How Much Horsepower Does One Point of Compression Add? We Measured It
Forums say 3–4% per point. We swept an LS7 and a 4.6 Modular from 8.5:1 to 13.5:1 with knock on and off, and found where the gain stops being free.
How much horsepower does one point of compression add? Ask on a forum and the answer arrives before you finish typing: three to four per cent, fading past 12.5:1. Nobody in the thread has measured it, because measuring it on a real engine means a new set of pistons for every data point.
In our simulator it is a slider. Load the LS7 and it makes 501.54 hp at its stock 11.0:1. Drag the slider to 12.0:1 and it makes 521.57. Drag it to 13.0:1 and it makes 540.09. Same heads, same cam, same fuel, with knock switched off, which on this engine's 98 octane changes nothing: the knock index reads 0.0000 at all three. Now put the same engine on 91 octane, hold it wide open at 4,500 rpm and watch the torque instead: 656.3 N·m at 11.5:1, 654.6 at 12.0, 647.6 at 13.0, 643.9 at 13.5. The slider kept going. The engine stopped following it.
Key Takeaways
- In our simulator one point of compression is worth about 5.6% of peak power at 8.5:1 and about 3.3% at 12.5:1, with knock switched off. The gain shrinks every half-point, and the curve is the same on a 7.0 L pushrod V8 and a 4.6 L SOHC V8 to within 0.1 point.
- Over the full sweep, 8.5:1 to 13.5:1, the LS7 went 443.24 → 548.86 hp (+23.8%) and the 4.6 Modular 256.08 → 316.34 hp (+23.5%).
- On pump fuel the gain ends early. Wide open at 4,500 rpm, torque on both engines peaks at 11.5:1 on 91 octane and 12:1 on 93, then falls as the knock derate takes back what compression gave.
- Premium fuel adds nothing until the engine is knock-limited: the 98-octane column and the knock-free control are identical everywhere the knock index reads 0.0000.
- The same 4.6 at 8.0:1, with the turbo the model will fit on 91 octane, makes 492.8 hp, against 316.34 for the naturally aspirated engine at 13.5:1 with knock switched off. A turbo engine runs low compression because that is where the power is.
On this page: Quick answer · Why compression makes power · The sweep, knock off · The sweep, knock on · Why turbo engines run lower compression · The folklore · What we do not model · Try it in the simulator · FAQ · How these numbers were made
Quick Answer: How Much Horsepower Does One Point of Compression Add
Between about 5.6% and about 3.3% of peak power per point in our simulator, depending on where you start: the first point above 8.5:1 is worth the most, and every point after it is worth less. Measured with knock switched off, an LS7 gained 23.8% from 8.5:1 to 13.5:1 and a Ford 4.6 Modular gained 23.5%, which compounds to 4.37% and 4.32% per point. Both engines land on the same per-step curve, so the number is a property of the thermodynamics rather than of either engine.
On pump fuel you do not get to keep all of it. Held wide open at 4,500 rpm, torque on both engines stops rising at 11.5:1 on 91 octane and at 12:1 on 93, and every half-point past that is a loss. Those crossover points belong to this model, with its spark timing held fixed. A real build that retards timing as compression rises will cross later.
Why Compression Makes Power
Compression ratio is the cylinder's volume at bottom dead centre divided by its volume at top dead centre: swept volume plus clearance volume, over clearance volume (JE Pistons, Compression ratio theory and how to calculate, retrieved 2026-09-10). Raising it squeezes the same charge into a smaller space before ignition, so the burnt gas expands through a larger ratio on the way back down and gives up more of its heat as work. A higher ratio improves thermal efficiency, and in a spark-ignition engine the limit on it is knock (Wikipedia, Compression ratio, retrieved 2026-09-10).
The ideal Otto cycle puts a number on it. Its thermal efficiency is η = 1 − 1/r^(γ−1), where r is the compression ratio and γ the ratio of specific heats of the working gas (Wikipedia, Otto cycle, retrieved 2026-09-10). The exponent is smaller than one, so efficiency rises with r and rises by less each time: the curve flattens. The flattening is why "per point" is the wrong unit to expect a constant answer from, and why the forum figure can be right at one compression ratio and wrong at another. You can watch the pressure trace change as the ratio moves in the Otto cycle explained, which has a live P-V diagram.
