How to Read a Dyno Graph

What a dyno graph plots

A dyno graph is not one number: it is two lines plotted against engine speed, drawn from a single wide-open-throttle pull. Engine speed in rpm runs along the bottom. Wheel horsepower usually reads up the left side, and torque, in pound-feet, reads up the right side, on its own scale. See WHP vs Crank HP: Reading Dyno Numbers for what a chassis dyno actually measures to produce that whp line in the first place.

Both lines climb from low rpm, reach a high point somewhere before redline, and then fall off as the pull ends. They rarely look identical, because horsepower and torque are two different quantities on two different scales, related by a fixed formula rather than by simply tracking each other. The next section walks through that formula.

The chart below shows a healthy pull from a naturally aspirated engine, one with no turbo or supercharger forcing extra air in. At marker 1, torque reaches its peak. At marker 2, power reaches its peak instead, later in the rev range, after torque has already started to fall. That gap between the two peaks is normal, and the next section explains why it happens.

How horsepower and torque relate

Horsepower and torque are tied together by one formula: horsepower equals torque, in pound-feet, multiplied by engine speed in rpm, divided by 5,252. Torque is a measure of twisting force, how hard the engine is pushing; horsepower is a measure of how quickly that force does work, which is why engine speed is part of the formula and torque alone is not.

Because rpm is part of the formula, horsepower keeps climbing even after torque has started to fall, as long as rpm is rising faster than torque is dropping. It only turns over and starts falling once torque's decline outpaces rpm's climb. That is exactly the gap between marker 1 and marker 2 on the chart above: torque peaks first, and power keeps building on rising rpm until it, too, runs out of room.

Why the shape of the curve matters more than one number

A shop headline usually quotes one peak whp figure and one peak torque figure, but those two numbers describe a single instant of the pull, not the whole thing. The full curve, everywhere between idle and redline, is what the car actually has available at every point a driver might use it: pulling out of a corner, merging onto a highway, or launching from a stop.

This is often summarized as the area under the curve: the total power delivered across the whole rev range, not just its highest point. Two engines can post the exact same peak whp number and still feel very different to drive if one holds strong output across a wide band of rpm and the other only reaches that number briefly before falling away.

This matters most away from the power peak itself, in the rpm range a driver spends the most time in on the street or at a track's slower corners. A wide, flat curve stays close to its best output through more of that range than a narrow, peaky one, even at an identical peak number. How to Prepare for a Dyno Day covers getting a pull that actually captures this shape cleanly.

Whether you need to read one, and what to watch for

You do not need to read a dyno graph to drive or even to modify a car; a shop or tuner can summarize what a pull showed. But being able to read the shape yourself means you can judge a claimed gain, spot when something looks wrong, and ask a more specific question if it does, instead of taking a single number on faith.

A graph that looks generally smooth, with power and torque climbing and falling in the shapes this guide describes, is usually a sign of a clean pull. A dip, a sudden notch, or a curve that falls off far earlier than it should are worth asking about, since they can point at anything from a momentary fuel or ignition event to a genuine mechanical problem.

None of these on their own means the car is damaged; some have simple causes and some do not. Datalogging: What to Log and Why and How to Read Datalog Graphs cover the channels, air-fuel ratio, knock, boost, that explain what actually happened at the moment a graph looks wrong. The advanced section of this guide comes back to reading these signs in more depth.

• A sudden dip or notch partway through the pull, rather than a smooth climb and fall
• A curve that falls off well before redline instead of near it, on an engine that should still be pulling
• Boost that keeps climbing above its target instead of holding flat, on a turbo car
• Power or torque noticeably lower than a past pull on the same car and same dyno, with nothing changed

Reading a before/after overlay

A before/after overlay plots two pulls on the same axes, usually one power line from before a change and one from after, so the two curves can be compared directly. Reading one well means looking at where the gap between the lines is widest, not just whether the after line ends up higher at its peak.

A change that adds midrange, a smaller or quicker-spooling turbo, a cam swap tuned for low-to-mid response, a manifold change, shows its biggest gap in the middle of the rev range, with the two lines nearly meeting again near redline. A change that adds top end instead, a bigger turbo, better high-rpm airflow, a head or exhaust change aimed at breathing at high rpm, shows the opposite: little gap low down and the widest gap near redline.

The chart below is a midrange-biased gain. At marker 1, the gap is already open by 3,000 rpm. At marker 2, it reaches its widest, about 40 more whp than before, in the middle of the range. At marker 3, the two lines are almost back together near redline: this specific change moved midrange power, not top-end power.

