Thursday, September 8, 2011

USB Thumb Drive Energy Use

This morning, just before removing an old Imation 64MB thumb drive (the expendable one I take with me to class) I checked the run time left on my netbook's battery: 5:57. Then after removing the flash drive, on a whim, I checked the charge remaining: 6 hours even. So apparently the power taken by the device is small but nonzero. Off and on I've been a little curious about this because sometimes when I remove a thumb drive, the drive is very warm.

So, how much energy do thumb drives use? The Kingston web site and Amazon offer no help, but there is a paper that is very informative, "Power and Performance Characteristics of USB Flash Drives," Kyle O’Brien, David C. Salyers, Aaron D. Striegel, Christian Poellabauer, Department of Computer Science and Engineering, Notre Dame, IN 46556 USA. this apparently appeared in the 2008 International Symposium on a World of Wireless, Mobile and Multimedia Networks, which is an IEEE venue. The paper is available from CiteSeer.

It makes sense that wireless folks would be interested in power consumption since so many wireless devices are battery-powered.

They looked at USB 1.1 and 2.0 devices from a variety of manufacturers and of a variety of sizes up to 4 GB. 4 GB is small now, but this was a few years ago.

USB 2.0 energy consumption was very interesting, with the idle consumption being over twice that of the USB 1.1 device (for a Patriot 4 GB device). If thumb drives are idle most of the time, then one can suppose that the USB 2.0 devices take twice as much power overall as compared to USB 1.1. The differences were approximately 22.4 mA for an idle USB 1.1 device as opposed to approximately 50.9 mA for a USB 2.0 device. The disparity between current required while writing was larger, almost 3:1, at 34 mA vs. 105 mA. Presumably the USB 2.0 device is writing faster so it may spend less time writing.

For a USB 1.1 Kingston 4 GB idle power was around 23 mA and power writing or reading was around 34 mA.

One can multiply all the mA numbers above by 5V to get mW, so even idle a thumb drive consumes a significant fraction of a watt. As a fraction of the power used by a typical PC, this is small, as one would expect.

A group in Vasser's Physics Department also studied this issue, with interesting results. They are not very specific about what hardware they are testing, though do note that they see little difference between different flash memories. Rather than measuring the amperage used by the flash drive itself, they measure the wattage used by the laptop to which the flash drive is attached.

It appears their laptop consumes about 11.7 W solo, or 13.4 W with a 4 GB flash drive plugged in, which seems low. I wonder if they were measuring the current from the wall socket or the current coming out of the power brick, which would be lower, or something else, like maybe power from the battery. I believe this is what they referred to as "the laptop at rest."

Though the power requirements for their laptop seem low, the extra power required by the thumb drive seems high. An extra 1.7 W? Are we to believe that the thumb drive tested burned 340 mW, which is much higher than the measurements provided by O'Brien, et al. One possible difference is the USB interface within the laptop. Perhaps the internal USB hub is in a low power state when nothing is plugged in.

Interestingly, as the thumb drive is unmounted, the power usage rises to around 20 W. The thumb drive will require additional power when write buffers are being flushed to it. Also, this would require the UI to be active (probably no screen dimming) and might bring the CPU briefly to full speed. I would be interested in the difference seen when ejecting a thumb drive if the CPU is already performing a compute-bound task.

Unsurprisingly, it took an extra fraction of a watt to play a file from hard drive as opposed to from their 4 GB flash drive. The flash drive has no moving parts. However, it's not clear whether the flash drive is plugged in when playing from disk.

Something else to look into: power consumption by SDHC cards as used in digital cameras. Battery life is important for digital cameras.

One caution: the numbers above all appear to be from a small sample of USB drives, which may or may not be representative.

1 comment:

Martin Bucknall said...

I've never thought about how much power a USB flash drive took because it's never been important until I started using a Raspberry Pi. Suddenly 100 mA is quite a lot; I plugged a (mechanical) hard drive and a wi-fi dongle into a powered hub on a Pi and it fell over if I tried to use both, only then did it turn out these two things were consuming pretty well an amp between them! As for flash drives it remains pretty hard to find out how much power they draw before you buy and test them, and indeed even then.