A real charge is not air, and its γ is lower than the air-standard 1.4. Heywood's fuel-air cycle analysis (Heywood, Internal Combustion Engine Fundamentals, McGraw-Hill, 1988, ch. 5) puts the efficiency of a real mixture below the ideal and still rising with compression. Our measured gain ought to sit near that curve.
The Sweep: Two V8s, Knock Off
Two engines, so the result is not one engine's quirk. v8_7L0 is our LS7, a 7.0 L pushrod V8 at 104.8 × 101.6 mm, stock 11.0:1 on 98 octane. sohc_v8_4L6 is our Ford Modular 4.6 2V, 90.2 × 90.0 mm, SOHC, stock 9.4:1 on 92. One oversquare big-bore, one undersquare. Stock, on 29-point wide-open sweeps, they make 501.54 hp at 5,875 rpm and 268.97 hp at 5,054.
Then compression from 8.5:1 to 13.5:1 in half-point steps, with knock switched off. Our knock model reads octane, so setting octane to 200 removes the derate without touching anything else. That control is mandatory here: every combustion lever in this simulator reads backwards through the knock derate otherwise, so the thermodynamic gain has to be read with knock out of the way, and the knock-on runs read against it. Spark stays at each preset's fixed value for every step.
| CR | LS7 hp | Δ per half-point | LS7 N·m | 4.6 hp | Δ per half-point | 4.6 N·m |
|---|---|---|---|---|---|---|
| 8.5 | 443.24 | — | 576.10 | 256.08 | — | 376.36 |
| 9.0 | 455.89 | +2.85% | 591.77 | 263.39 | +2.85% | 386.71 |
| 9.5 | 468.11 | +2.68% | 606.71 | 270.33 | +2.64% | 396.54 |
| 10.0 | 479.74 | +2.49% | 620.96 | 276.97 | +2.46% | 405.94 |
| 10.5 | 490.90 | +2.33% | 634.55 | 283.30 | +2.28% | 414.98 |
| 11.0 | 501.54 | +2.17% | 647.59 | 289.38 | +2.14% | 423.58 |
| 11.5 | 511.78 | +2.04% | 660.05 | 295.18 | +2.01% | 431.87 |
| 12.0 | 521.57 | +1.91% | 672.09 | 300.78 | +1.90% | 439.78 |
| 12.5 | 531.02 | +1.81% | 683.56 | 306.15 | +1.78% | 447.43 |
| 13.0 | 540.09 | +1.71% | 694.78 | 311.33 | +1.69% | 454.80 |
| 13.5 | 548.86 | +1.62% | 705.51 | 316.34 | +1.61% | 461.85 |
Read the two Δ columns against each other. At every step they agree to within 0.1 point, from +2.85% on both engines at 8.5 → 9 to +1.62% and +1.61% at 13 → 13.5. The LS7 and the 4.6 share no geometry, no cam and no head; what they share is the Otto cycle. Per full point, that is about 5.6% at the bottom of the sweep and about 3.3% at the top.
The peaks hold still. The LS7's power peak sits at 5,875 rpm from 9.5:1 up and its torque peak at 4,525 until 13:1, where it drifts to 4,750; the 4.6's power peak stays at 5,054 until the last step. Compression raises the whole curve rather than reshaping it.
The Ideal Cycle Beside the Measurement
The sanity check is whether the simulator is doing physics or a lookup. Take η = 1 − r^(1−γ) across each half-point step and compare the fractional change in efficiency with the fractional change in measured LS7 power, for the air-standard γ of 1.4 and for 1.3, closer to a fuel-air charge:
| Step | Δη/η, γ 1.4 | Δη/η, γ 1.3 | Measured Δhp/hp (LS7) |
|---|---|---|---|
| 8.5 → 9 | +1.67% | +1.89% | +2.85% |
| 9 → 9.5 | +1.52% | +1.72% | +2.68% |
| 9.5 → 10 | +1.39% | +1.58% | +2.49% |
| 10 → 10.5 | +1.28% | +1.46% | +2.33% |
| 10.5 → 11 | +1.18% | +1.35% | +2.17% |
| 11 → 11.5 | +1.09% | +1.26% | +2.04% |
| 11.5 → 12 | +1.02% | +1.17% | +1.91% |
| 12 → 12.5 | +0.95% | +1.10% | +1.81% |
| 12.5 → 13 | +0.89% | +1.03% | +1.71% |
| 13 → 13.5 | +0.84% | +0.97% | +1.62% |
The measured gain runs above the ideal-cycle gain at every step, by a steady margin, and the two decline together. Two things account for the margin. First, combustion here is a Wiebe burn spread over 60° of crank on the LS7, with heat transfer to the walls, not the instantaneous constant-volume heat release the ideal cycle assumes. Its efficiency sits below the ideal, and a cycle that starts lower gains a larger fraction from the same improvement. Second, friction in this model is charged on mean piston speed, which compression does not change, so the friction bill is the same at 8.5:1 as at 13.5:1 and brake power moves by more than indicated power does. The ideal column is the textbook, not this model's prediction.