Reading which shape a change produced tells you something a single before/after peak number cannot: whether the car will feel stronger pulling out of a slow corner or merging at part throttle, versus whether it will feel stronger only once it is already spinning hard near redline.

Naturally aspirated vs turbo curve shapes

A naturally aspirated engine's torque curve tends to build gradually, peak somewhere in the middle of the rev range, and taper off toward redline as the engine's ability to fill its cylinders with air falls off at high rpm. Its power curve, built from that torque curve, climbs later and often peaks close to redline, since rpm keeps compounding even as torque is already declining.

A turbo engine's torque curve usually looks different: it climbs quickly once the turbo has spooled, then holds close to flat across a wide plateau while the ECU holds its boost target, before tapering as boost starts to fall off near redline. See Turbo vs Supercharger: How They Differ for how forced induction shapes the curve in the first place.

The chart below shows a turbo car's curve. At marker 1, torque is already near its plateau by 4,000 rpm, well earlier than a naturally aspirated engine's torque peak usually arrives. At marker 2, power reaches its peak by 6,000 rpm and holds essentially flat before it starts to fall, since rpm keeps climbing after torque has stopped until it, too, runs out of room. At marker 3, notice that the lines never touch, even near redline: this graph's left axis runs to 400 and its right axis runs to 500, a different scale on each side, so equal numbers do not land at the same height. That is a property of how the graph is drawn, not something about the engine, and it is worth checking before assuming two lines crossing, or not crossing, means anything on its own.

Reading boost on its own graph

On a turbo car, boost is usually plotted on its own graph rather than sharing an axis with power or torque, since psi is a different unit on a different scale entirely. A boost trace commonly shows two lines: the target the ECU is trying to hold, and the actual boost the sensor measured, so a reader can see whether the two matched.

A healthy trace holds actual boost close to target through the middle of the pull, with a brief ramp while the turbo spools at the bottom of the rev range. Boost creep is when actual boost climbs above target and stays there instead of holding flat, usually because the wastegate can no longer fully bypass the exhaust flow at high rpm and boost that should have been vented keeps building instead.

The chart below shows boost creep starting to appear. At marker 1, boost is locked right on its 18 psi target through the midrange. At marker 2, actual boost starts drifting above target, the first sign of creep. At marker 3, by 6,500 rpm actual boost is running about 3 psi over target and holding there, rather than continuing to climb further.

Correction factors on the sheet: SAE, STD and uncorrected

A dyno sheet's numbers depend partly on the weather the day the pull happened, since colder, denser air lets an engine make more power for the same input than hot, thin air does. A correction factor scales a raw pull to a standard set of reference conditions so two pulls on different days can be compared fairly.

SAE is the current automotive industry correction standard; STD is an older standard with more generous reference conditions that generally reads higher than SAE from the identical raw pull. An uncorrected number is the raw reading on the day, in that weather, with no adjustment at all. See WHP vs Crank HP: Reading Dyno Numbers for how much a correction factor can move a number on a day far from the reference conditions.

A dyno sheet should state which correction, if any, was applied. Two curves plotted on the same graph should use the same correction; comparing an SAE-corrected line to an uncorrected one on the same chart mixes a real change in the car with a difference in how the two lines were calculated.

Comparing sheets from different dynos

Two dyno sheets are only safely comparable when they come from the same physical dyno, use the same correction factor, and were pulled in the same gear. A different dyno design, an inertia dyno versus a load-bearing dyno, or even a different unit of the same design, can read the identical car differently, which has nothing to do with the car itself.

This matters most for a before/after comparison: a gain that looks large on paper can simply be the difference between two dynos rather than anything the part or tune changed, if the before and after pulls did not happen on the same unit.

None of this means a dyno sheet from a different shop is wrong; it is doing exactly what it is designed to do. It just means a number pulled from one sheet should not be lined up against a number from a different sheet as if the two measured the same thing the same way.

• Same physical dyno for both pulls, not just the same brand
• Same correction factor, SAE or STD, stated on both sheets
• Same gear used for both pulls
• Similar weather and engine temperature between the two runs

The horsepower formula, and why power can climb while torque falls

The formula behind every dyno graph is fixed: horsepower equals torque in pound-feet, multiplied by rpm, divided by 5,252. Because 5,252 is the constant in that formula, a torque line in pound-feet and a horsepower line plotted on that exact numeric scale always work out equal at exactly 5,252 rpm. Whether the two lines actually appear to cross on the page depends on whether the graph's left and right axes share that same numeric scale, which is not guaranteed, and was not the case on the turbo chart earlier in this guide.

