno; we measure power at the drive wheels with an inertia dyno. Calculating one number from the other requires detailed knowledge of all the stuff between the engine and the wheels. You have to know precisely how much friction comes from the gearbox, CV joints, wheel bearings and tires, and exactly how much rotational inertia those parts have. We don't know any of that.
We do, however, notice patterns. When every 140-hp 2000 Miata made about 110 hp, and every 155-hp 2001 Miata tested also made 110 hp, we knew something was wrong. Mazda did the same re-rate, rebate, and buyback routine on that car, too. The same happened with Nissan when the QR25 and VQ35 engines debuted, and there have been countless smaller, less obvious discrepancies from several other carmakers that didn't justify a re-rating.
On the RX-8, Mazda is cursed with a flywheel power rating that's uncomfortably close to that of a dyno veteran's. The Honda S2000 (at least until the 2004 model) is rated 2 hp stronger than the RX-8, with a power peak at about the same rpm. The two have the same basic drivetrain layout, though none of the components are actually the same. The Mazda drivetrain could have more drag, and the RX-8's 18-inch wheels have more rotational inertia than the S2000's 16s. There are enough differences that you wouldn't expect the two to make exactly the same power at the wheels.
But the difference is 14 hp. Could the S2000 have 16 percent driveline loss and the RX-8 have 21 percent? Maybe, but it doesn't seem likely. Mazda agrees. Multiple sources at Mazda tell us the RX-8 is so computer laden, it's nearly self-aware. When it sees the rear wheels spinning and the fronts sitting still, it realizes there won't be much air flowing across the catalytic converter. It also realizes that 50,000-mile emissions system durability is a federal mandate, and that if it overheats its cat with screaming-hot rotary exhaust, it might get in trouble with the feds. So the RX-8, according to Mazda, goes into don't-overheat-the-cat mode and tries lowering the exhaust temperature by running extra rich. This lowers power output under chassis dyno conditions.
If we're measuring some cat-safe mode, how do you measure the real, full-power mode? I spent most of a weekend trying to trick the RX-8 by running jumper wires from the rear wheel speed sensors to the ABS computer's front wheel speed input. This, I hoped, would let the RX-8 think all four wheels were turning at the same speed. Getting my ruse past the car's multiple self-diagnostic systems proved difficult, however. I was able to get the ABS computer to accept the left rear wheel signal to both left-side inputs, but when I did the same on the right, there was a fireworks show of warning lights on the dash.
Undaunted, I next tried measuring power on the road with an accelerometer. Tesla Electronics' new G-Tech Pro Competition is a shockingly sophisticated little box that, among other things, is able to measure horsepower on the road in essentially the same way a Dynojet does, by watching how quickly the engine can accelerate a known mass. In the Dynojet's case, the known mass is a set of rollers (3,600 pounds). In the G-Tech's case, the known mass is the car (2,950 pounds) and its occupants (394 pounds). Wind resistance, rolling resistance (two more tires are turning) and the greater rotational inertia losses from testing in second gear (on the road) vs. third gear (on the dyno) mean the two will produce different numbers, but if I could trick the RX-8 into thinking it was a dyno,perhaps I could see the difference between go-fast mode and don't-overheat-the-cat mode.
It should be noted that I have far less experience with the G-Tech than with the Dynojet, and therefore trust its results less. A 14-hp difference between these two theoretical operating modes, though, should be easy to spot.
In six runs with all four wheel-speed sensors working, most of the runs (4 of the 6) looked almost exactly like the orange line on this dyno chart. The other two were higher by about 5 hp. Six more runs with the front ABS sensors disconnected yielded similar, though slightly more erratic, readings.
No sign of a different mode here, but the conditions of our test aren't a perfect match for dyno conditions. On startup, most ABS computers do a simple impedance check to make sure all the sensors are there. On the dyno, it knows they're there, but doesn't see them turning. In our test, it knows the sensors aren't there. (In both cases, the ABS malfunction light comes on.)
In hopes of circumventing the startup check, we plugged the sensors back in, started the car, drove it long enough to convince the computer that all was well, then unplugged them while the car was still running. The ABS light came on again, but miraculously, power dropped about 20 hp for the next two runs. The blue line on the chart is the second of these runs. Is this the elusive don't-overheat-the-cat mode?
The next two runs, with the same setup, were back to normal, almost identical to the first six runs. Then we tried our plug-drive-unplug ruse again, but were unable to duplicate the low-output runs. They could be a fluke, an error in zeroing the G-Tech's accelerometer, or the RX-8's salvation. The Dynojet results, and the exhaust temperature logic both suggest that a cat-saving mode would only appear at high rpm, not everywhere as the low G-Tech runs show.
So how much power does an RX-8 make?I have no idea. The only thing I'm sure of is that on the road, the whole number quest seems silly. The RX-8 is fast, and the Renesis is smooth, flexible, and torquey compared to its dyno rival, the (pre-'04) S2000. Spend half an hour weaving apexes together at 8000 rpm and you forget all about numbers and charts. I, for one, am more interested in seeing what this sweet little motor will do uncorked than what it does now.