The Sweep Again, Knock On
Now put the fuel back. Same compression ladder, each point held wide open at 4,500 rpm for six seconds with torque averaged and knock read as the maximum over the last five of them, on 91, 93 and 98 octane. The index is our model's own intensity number: 0 is nothing, and 0.35 is what the gauge calls heavy knock. Below 11:1 every column is identical to the knock-free control on both engines and the index reads 0.0000 all the way down, so the tables start where something happens.
LS7, torque at 4,500 rpm:
| CR | Knock off | 91 | Knock, 91 | 93 | Knock, 93 | 98 | Knock, 98 |
|---|---|---|---|---|---|---|---|
| 11.0 | 647.7 | 647.7 | 0.0000 | 647.7 | 0.0000 | 647.7 | 0.0000 |
| 11.5 | 660.2 | 656.3 | 0.0239 | 660.2 | 0.0000 | 660.2 | 0.0000 |
| 12.0 | 672.1 | 654.6 | 0.0724 | 670.3 | 0.0144 | 672.1 | 0.0000 |
| 12.5 | 683.7 | 651.4 | 0.1188 | 668.7 | 0.0622 | 683.7 | 0.0000 |
| 13.0 | 694.8 | 647.6 | 0.1669 | 665.5 | 0.1089 | 694.8 | 0.0000 |
| 13.5 | 705.5 | 643.9 | 0.2084 | 661.9 | 0.1487 | 705.4 | 0.0046 |
4.6 Modular, torque at 4,500 rpm:
| CR | Knock off | 91 | Knock, 91 | 93 | Knock, 93 | 98 | Knock, 98 |
|---|---|---|---|---|---|---|---|
| 11.0 | 421.7 | 421.7 | 0.0025 | 421.7 | 0.0000 | 421.7 | 0.0000 |
| 11.5 | 430.1 | 422.6 | 0.0505 | 430.1 | 0.0011 | 430.1 | 0.0000 |
| 12.0 | 438.1 | 420.6 | 0.1013 | 432.5 | 0.0411 | 438.1 | 0.0000 |
| 12.5 | 445.8 | 418.3 | 0.1507 | 430.7 | 0.0914 | 445.8 | 0.0000 |
| 13.0 | 453.2 | 416.0 | 0.1972 | 428.4 | 0.1357 | 453.2 | 0.0000 |
| 13.5 | 460.3 | 413.9 | 0.2348 | 426.1 | 0.1786 | 457.1 | 0.0245 |
Where Torque Stops Rising, Per Octane
On 91 octane, torque at 4,500 rpm peaks at 11.5:1 on both engines: 656.3 N·m on the LS7, 422.6 on the 4.6. On 93 it peaks at 12:1, 670.3 and 432.5. On 98 it is still rising at 13.5:1, the top of the sweep, with a first flicker of knock at 0.0046 on the LS7 and 0.0245 on the 4.6.
Past the crossover the compression keeps adding and the derate keeps subtracting, and the derate wins. The LS7 on 91 at 13.5:1 makes 643.9 N·m at 4,500 against 705.5 with knock off, a gap of 60 N·m: the derate has taken back everything compression added above 11.5:1, and more. Do not read that as a crossover for a real LS7. Ours is a model's number, with spark fixed at −12° for every step, no squish, no chamber shape, and an octane number that is a single scalar rather than a sensitivity. All of that is in the honesty section below.