Torque falling does not mean power is falling; it means the rate torque is falling has to be compared against the rate rpm is rising to know which way power is heading.

This is why power peaking well after torque, often by somewhere in the range of 1,500 to 2,000 rpm on a naturally aspirated engine though the actual gap varies a lot by engine, is completely normal rather than a sign anything is wrong. It is also why comparing where a competitor's torque peaks to where your own power peaks is comparing two different kinds of point on the curve.

Smoothing and correction factors, in more depth

A raw dyno pull is noisier than the clean curve a sheet usually shows: small variations from cylinder to cylinder, momentary sensor noise, and the mechanics of the rollers themselves all add small bumps to a raw trace. Dyno software typically applies a smoothing filter before printing the final curve, averaging out that noise so the underlying shape, not every small fluctuation, is what a reader sees.

This is worth knowing because the sample graphs in this guide are drawn smooth on purpose, to teach the shapes clearly; a real datalog or a raw dyno trace is usually noisier than what is shown here, with small jitter riding on top of the same overall shape. How to Read Datalog Graphs covers reading a noisier, real trace rather than a smoothed teaching curve.

Smoothing and correction are two separate steps, not the same thing. Smoothing cleans up noise within a single pull; correction (SAE, STD, or none) adjusts the whole pull's level to account for the day's weather. A sheet can be smoothed and uncorrected, corrected and unsmoothed, or both, and knowing which was applied matters for comparing it to another sheet.

Over-aggressive smoothing can hide a real, brief event, like a short misfire, by averaging it into the surrounding data rather than showing it as the dip it actually was. A dyno operator who can show the raw trace alongside the smoothed one gives a more complete picture than the smoothed curve alone, especially when something on the graph looks like it needs explaining.

Spotting problems on the sheet

A dip or notch that interrupts an otherwise smooth climb, rather than a curve that falls off and stays down, usually points at a brief event during the pull: a misfire, a boost spike followed by the ECU cutting timing or fuel to protect the engine, or knock retard pulling ignition timing at that exact moment. These tend to recover: the curve dips and then climbs back on toward a normal peak, as it does on the chart below.

A dip can also have nothing to do with the engine at all: a tie-down strap or a tire briefly slipping on the rollers reads as a momentary loss of power in exactly the same way an engine event does. This is one more reason a repeat pull is worth asking for before treating a single dip as a real, repeatable problem.

A curve that falls off well before redline and stays down, instead of dipping and recovering, points at something different: the engine running out of fuel or airflow to support more power at that rpm, rather than a single brief event. This is a limit, not a glitch, and it tends to repeat at the same rpm on every pull rather than showing up once.

Boost creep, covered in the intermediate section above, is a third pattern again: not a dip in power at all, but boost quietly climbing past its target near the top of the rev range. Left unaddressed it can push cylinder pressure and knock risk higher than the tune was built around, which is why a tuner watching a boost trace treats sustained creep as worth correcting rather than a harmless quirk of the graph.

The chart below marks a dip caused by a brief event partway through the pull. At marker 1, power drops even though rpm is still climbing. At marker 2, torque notches down at that exact same rpm, showing the event acted on the engine directly rather than only showing up in the power calculation. At marker 3, the curve recovers and climbs on to a normal peak near redline, which is what marks this as a brief event rather than the car falling off for good.

What experienced tuners check before trusting a number

An experienced tuner rarely trusts a single pull on its own. Running the same pull two or three times back to back and confirming the curve repeats closely is a stronger basis for trusting a number than accepting whatever the first pull showed, since a one-off pull can be affected by a passing condition that will not repeat.

Beyond the graph itself, a tuner cross-references it against the datalog captured during the same pull: air-fuel ratio, knock retard, and boost all logged alongside power and torque. A curve that looks fine but sits on top of AFR running leaner than intended, or knock retard pulling meaningful timing, is not actually a clean pull, even though the power line itself looks smooth. See Datalogging: What to Log and Why and How to Read Datalog Graphs for reading those channels alongside the dyno's own output.

Preparing the car properly before a pull, consistent tire pressure, a full enough fuel load, an engine that is warmed up but not overheated, removes some of the variability that would otherwise be mistaken for a real result. How to Prepare for a Dyno Day covers that preparation in more detail, and pairs with this guide once the pull is done and the sheet is in hand.