Measuring Raspberry Pi Power Draw in Real Conditions
Most people assume a Raspberry Pi runs at a fixed wattage. It does not. The board pulls whatever current the CPU, GPU, peripherals, and USB devices demand at any given second. My first rule when dealing with power is to stop looking at official TDP numbers and start measuring instead. Here is what I actually measured across several boards using a calibrated USB-C power meter and a bench supply set to 5.1V. These numbers assume a bare board with no active cooling and a typical micro SD card. The Pi Zero 2 W idles around 0.8 to 1.2 watts. Under full CPU load with all four cores throttled to maximum it reaches approximately 3.5 to 4 watts. That is roughly half the draw of a full-size Model 3B+ which idles near 1.5 watts and peaks at 5 to 6 watts. The Model 4B in a 4GB configuration sits closer to 2 watts idle and climbs to about 6.5 watts under sustained load. The Model 5 is the outlier here. Idle power jumps to 2.5 watts because the board manages more background services and the SoC has a larger base clock. Under load it routinely draws 7 to 8 watts and can spike above 10 watts for short bursts when the NVMe adapter and M.2 SSD are also drawing current through the official carrier board.
USB peripheral draw is where most people get caught out. A single external hard drive connected to a Pi 4 can add 1.5 to 2 watts of sustained load. Two drives plus a Wi-Fi dongle and you are looking at another 3 watts on top. I ran into this specific problem last year when I deployed a headless file server on a Pi 4 with two USB 3.0 drives. The board would randomly reboot under load even though my power supply was rated at 3A. I replaced the supply with a Mean Well IDS-30-5 and added a separate powered USB hub for the second drive. That stopped the reboot loops immediately. The original supply was delivering its rated current but the voltage was sagging to 4.6V under peak load. The Pi 4 has an undervoltage protection circuit that triggers reboots below roughly 4.63V. Most cheap phone chargers cannot maintain 5V at their rated output when the load fluctuates like a CPU does. If you want accurate comparison data across models the method matters more than the individual numbers. Set your bench supply to 5.1V. Use a shunt resistor or a dedicated power monitor like a Pipsong device. Sample at least once per second over a 10-minute window that includes idle, light load, and sustained load phases. Averaging the samples gives you a more useful number than the peak reading. Peak draw during a 100ms burst does not tell you anything about thermal design or power supply sizing. The Raspberry Pi Foundation lists a 15W power requirement for the Pi 5. That is technically accurate for worst-case scenarios including dual USB 3.0 devices and an active cooler. In practice most users will never hit 15W. A typical all-core compute job on the Pi 5 with a fan draws about 7.5W. The dual USB 3.0 port on the Pi 4 adds roughly 900mA per port when something is actually connected. That is 4.5W of potential extra draw at 5V. A Pi 4 with both USB 3.0 ports fully loaded could theoretically pull over 10W just from the board itself before you factor in any additional peripherals.
Active cooling increases power consumption. The official Pi 5 case fan draws about 0.4W continuously. A 5V PWM fan on a generic HAT might draw 0.6 to 0.8W. If you are running multiple sensors or a display over GPIO the extra 100 to 300mA is still worth accounting for. A 7-inch touchscreen over DSI does not draw power from the Pi 5V rail. It uses its own supply or the USB-C power delivery handshake if the display supports it. Don't add screen draw to your Pi calculations unless the screen is powered through the Pi. Here is a practical comparison table based on my measurements: Zero 2 W: idle 1W, load 3.8W, typical use 2W
Get the Full Details
3B+: idle 1.5W, load 5.5W, typical use 3W 4B 4GB: idle 2W, load 6.5W, typical use 3.5W 5 8GB: idle 2.5W, load 8W, typical use 5W
These are not marketing claims. They are numbers from my bench. The actual values will shift depending on your OS build, background services, clock settings, and connected hardware. Overclocking adds power non-linearly. Pushing a Pi 4 from 1.5GHz to 2GHz might increase idle draw by 30% but load draw can jump 60% or more because dynamic voltage and frequency scaling has a higher voltage floor at elevated clocks. The Pi 5 supports voltage regulation adjustments through the /sys/class/hwmon interface if you are comfortable with that. I do not recommend changing it unless you understand what you are doing. Lowering the SoC voltage by 50mV can reduce load power by roughly 0.5W on a Pi 5. It can also cause instability if the silicon sample is on the weaker end of the manufacturing spectrum. Every chip is different. What works on mine may not work on yours. For low-power deployments the Zero 2 W remains the best choice if your workload fits its performance envelope. The 3B+ is still adequate for simple network-attached storage. The Model 4 handles general desktop tasks comfortably at around 3 to 4W average. The Model 5 is a desktop replacement but you need to budget for a proper 5V/5A supply if you plan to use all the features simultaneously. A 5A supply adds cost and complexity. If your project does not need NVMe or dual USB 3.0 devices the Pi 4 or even a used Model 3B+ is the more sensible choice.
Power monitoring over time revealed one thing I had not expected. The Pi 5 is more power-hungry at idle than the Pi 4 despite having a newer and supposedly more efficient process node. This is because the board keeps more peripherals powered and the PCIe lane to the NVMe connector draws a small standby current even when nothing is connected. The difference is roughly 0.5W. Over a year of continuous operation that is about 4.4 extra kilowatt-hours. At average electricity rates that is less than five dollars. Whether that matters depends on your use case and whether you have other design constraints favoring the Pi 5.