The unburnt end gas ahead of the flame front, compressed by the piston and by the burning charge behind it, reaches its autoignition temperature and lights on its own, and a fuel's octane rating is its resistance to doing that (Wikipedia, Engine knocking, retrieved 2026-09-10). Our threshold for it is built on the Douaud–Eyzat four-octane-number method (Douaud and Eyzat, SAE Technical Paper 780080, 1978), which is where the rpm and pressure terms in the next subsection come from. The longer version, including what the knock index measures, is in what engine knock is.
Knock Is a Mid-rpm Event Here
Knock in this model depends on rpm, so the 4,500 rpm hold is one reading of it. Power at each engine's stock power-peak rpm, 5,875 on the LS7 and 5,054 on the 4.6, knock on and off:
| CR | LS7 off | LS7 91 | LS7 93 | 4.6 off | 4.6 91 | 4.6 93 |
|---|---|---|---|---|---|---|
| 11.5 | 511.8 | 511.8 | 511.8 | 295.2 | 294.9 | 295.2 |
| 12.0 | 521.6 | 521.6 | 521.6 | 300.8 | 299.7 | 300.6 |
| 12.5 | 531.0 | 530.9 | 531.0 | 306.1 | 302.6 | 305.3 |
| 13.0 | 540.1 | 539.8 | 540.0 | 311.3 | 304.2 | 308.6 |
| 13.5 | 548.9 | 548.1 | 548.7 | 316.3 | 303.3 | 310.0 |
The LS7 loses almost nothing at its power peak on 91 octane, 548.1 hp against 548.9 at 13.5:1, while its torque at 4,500 is already 60 N·m down. The 4.6, whose power peak sits at 5,054 rpm, loses 4% of peak power on 91 at 13.5:1, and on that fuel its peak power tops out at 13:1. The reason is a residence-time term in the threshold, (rpm/4400)^0.19, from Douaud and Eyzat: the faster the engine turns, the less time the end gas spends hot, so 5,875 rpm is cleaner than 4,500 and 5,054 sits in between.
Do not generalise that to "knock does not matter at high rpm". The threshold is calibrated at 4,400 rpm and a manifold pressure of 0.95 bar. Below 4,400 the rpm term is a clamp, so the model says nothing about knock there, and above it the residence-time argument is the model's own, fitted at one point. The same term is what lets our Cosworth CA2006 hold 13.3:1 at 19,250 rpm on 102 RON, and why F1 engines rev to 20,000 rpm is partly a story about residence time.
Why Turbo Engines Run Lower Compression
Boost is compression by another route. A turbocharger raises the pressure of the charge before the piston touches it, so the end gas at the moment of ignition is hotter and denser than the geometric compression ratio alone would make it, and the knock limit arrives sooner. Every point of compression you keep in the pistons is boost you cannot run.
Our simulator's FIT button makes that arithmetic explicit. boostCeilingBar takes the engine's compression ratio and octane, inverts the same end-gas knock chain the derate uses, and returns the highest manifold pressure at which settled knock stays at zero. A fitted turbo is then aimed 0.3 bar under that ceiling, so wastegate creep lands on it rather than past it. The 4.6 across its compression range on three octanes: the ceiling in bar absolute (1.00 is no boost), whether the button will fit a turbo at all, and the fitted engine's dyno peak.
| CR | Ceiling, 91 | Fit? | Net hp, 91 | Ceiling, 93 | Fit? | Net hp, 93 | Ceiling, 98 | Fit? | Net hp, 98 |
|---|---|---|---|---|---|---|---|---|---|
| 8.0 | 1.99 | yes | 492.8 | 2.16 | yes | 533.6 | 2.63 | yes | 629.0 |
| 8.5 | 1.70 | yes | 430.2 | 1.85 | yes | 471.1 | 2.27 | yes | 574.0 |
| 9.0 | 1.47 | yes | 374.6 | 1.60 | yes | 411.8 | 1.96 | yes | 513.6 |
| 9.4 | 1.30 | yes | 338.5 | 1.42 | yes | 370.5 | 1.76 | yes | 466.1 |
| 10.0 | 1.10 | no | — | 1.20 | yes | 347.4 | 1.49 | yes | 401.6 |
| 10.5 | 1.00 | no | — | 1.05 | no | — | 1.31 | yes | 358.1 |
| 11.0 | 1.00 | no | — | 1.00 | no | — | 1.15 | yes | 362.5 |
| 11.5 | 1.00 | no | — | 1.00 | no | — | 1.02 | no | — |
| 12.0 | 1.00 | no | — | 1.00 | no | — | 1.00 | no | — |
On 91 octane the fit is refused from 10.0:1 up; on 93 from 10.5:1; on 98 from 11.5:1. At 12:1 the ceiling reads 1.00 on all three fuels, which means the model finds no boost at all that this engine can carry on pump gasoline. Every fitted engine's settled knock reads 0.0000 at its power peak, by construction: the ceiling and the derate are the same chain read in opposite directions.
