Everyone thinks bigger ports flow more air and make more power. That's half true — and the other half is why so many "ported" heads make less torque than stock. Porting is really a fight between cross-sectional area and port velocity, and getting it wrong ruins the engine's character.

Why velocity matters more than CFM: The intake charge has mass. Once it's moving through a small port at high speed, its inertia keeps packing the cylinder even after the piston starts moving up (this is ram effect). Open the port up too much, velocity drops, inertia disappears, and cylinder filling actually gets worse at low and mid RPM. You gained peak flow and lost the powerband.

The three porting jobs, in order of return-on-investment:

• Bowl blend / valve job cleanup — the area just under the valve seat is usually the worst restriction. A good three-angle valve job plus bowl blending often nets 15-25% more flow with zero downsides. This is the free lunch.
• Short-side radius reshaping — the sharp turn on the floor of the port where it dives toward the valve is where flow separates. Smoothing this radius is high-value and doesn't add cross-section.
• Full port enlargement — grinding the walls back. High risk. You lose velocity, you can hit water jackets, and gains only show above ~5500 RPM.

Real-world example: Chevy LS truck heads (317 castings) flow around 235 CFM at 0.600" lift stock. A pocket-port with valve job job takes them to ~280 CFM — a 19% gain — while keeping port cross-section stock. Full CNC porting pushes them to 310+ CFM but drops mean port velocity below 240 ft/sec, and dyno results routinely show less torque under 4000 RPM on a street truck.

Rule of thumb — target port velocity: Aim for 240-280 ft/sec mean velocity at peak torque RPM. Calculate required cross-section:

Area (sq in) = (CID × RPM) / (V × 3456)

For a 350ci V8 targeting peak torque at 4500 RPM with 260 ft/sec velocity: Area = (350 × 4500) / (260 × 3456) ÷ 8 cylinders ≈ 2.19 sq in per port. Any bigger and you're trading torque for a peak-HP number on a flow bench that the engine can't actually use.

Diminishing returns are brutal: The last 10% of flow gain typically costs 60% of the labor. And every CFM above what the cam, intake, and exhaust can use is wasted — the whole intake system is a chain of restrictions, and porting one link past the others just moves the bottleneck.