Aluminum heat sink CNC machining is the shortest route from a thermal problem to a part you can test: no extrusion die, no tooling lead time, and fin geometry that is not limited by a standard profile. It is also easy to over-specify, because every extra fin and every micron of flatness costs cycle time. This guide sets out the geometry rules that machines well, the tolerances that actually change thermal performance, and how finishing affects both.
Why machine a heat sink instead of extruding or casting it
Extrusion is cheap but fixed: a die gives you one cross-section, and fin height-to-gap ratios above roughly 10:1 to 15:1 are out of reach. Die casting can produce complex three-dimensional housings with integrated fins, but the usual casting alloys carry 10–12% silicon, which drops thermal conductivity to around 96 W/m·K versus 167 W/m·K for 6061-T6 — and the same silicon makes anodising grey and inconsistent.
Machining sits in a different place on the trade-off curve, and 5-axis capability widens it further: a five-axis CNC machining centre can cut a fin field that follows a curved or stepped base, which no extrusion profile can do.
| Route | Conductivity | Geometry freedom | Tooling | Best volume |
|---|---|---|---|---|
| Extrusion + cut-off | ~200 W/m·K (6063) | One cross-section | Low die cost | 1,000+ |
| CNC machining | 167 W/m·K (6061-T6) | Fully custom | None | 1–2,000 |
| Die casting | ~96 W/m·K (ADC12) | Complex 3D | High | 10,000+ |
| Skived fin | 200–240 W/m·K | Very high aspect | Special machine | 500+ |
For CNC prototyping, low-volume production, and any design where the fin field must follow an odd footprint — a cavity, a curved base, a component keep-out — machining wins. Our low volume CNC machining programme covers the 20–500 piece band most thermal programmes live in.
Material selection
| Alloy | Conductivity | Machinability | When to use |
|---|---|---|---|
| 6061-T6 | ~167 W/m·K | Excellent | Default choice; structural plus thermal |
| 6063-T5 | ~200 W/m·K | Excellent | Best conductivity-to-cost for fin arrays |
| 6063-T6 | ~200 W/m·K | Excellent | When the part also carries load |
| 7075-T6 | ~130 W/m·K | Good | Weight-critical aerospace and robotics |
| Copper C110 | ~390 W/m·K | Moderate, gummy cut | High-density power modules, spreaders |
Copper conducts more than twice as well as aluminium, but it weighs about 3.3× more, costs more per kilogram, and cuts poorly — longer cycle times and shorter tool life. The usual compromise is a copper base plate or vapour chamber with an aluminium fin array above it. More on alloy behaviour in CNC machining materials.
Geometry rules that keep the price sane
Thin, tall fins deflect and chatter. These are the rules our programmers work to, and the ones that move the quote:
- Fin thickness: 1.0–3.0 mm. Below 1.0 mm needs high-speed machining with light radial engagement and pushes cost up steeply.
- Fin spacing: 2–5 mm for forced air, 5–10 mm for natural convection. Wider gaps for passive cooling are counter-intuitive but correct — narrow channels stall natural convection.
- Aspect ratio: keep height-to-thickness at or below 30:1. A 30 mm tall fin should be at least 1.0 mm thick.
- Fin height: up to 50 mm is practical; beyond that, fins become inefficient anyway because the tip runs close to air temperature.
- Base thickness: 3–6 mm. Thinner bases warp during machining and then will not sit flat on the die; thicker adds weight and cost for no thermal gain.
- Fin tips: add a 0.2–0.5 mm chamfer. It removes a sharp edge, reduces stress concentration, and survives handling better.
If you need extreme surface area in a fixed footprint, the hybrid answer is usually cheaper: machine the base and mounting features, then bond or press-fit an extruded or skived fin array. That is a common structure for high-power LED and telecom hardware.