Now read that table against the naturally aspirated sweep. With knock switched off, the 4.6 tops out at 316.34 hp at 13.5:1. At its stock 9.4:1, with the turbo the model fits on 91 octane, it makes 338.5. At 8.5:1, 430.2. At 8.0:1, 492.8, more than half again what compression alone could reach with knock removed. The ceiling pays better than the pistons do, and it pays best where the pistons pay worst. The folk range for a boosted engine on pump gas is 8.5–10:1; our fitting rule lands at the low end of it on 91 and only reaches 10:1 on 93. How the compressor makes that pressure, and why a turbo's limit differs from a blower's, is in how a turbocharger works and Roots vs centrifugal superchargers.
What the Folklore Gets Right and Wrong
"Three to four per cent per point." This is the rule the piston makers print, with the caveat that returns diminish near 14:1 (JE Pistons, Compression ratio theory and how to calculate, retrieved 2026-09-10), and the one threads like Speed-Talk's "Debating the % increase in HP per point" (retrieved 2026-09-10) argue over at length and measure never. On our numbers it is right in the 11–13:1 range and an understatement below 10:1, where a point is worth about 5.6%. The rule is one reading of a curve, taken near the compression ratios most people were arguing about.
"It fades past 12.5:1." That sentence folds two claims together, and only one holds here. The thermodynamic gain does not fade at 12.5. It declines from the first step and is still worth about 3.3% per point at the top of the sweep with knock switched off. What fades is what you keep on pump fuel, and on our numbers that happens at 11.5:1 on 91 octane and 12:1 on 93, earlier than the folklore's line. The 12.5 figure has a life of its own in this simulator too: the spec panel warns on any gasoline engine above 12.5:1, and that warning is a rule of thumb that does not read octane. It is not one of these measurements.
"Higher compression needs premium." Half right, and the wrong half costs money. Premium fuel carries no more energy; what it carries is knock resistance, and knock resistance is worth nothing to an engine that is not knocking. In our tables the 98-octane column and the knock-free control are identical everywhere the index reads zero, which on 98 is every half-point up to 13:1 on both engines. AAA's 2016 study found no benefit from premium in vehicles designed for regular (AAA, U.S. Drivers Waste $2.1 Billion Annually on Premium Gasoline, retrieved 2026-09-10). Raise the compression until the engine knocks on 91 and premium becomes the cheapest horsepower there is. Below that line you are paying for octane the engine never uses.
What This Simulator Does Not Model
The crossover points above describe this model, not an LS7 on a dyno. Five specific gaps:
- Spark is one number per engine. Each preset carries a single ignition advance, the gasoline default of −12° on both of these engines, held for every compression ratio. There is no spark map. A real build retards timing as compression rises, trading a little top-end to hold off knock, so the knock-on curves here are pessimistic above each engine's stock ratio and the crossovers would move later with timing free to move.
- The knock threshold is calibrated at one point. 4,400 rpm and 0.95 bar manifold pressure, from the Douaud–Eyzat method. Above that rpm the residence-time term is the model's own; below it, and at part throttle, the terms clamp and the model is silent. Every knock figure above was read at wide-open throttle at or above 4,500 rpm, inside the fitted region. Nothing here is a claim about knock at 3,000 rpm or on a light throttle.
- The 12.5:1 warning in the UI is not a measurement. The spec panel warns on any gasoline engine above 12.5:1 regardless of octane. A 13:1 engine on 98 that reads 0.0000 knock in the table above will still trigger it. The measurements are in the tables; the warning is a label.
- No squish, no chamber shape, no quench, no piston-crown temperature. All of them move where a real engine knocks, and the forum "3–4%" bundles all of them into one number. Ours is the thermodynamic share plus a knock derate, and nothing else.
