View Single Post

Old 03-01-2008, 06:51 PM
Menneisyys Menneisyys is offline
Registered User
Join Date: Oct 2007
Location: Europe
Posts: 433
Menneisyys is on a distinguished road
UPDATE (01/03/2008): as an addition to bullet 3 of section “1.2 When you prefer optimizing an, otherwise, working stream...”, I’ve made some 3G vs 2G power consumption tests on the HTC Trinity / P3600 running the official February 2007 ROM version so that I can prove that the “streaming over 3G consumes way more power than streaming over pre-3G connections, even at the same bitrate” rule-of-thumb is right with every single chipset used in contemporary handsets. The results speak for themselves:

The first section shows playing back a stream over HSDPA, the second shows playing over exactly the same HE-AACv2 stream over GPRS, in TCPMP.

Let’s do some math: ~500 mA / ~280 mA = 78%! That is, if you stick with 3G on the Trinity for streaming, your battery life will be about 80% worse!

CPU usage vs power usage

Still looking at the above screenshots, the third (and all section after that) shows switching from the more CPU-intensive HE-AACv2 stream to an MP3 stream (still using GPRS – as in the next benchmarks) to find out whether the decreased CPU usage results in any change in the power usage. As can clearly be seen, there’s not much difference – they’re unmeasurable.

In the following screenshot:

I’ve enabled A2DP audio transfer to stereo Bluetooth headphones. A2DP, for some (for me, unknown) reason, is very CPU-intensive on all Samsung-based Windows Mobile devices. As I’ve already stated in some of my articles (see for example THIS), it raises the CPU usage by about 50 (fifty) percents. (For comparison, at 400 MHz, the same Microsoft A2DP implementation “only” results between 10...20% of CPU usage increase on Intel Xscale (pre-PXA-3xx, where I still haven’t tested this) and on TI OMAP CPU’s.) Therefore, switching on A2DP is a great way to quickly “bump up” the CPU usage. In the first (short) test, I’ve continued playing the same MP3 stream (for quite a short time; therefore, the (relative) results aren't very dependable); in the second (much longer) test, I switched back to the original, more CPU-intensive HE-AACv2 stream. Then, the CPU usage was constantly at 100% (again, "thanks" to the huge CPU usage of A2DP on all Samsung-based Windows Mobile handsets) and the audio started to stutter. As can be seen, there was no measurable increase in power usage, compared to the case of only about 30% CPU usage.

This certainly shows the Samsung CPU’s do have some strengths; albeit, in many other ways, they’re definitely behind the current crop of CPU’s. For example, they still use the ancient ARMv4T architecture (approximately the same as that of the 9-year-old Intel StrongARM - yes, the CPU used in the first, h31xx/h36xx-series iPAQ's released in May 2000), while
  1. even the oldest (much older than those of Samsung!) Intel PXA-series CPU’s are already ARMv5TE,
  2. let alone the latest Texas Instruments OMAP2420 generation (used in, for example, the Nokia N95 or, as far as Windows Mobile devices are concerned, (only) the Samsung SGH-i616, BlackJack II and the Moto Q9h Global – note that the Texas Instruments OMAP 850 Windows Mobile phones almost exclusively used in HTC’s “lower-end”, and the OMAP 1710 in some MS Smartphones are still ARMv5TEJ only)
  3. and the Qualcomm MSM7200 (used in many contemporary, higher-end HTC devices - HTC Kaiser etc.), which are ARMv6. That is, two generations more advanced than those of Samsung.