CNC radiator processing: how machined heatsinks actually work

• 1
  What CNC radiator processing changes in the heat path
• 2
  Fin geometry limits in CNC radiator processing
• 3
  Base flatness, bond lines, and interface resistance
• 4
  Choosing aluminum, copper, or a mix
• 5
  When CNC radiator processing is the wrong answer
• 6
  How a machined heatsink moves through the shop
• 7
  Machining against other heatsink processes

A radiator does two jobs: pull heat out of a source and hand it to moving air or liquid. CNC radiator processing mainly improves the first job. Machining a base flat and thin shortens the distance heat travels before it reaches the fins. That distance is the bottleneck in most small heatsinks, not the fin count.

The second job is surface area. Fins add area, but only if air can reach between them. A 0.5 mm gap blocks more airflow than a 1 mm gap at the same fan pressure, so the extra area can be wasted. Machining lets you hold a consistent gap along the whole fin length instead of a nominal value that drifts.

Thermal resistance is the number that matters. It combines conduction through the base, spreading into the fin field, and convection off the fin walls. CNC work touches the first two. If your measured resistance is far above the calculated value, the cause is usually a flatness error or a bond line, not a shortage of fins.

Copper and aluminum both machine well, but they behave differently at the tool tip. Copper grabs the cutter and builds heat fast. Aluminum 6061 or 6063 cuts clean at higher spindle speeds. Choose the material by the heat flux first, then accept the machining cost that follows.

• 1
  Base thickness sets spreadingThin bases help point sources, thick bases help wide sources.
• 2
  Gap tolerance sets airflowA drifting slot width chokes the fin field unevenly.
• 3
  Bond lines add resistanceEvery interface layer costs you a fraction of a degree per watt.

The narrowest slot you can mill is set by tool diameter and depth-to-diameter ratio. A 2 mm cutter can reach roughly 10 mm to 12 mm deep before it deflects. Push past that and the slot walls taper, which changes the fin thickness along its height. Skiving handles 0.2 mm fins because it cuts rather than mills.

Fin height matters for the same reason. Tall thin fins vibrate during milling and leave chatter marks on the flanks. Those marks are not just cosmetic, they raise surface roughness and slightly reduce effective area. We run finishing passes at Ra 0.8–1.6 μm on fin flanks to keep the surface predictable.

Pin fins are the other common shape. They break up the boundary layer and work well in forced air with a fan, but they are slower to machine because the tool has to lift between every pin. A plate fin array with the same area usually costs less and performs close enough in ducted flow.

Your width-to-height ratio decides the process. Below about 1:8, milling is fine. Between 1:8 and 1:20, expect to pay for extra passes and slower feed. Past 1:20, look at skiving, extrusion, or a bonded assembly instead of cutting the fins from solid.

• 1
  Tool diameter caps slot widthA 2 mm cutter holds about 0.2 mm tolerance at 10 mm depth.
• 2
  Chatter shows on fin flanksRough flanks cut area and look bad in customer inspection.
• 3
  Pin fins cost more per unit areaEvery pin needs a lift, a move, and a plunge.
• 4
  Aspect ratio is the switchPast 1:20, another process usually wins.

The base is where most machined heatsinks lose performance. A base that is flat to 0.02 mm over 100 mm makes good contact with a thermal pad. A base that bows 0.1 mm leaves an air gap in the middle. Air conducts poorly, so the source runs hot even though the fin field is fine.

We hold ±0.005 mm on critical features and check flatness on the base face before shipment. For a large plate, that means measuring across the diagonal, not just at the corners. The middle of a plate is where bowing hides.

Bonded fins are common when you want copper fins on an aluminum base or aluminum fins on a copper cold plate. The joint is a thin layer of solder or epoxy. It adds a small resistance per interface, and it can fail under repeated thermal cycling if the two metals expand at different rates.

For high-reliability parts, cut the fins and base from one block. There is no bond line to inspect and no differential expansion. The cost is more machining time and more material removed. For most electronics duty, a bonded stack is a reasonable trade, provided the bond layer is thin and void-free.

• 1
  Flatness beats fin countA bowed base wastes the whole fin field.
• 2
  Measure the diagonalCorner readings miss center bow.
• 3
  Monolithic costs moreNo bond line, but more material removed.

Aluminum is the default. Grade 6061 machines cleanly, takes anodizing, and weighs about a third of copper. Grade 6063 extrudes well but is softer, so it can tear during aggressive milling. Grade 7075 is stronger and used when the heatsink is also a structural bracket.

Copper moves heat about twice as fast as aluminum by thermal conductivity, but it is denser and gummier. A copper heatsink for the same fin count weighs roughly three times as much. Copper also needs more care at the cutter, with lower surface speed and sharp tools.

The practical compromise is a copper base with aluminum fins. The copper spreads heat from a small die, and the aluminum fins reject it to air. The base-to-fin joint is the weak point, so the bond must be thin, uniform, and inspected.

For harsh environments, we use 316L stainless or Inconel when corrosion or heat dominates. Stainless conducts poorly, so it only makes sense when the heatsink is also a housing wall. In that case the thermal path is short and the corrosion resistance is worth the loss.

• 1
  6061 is the safe choiceGood finish, good strength, easy anodizing.
• 2
  Copper needs slower speedsWatch chip evacuation and tool wear.
• 3
  Mixed metal joints need inspectionVoids and thin spots drive resistance up.

Machining is great for low volume and complex shapes. It is a poor fit for a simple constant profile sold in thousands. If your heatsink is a straight extrusion with no machined pockets, an extrusion die pays for itself quickly and the per-part cost drops well below milling.

Very thin, very tall fin fields also resist milling. Once the aspect ratio climbs past about 1:20, tool deflection and chatter make the geometry hard to hold. Skiving or a bonded stack gets you there with better results.

Large flat plates with a dense pin field are another mismatch. The tool path is long, the chip volume is high, and the part tends to move as stress releases. A cast or forged near-net blank with a light finish cut often costs less.

The honest rule is this. Machining wins when the part has features that a die cannot form, when the volume is low, and when flatness or tolerance matters more than unit cost. Outside that, another process usually serves you better.

• 1
  Simple constant profileExtrusion beats milling at volume.
• 2
  Aspect ratio past 1:20Skiving or bonding holds the fins better.
• 3
  Dense pin field on a big plateCast or forged blanks cut cost and distortion.