- Octane is a scalar. Real fuels have a sensitivity, the gap between research and motor octane, and real engines respond to it differently. Here 91 is one number in one threshold.
The gain that runs above the ideal Otto cycle in the second table has an explanation, not an excuse: a Wiebe burn with heat transfer starts below the ideal and gains a larger fraction from the same step, and friction charged on piston speed does not move with compression, so brake power moves by more than indicated power does. The full scope, and the rest of what we get wrong, is in how we build the simulator.
Try It in the Simulator
The knock-free sweep is a probe setting and cannot be reproduced in the product. Everything on real octane can.
- Open the LS7 and run the dyno at stock, 11.0:1 on 98 octane. Then find Compression in the spec panel and drag it to 12.0, 13.0 and 13.5. Watch the peak climb and the knock chip stay dark until the last stop; the chip reads "knock" from an index of 0.05 and "KNOCK!" at 0.35. The stock slider runs from 8:1 to 22:1, and the top of that range is a different article.
- Open the 4.6 Modular. It ships on 92 octane, between the two pump-fuel columns above, so its crossover sits between 11.5:1 and 12:1. Drag compression up from 9.4 and read torque at 4,500 rpm off the dyno curve rather than the peak. Torque at 4,500 is where the derate shows first, because the power peak at 5,054 rpm is cleaner.
- Change the fuel. Octane on a piston engine is set in the build wizard, on the Fuel step, anywhere from 80 to 110. Build a V8 there on 91, then again on 98, and run the same compression ladder on each.
- Fit a turbo to the 4.6 and read the line under the FIT button, which names the ceiling for the current compression ratio and octane. Drag compression up half a point at a time and watch the ceiling fall, and keep going until the line says the engine leaves no room for boost.
For an engine of your own, the compression ratio calculator turns bore, stroke, chamber volume, gasket and deck height into the number the slider wants, or runs the other way and tells you what chamber volume reaches a target ratio.
Run the LS7 · Run the 4.6 Modular · Browse the full roster
Frequently Asked Questions
Does higher compression increase horsepower?
Yes, at every step we measured, as long as the engine is not knocking. With knock switched off, our LS7 gained power at every half-point from 8.5:1 to 13.5:1 and so did our 4.6 Modular, by the same fraction at each step to within 0.1 point. The gain shrinks as the ratio rises, because the Otto-cycle efficiency curve flattens, but it never turns negative on its own. What turns it negative is knock, which is a fuel problem rather than a compression problem.
How much power does one point of compression add?
In our simulator, about 5.6% going from 8.5:1 to 9.5:1 and about 3.3% going from 12.5:1 to 13.5:1, with the steps between falling on a smooth curve. Across the whole 8.5-to-13.5 sweep the LS7 gained 23.8% and the 4.6 gained 23.5%, which compounds to 4.37% and 4.32% per point. Those are this model's figures with knock removed and spark held fixed. A real engine's number bundles in chamber shape, squish and timing, none of which are modelled here.
What compression ratio is safe on 93 octane?
We can only answer for this model, and the answer is 12:1 on both engines, wide open at 4,500 rpm: torque peaks there on 93 and falls at every half-point above it, with the knock index reading 0.0144 on the LS7 and 0.0411 on the 4.6 at that point. That is a crossover with spark fixed and no squish or chamber modelled, so a real build with its timing adjusted will cross somewhere else. The tables above give the whole curve on 91, 93 and 98 for anyone who wants to pick their own line.
Why do turbo engines run lower compression?
Because boost and compression spend the same knock budget, and boost spends it better. On our 4.6 the fitting rule finds a 1.99 bar ceiling at 8.0:1 on 91 octane and refuses to fit a turbo at all from 10.0:1 up; the fitted engine at 8.0:1 makes 492.8 hp, against 316.34 for the naturally aspirated engine at 13.5:1 with knock switched off. The folk range for a boosted engine on pump gas is 8.5–10:1, and our fitting rule lands inside it for the same reason, though the refusal points are the model's, not any manufacturer's.
Does higher compression need premium fuel?
Only once the engine is knock-limited on the cheaper fuel. Premium carries no extra energy, only knock resistance, and in our tables the 98-octane column matches the knock-free control at every point where the index reads zero. AAA's 2016 study found no benefit from premium in vehicles designed for regular (AAA, U.S. Drivers Waste $2.1 Billion Annually on Premium Gasoline, retrieved 2026-09-10). On our LS7 the switch from 91 to 98 is worth nothing at 11:1 and about 60 N·m at 4,500 rpm at 13.5:1. The question is which side of that line your engine sits on.
