The Raspberry Pi 5 BCM2712 SoC begins thermal throttling at 85°C and initiates a hard shutdown at 100°C. While a thermal camera for Raspberry Pi debugging (like the InfiRay P2 Pro or FLIR ONE) is the fastest way to spot board-level hotspots, raw infrared pixels are notoriously misleading without emissivity correction and junction-to-ambient thermal resistance math. To actually prevent throttling, you must translate those surface temperatures into internal junction temperatures using Rθ values and select a heatsink based on worst-case wattage, not just physical size.
Translating Thermal Camera Pixels to Junction Math
The most common mistake makers make with thermal imaging is assuming the camera reads the silicon's internal temperature. It does not. A thermal camera measures the surface temperature of the Integrated Heat Spreader (IHS) or the plastic package. To find out if your Pi is actually throttling internally, you need the thermal path equation.
Bare metal IHS lids and solder joints have very low emissivity (typically 0.1 to 0.3), meaning they reflect ambient room IR rather than emitting their own. Your thermal camera will read them as artificially cold. Always apply a small square of Kapton tape (emissivity ~0.95) or a dab of matte black electrical paint to the SoC and PMIC before taking readings.
According to Texas Instruments' thermal design guidelines, the junction temperature (TJ) is calculated as:
TJ = TA + PD × (θJC + θCS + θSA)
- TA: Ambient air temperature (°C).
- PD: Power dissipation (Watts). The Pi 5 BCM2712 can pull up to 12W under heavy synthetic load, but a realistic sustained compile load is around 10W.
- θJC (Junction-to-Case): Internal resistance of the silicon to the metal lid. For the BCM2712 BGA package, this is roughly 2.5 °C/W.
- θCS (Case-to-Sink): Resistance of your thermal interface material (TIM). A standard 1mm silicone thermal pad is ~0.5 °C/W; high-end thermal paste is ~0.1 °C/W.
- θSA (Sink-to-Ambient): The thermal resistance of the heatsink itself. This is the variable you control.
Worked Example: You are stress-testing a Pi 5 drawing 10W in a 25°C room. You install a mid-tier passive extruded aluminum heatsink with a θSA of 3.5 °C/W and use a standard thermal pad (0.5 °C/W).
TJ = 25 + 10 × (2.5 + 0.5 + 3.5)
TJ = 25 + 10 × 6.5 = 90°C
Even though your thermal camera might show the metal IHS lid sitting at a comfortable 60°C, the internal silicon junction is at 90°C—well past the 85°C soft-throttle limit. You need a lower θSA to fix this.
Heatsink Selection and Derating Curves
Choosing a heatsink requires matching the θSA to your wattage budget. If you need TJ to stay under 80°C at 10W in a 25°C room, your total allowable thermal resistance (θJC + θCS + θSA) is 5.5 °C/W. Subtracting the package and TIM leaves a maximum θSA of 2.5 °C/W.
| Cooling Type | Example Part / Model | Approx. θSA (°C/W) | Est. TJ at 10W (25°C Ambient) | Verdict |
|---|---|---|---|---|
| Stamped Aluminum (Passive) | Generic 14x14x5mm stick-on | 15.0 | 180°C (Hard Shutdown) | Useless for Pi 5 SoC |
| Extruded Aluminum (Passive) | Aavid Thermalloy 576802B00000G | 4.5 | 100°C (Throttling) | OK for Pi 4, fails Pi 5 load |
| CNC Fin Array (Passive) | Geekworm C5 Pro Aluminum | 3.2 | 87°C (Soft Throttle) | Good for light/medium duty |
| Tower Cooler (Active 5V PWM) | Geekworm C7 + Noctua NF-A4x10 | 1.8 | 73°C (No Throttle) | Required for sustained heavy compute |
Interpreting the Derating Curve:
Heatsink datasheets provide derating curves showing how θSA increases as natural convection stalls or ambient temperature rises. If you put your Pi inside a sealed ABS plastic enclosure, the local ambient air (TA) inside the box will rise. If the internal enclosure air reaches 45°C, that same active tower cooler (θSA 1.8) yields:
TJ = 45 + 10 × (2.5 + 0.5 + 1.8) = 93°C.
The fan is moving air, but it is moving hot air. The derating curve dictates that without enclosure ventilation, active cooling loses its primary advantage.
Airflow, Enclosures, and Thermal Failure Signatures
What do airflow and enclosure modifications actually buy you? Moving air disrupts the stagnant boundary layer of heat clinging to the fins. Even a low-CFM 30mm fan drops the effective θSA of a passive extruded sink by 40% to 60%. For enclosures, you must calculate the Delta-T (ΔT). A good rule of thumb for embedded enclosures is providing 1 square inch of exhaust venting for every 5W of system dissipation to prevent internal ambient buildup.
Ignoring thermal management doesn't just cause annoying CPU throttling; it causes permanent hardware degradation. Watch your thermal camera for these failure signatures:
- BGA Solder Ball Fatigue: Repeated thermal cycling (swinging from 40°C idle to 85°C load) causes the silicon die and the PCB to expand at different rates (Coefficient of Thermal Expansion mismatch). Over months, this shears the microscopic BGA solder balls, leading to intermittent I2C/SPI bus errors or dead memory channels.
- Electromigration: At junction temperatures consistently above 90°C, high current density physically moves metal atoms inside the silicon traces. This thins the conductors over time, eventually causing open circuits.
- Electrolytic Capacitor Drying: The Arrhenius equation dictates that the lifespan of electrolytic capacitors (often found on the Pi's power input or attached HATs) halves for every 10°C increase in operating temperature. A capacitor rated for 2,000 hours at 105°C will fail in just 250 hours if your thermal camera shows it baking at 125°C near a hot voltage regulator.
Raspberry Pi Thermal Camera FAQ
Can a smartphone thermal camera accurately read Raspberry Pi SoC temperatures?
Yes, but only if you correct for emissivity and resolution. Entry-level thermal cameras (like the basic FLIR ONE) often have a resolution of 80x60 pixels. Because the Raspberry Pi SoC is small, a low-resolution sensor will average the hot IHS lid with the cooler surrounding PCB, reporting a falsely low temperature. For accurate bench debugging, use a higher-resolution sensor like the InfiRay P2 Pro (256x192) or Topdon TS001, and always apply Kapton tape to the target area to force an emissivity of ~0.95.
How hot is too hot for the Raspberry Pi 5 PMIC and SoC?
For the BCM2712 SoC, 85°C is the soft-throttle limit and 100°C is the hard-shutdown limit. However, the Pi 5's DA9099 PMIC (Power Management IC) and the surrounding MLCC capacitors are also critical. While the PMIC silicon might survive 125°C internally, running it that hot will rapidly degrade the surrounding passive components and the PCB substrate. If your thermal camera shows the PMIC surface exceeding 80°C, you need to add a small stamped heatsink or direct a low-profile fan across the board.
Why does my thermal camera show the Raspberry Pi voltage regulator hotter than the CPU?
This is incredibly common and usually points to power dissipation inefficiencies. Linear voltage regulators drop excess voltage as pure heat (P = Vdrop × I). If you are feeding your Pi 5V via the GPIO header from a 12V-to-5V buck converter that is poorly tuned, or if a connected USB peripheral is drawing heavy current through the board's internal power gating, the PMIC and surrounding MOSFETs will dissipate massive amounts of heat. Switch to a high-efficiency USB-C PD power supply to bypass the board's internal 5V routing losses, and verify your thermal camera readings drop back to normal.