Tolerances that matter — and ones that do not
Not every dimension on a heat sink is a thermal dimension. Over-tolerancing the fin field is the single most common cost mistake we see on thermal drawings.
| Feature | Recommended tolerance | Why |
|---|---|---|
| Thermal interface face flatness | 0.02–0.05 mm over the contact area | Directly changes thermal resistance |
| Thermal interface roughness | Ra 0.8–1.6 µm | Sets contact resistance with the TIM |
| Mounting hole position | ±0.05 mm | Must align with the PCB or device |
| Fin thickness | ±0.05 mm | Affects airflow, not assembly |
| Fin spacing | ±0.03–0.05 mm | Affects airflow and pressure drop |
| Overall envelope | ±0.1–0.2 mm | Rarely functional |
Flatness is the one worth paying for. Going from 0.10 mm to 0.05 mm flatness on the interface face typically reduces thermal resistance by 8–15%, because the interface material fills less of a gap. Specifying it separately from the general profile tolerance — rather than hiding it in a blanket title-block tolerance — is what makes the intent clear. The mechanics are explained in our CNC machining tolerance guide.
Finishing: anodising is a thermal decision
| Finish | Thickness | Thermal effect | Use |
|---|---|---|---|
| As-machined | — | Baseline, emissivity ~0.1 | Forced-air, hidden parts |
| Bead blast | — | Slightly larger surface area | Matte cosmetic, mild gain |
| Clear anodise, Type II | 5–15 µm | Negligible thermal change, adds electrical insulation | Default protection |
| Black anodise, Type II | 5–15 µm | Emissivity ~0.8, 10–15% better passive performance | Fanless designs |
| Hard anodise, Type III | 25–50 µm | Dielectric, wear resistant | Rarely needed for heat sinks |
For a fanless, natural-convection design, black anodising is close to free performance: emissivity rises from about 0.1 on bare aluminium to about 0.8, which typically buys 10–15% lower thermal resistance for a small finishing premium. Remember the dimensional allowance — Type II at 5–15 µm changes a Ø6 mm hole by only 0.01–0.03 mm, but Type III at 25–50 µm grows a bore by 50–100 µm on diameter. Details in anodizing aluminum and CNC parts finishing.
Machining strategy
Heat sinks are not solid blocks, and the toolpath has to respect that:
- High-speed milling with 5–10% radial engagement and high axial depth keeps cutting forces low so thin fins do not deflect.
- Trochoidal toolpaths maintain constant engagement and eliminate chatter in deep fin channels.
- Sharp 2–3 flute carbide end mills with AlTiN coating reduce cutting force and leave cleaner fin walls.
- Vacuum fixturing holds thin base plates flat without clamping distortion — critical when flatness is the acceptance criterion.
- Rough then finish: clear the channels with a larger tool, then finish fin walls with a smaller one.
- Temporary support ribs for fins over 30 mm tall, removed after machining.
Where machined heat sinks get used
- Power electronics: IGBT and MOSFET base plates, inverter cold plates, UPS and power supply cooling
- LED lighting: street light and stadium fixture housings where the housing *is* the heat sink
- EV and battery: module cold plates and spreaders requiring tight flatness for thermal interface contact — see EV motor housing machining and automotive CNC machining
- Consumer electronics and 3C: CPU, GPU and router cooling where the enclosure is part of the thermal path — see consumer electronics machining
- Telecom and 5G: base station and radio unit housings, often with an integrated fin field
- Audio and pro AV: amplifier chassis where the case doubles as the heat path — see amplifier front panel machining
- Robotics and automation: compact 7075 sinks for motor drivers and servo controllers
Getting a heat sink quoted
Send a 3D model, a drawing with the interface flatness called out separately, the alloy, the finish, and the quantity. A prototype is typically 3–7 days; production runs are 7–15 days depending on fin count and finishing. We run thermal-critical parts under IATF 16949, ISO 9001:2015 and ISO 13485, hold ±0.01 mm on critical features, and quote with a free DFM analysis — for thermal parts that usually means a fin geometry tweak that removes cycle time without losing performance. See the about page for our equipment and certifications, or browse CNC milling services.