How These Numbers Were Made
It matters which kind of claim each figure is:
- Physical principles that hold on real hardware: compression ratio raises thermal efficiency by a diminishing amount, the Otto formula η = 1 − 1/r^(γ−1), end-gas autoignition as the mechanism of knock, boost and compression drawing on the same knock margin. Corroborated by the cited sources, not derived from our code.
- Simulator calibration: every horsepower, torque, knock-index and boost-ceiling figure above. The per-point percentages, the 11.5:1 and 12:1 crossovers, the refusal points and the 60 N·m gap describe two engines as this model calibrates them, with spark fixed and a knock threshold fitted at 4,400 rpm and 0.95 bar. The ideal-cycle column is a textbook comparison, not this model's prediction.
- Method: 29-point wide-open-throttle sweeps from idle to redline, each point settled before sampling, in fixed reference air of 1.000 bar and 293 K with no correction applied. Each compression ratio was built the way the preset builder builds it, so injectors and intake were re-sized to match. Knock-on holds ran wide open at a fixed rpm for six seconds with torque averaged and knock read as the maximum over the last five, never from the instant the throttle opens, because a transient reading can be fifty-five times the settled one. The knock-free control is octane 200, which switches the derate off without touching anything else. Boost ceilings come from the same functions the FIT button calls.
- The 29-point grid reads 501.54 hp and 268.97 hp stock, against 502.29 and 269.27 in our published peaks, which use a 24-point grid. Same engines, different sampling.
- No real-engine power, torque or crossover figure appears above. Where we describe what a real engine or a real fuel does, it carries a citation.
Every knock-on run reproduces from shipped presets: load one, set the compression, and you should land on the same curve, because the knock model's random component is seeded and every figure here repeats. If you do not, send it to us. The dyno correction calculator explains how to compare our reference-day figures against a real dyno sheet.
About this article
Written by the Engine Simulator Team, who build and calibrate the physics engine behind these numbers. We publish the model's limits beside its results, because a simulated figure is worth nothing without them. Found an error? Contact the team.
Sources and Further Reading
- Wikipedia, Compression ratio — the definition, and the statement that a higher ratio improves thermal efficiency and is limited by knock (retrieved 2026-09-10)
- Wikipedia, Otto cycle — the ideal-cycle efficiency formula η = 1 − 1/r^(γ−1) (retrieved 2026-09-10)
- Wikipedia, Engine knocking — end-gas autoignition and what an octane rating measures (retrieved 2026-09-10)
- JE Pistons, Compression ratio theory and how to calculate — the common guide to computing a ratio, and the source of the "3 to 4 percent per point" rule with its diminishing-returns caveat (retrieved 2026-09-10)
- Speed-Talk, Debating the % increase in HP per point — the folklore being argued over, cited as folklore (retrieved 2026-09-10)
- AAA, U.S. Drivers Waste $2.1 Billion Annually on Premium Gasoline (2016) — no benefit from premium in vehicles designed for regular, on performance, fuel economy or emissions (retrieved 2026-09-10)
- Heywood, J. B., Internal Combustion Engine Fundamentals, McGraw-Hill (1988), chapters 5 and 9 — fuel-air cycle efficiency against compression ratio, and knock
- Douaud, A. M. and Eyzat, P., "Four-Octane-Number Method for Predicting the Anti-Knock Behavior of Fuels and Engines", SAE Technical Paper 780080 (1978) — the basis of this simulator's knock threshold
Related Reading
- What engine knock is — the derate that ends the gain, and what the knock index measures
- The Otto cycle explained — the efficiency curve behind the first table, with a live P-V diagram
- How a turbocharger works — the other way to raise cylinder pressure, and why it wants low pistons
- Roots vs centrifugal superchargers — the same knock ceiling applied to a belt-driven compressor
- Why F1 engines rev to 20,000 rpm — 13.3:1 at 19,250 rpm, and the residence time that makes it possible
- Compression ratio calculator — the ratio for your own bore, stroke and chamber
- Browse every engine in the simulator — filter by layout to find the V8s named here