Everything you will ever need to know about the power consumption of Pocket PC audio players

In my well-known Pocket PC & Smartphone Magazine article Maximize Battery Life by Minimizing Power Consumption! and, for example, Pocket PC Thoughts-frontpaged Some new power consumption measurements (Dell Axim x51v, HP iPAQ hx4700, Fujitsu-Siemens Pocket Loox 720, HTC Universal, HTC Wizard), I’ve elaborated on how important it is to reduce the processor (CPU) usage of a given application to gain the best battery life possible and/or force the CPU to run at a lower clock speed.

In the current article, I elaborate on how all the known MP3-capable Pocket PC multimedia players fare in this respect. You will really need to read this article if you regularly listen to for example MP3 files for more than, say, half an hour between recharges. You can save even hours of battery uptime if you choose your multimedia player with the CPU usage in mind. As you’ll see, current multimedia players have vastly different CPU usage (and, therefore, power consumption), particularly if you enable for example equalization, digital signal processing (DSP) and the like.

I’ve previously also made some similar measurements with Pocket PC multimedia players (see for example this article) and sound recorders (see the well-known (see for example these remarks) Pocket PC Audio Recording Bible), but have never compared ALL MP3 players directly to each other, on the same Pocket PC, at the same time. Now, this omission has been fixed.

In the comparison chart (click the link!), I’ve collected (after some really thorough measuring) the following characteristics of each and every MP3-capablePocket PC-based multimedia player.

• Upon creating the chart, I've measured the CPU usage at 208 MHz to directly and reliably compare the CPU usage of each and every application. I’ve chosen such a low CPU frequency to emphasize the CPU usage differences (at higher CPU speeds, the differences would be smaller and more prone to benchmark errors. As can be seen, the, in this respect, there are four clearly separate groups. As with all the other figures in the test, the lower the given percentage, the better.

1. The first group, consisting of three applications, offers, CPU-usage wise, about twice the runtime than the better titles (for example, the built-in Windows Media Player (WMP), LGC Jukebox and GSPlayer in the second group). The best-behaving applications is definitely 40th Floor's iPlay (sporting 11% CPU usage), closely followed by TCPMP and Resco Audio Recorder (12 and 12.5% CPU usage, respectively). Always try preferring these three players if you absolutely need the best battery life!
2. The second group contains many more titles and is started with a brand new title, LGC Jukebox (sporting 20% CPU usage), which is, then, closely followed by the widely-known, excellent, with third party add-on plug-ins, even midi- and mod-capable freeware GSPlayer (21%), VITO AudioPlayer (21%) (when minimized), WMP (21.4%), the (as opposed to all the listed titles so far, not taken iPlay and, partly, TCPMP into account) AVRCP-capable, free and and excellent MortPlayer (22%). Also, some lesser-known titles (for example, the no-longer developed TodayPlayer) are also in this group; so is NoteM, the excellent, free MP3 recorder. (Also note that NoteM is particularly sensitive to skips, which may make it to a non-recommended player in certain circumstances.)
3. The third group consists of titles like WinVibePro (more on this title later!), Conduits' Pocket Player, VITO SoundExplorer, PocketMind's PocketMusic Bundle , iMusic and withMP3. The CPU usage of these titles is between 24 and 27.5% and are definitely less recommended than even the players in the second group, unless you REALLY want to take advantage of the advanced features of, say, Conduits' Pocket Player and VITO SoundExplorer.
4. The fourth (worst) group consists of Nero Mobile (31%), Platform4 Player (37%) and absolutely the worst title, WinamPAQ (40%). Note that WinVibePro should also belong to this group because its CPU usage is pretty tricky and is pretty hard to predict whether it will really "only" consume 24% of the CPU cycles, or, will it consume way more.

• I've also measured the CPU usage with enabled equalizer (I've tried to "cook" the same very-strong-at-highs and slightly-stronger-at-basses with all the players so that they sound at least similarly the same with my Plantronics 590A, which, by default, pretty much lacks the highs) and bass boost, both when available. These are listed in the third column.

  As can be seen, using built-in equalizers (EQ's) definitely raise the power consumption with most (but not all; there are some exceptions like GSPlayer and MortPlayer (they use exactly the same core; hence the minimal additional CPU usage), the no-longer-maintained TodayPlayer and the otherwise absolutely bad WinamPAQ). This means if you use some other player, you may want to consider using built-in, hardware-level equalizer capabilities of your Pocket PC if and only if it supports it AND you don't listen to music via Bluetooth A2DP. In some devices (for example, the HP iPAQ hx4700, the HP iPAQ 2210 (even if the latter only has a really basic bass/middle/tremble setter) in Start / Settings / System / iPAQ Audio), there is already a built-in equalizer; in other devices (for example, the Dell Axim x50(v) / x51(v) or the old Compaq iPAQ 36xx/37xx series, you can get access to them with external tools like x50mix and UdaEq 1.1, respectively. They won’t cause any additional CPU usage, as opposed to software-based solutions.

  Note that the above-introduced x50mix will not only work OK on Dell Axims but also on some other models; for example, the Fujitsu-Siemens Pocket Loox 720. Feel free to give it a test ride on your particular Pocket PC model – it won’t render your Pocket PC a brick. (Note that it didn’t work on the HTC Universal.)

  Except for the, in this test, excellently behaving GSPlayer, MortPlayer, TodayPlayer and the (otherwise, in no way recommended) WinamPAQ, enabling EQ may result in really bad CPU usage increase. This is definitely the case with, for example, PocketMind's PocketMusic (Bundle), withMP3, the non-recommended Nero Mobile, and, finally, depending on the number of points you use (for example, if you only raise the highs with only one point, you can save a lot of battery - but, still, it's better to use another media player if battery life is a concern), VITO SoundExplorer. Note that VITO AudioPlayer, unlike its "big" brother, doesn't have any DSP or equalizer; it, however, sports a (fixed) bass boost, which, unfortunately, is pretty CPU-hungry.

  Also note that the two multimedia players (iPlay, TCPMP) belonging to the "least CPU-hungry" group and also having a built-in EQ (Resco Audio Recorder doesn't support EQ) become much more CPU-hungry when you enable the built-in EQ. iPlay's CPU usage almost doubles, and TCPMP's CPU usage increases by about 40%. This also means you almost completely lose the CPU usage advantage of, say, iPlay if you DO use the built-in equalizer. Again and again, if you use wired headphones, check first if your particular Pocket PC model already has support for system-level EQ settings. (Unfortunately, this, as has already been pointed out, won't work through A2DP.) Alternatively, try to avoid EQ's - remember, real audiophiles (like I used to be) don't use any kind of equalizers at all ;-)

• The fourth column lists a well-known, nice and really useful DSP, reverb. While most reverb (in some players, there are only some similar DSP's like Pocket Player's "echo") implementations (except for that of iPlay) are pretty bad, I've still found this test necessary to find out how much additional CPU load they cause. I was particularly interested in the figures of iPlay, which has a wonderful reverb DSP you'll love to keep enabled. As can be seen, with iPlay, enabling reverb almost triples the CPU usage. This is, however, in my opinion, is a good tradeoff, taken into the quality of the reverb. It should be pointed out that "massive" (maximal) reverb causes the same CPU usage as "more". This means you won't really extend battery life if you refrain from using "massive" reverb effects and stick to lower-level effects.
• The fifth column lists whether the bad side effects of enabled visualization (Real-Time Spectrum Analyzers (RTA's; I'll also refer to this as "spectrums"), peaks) can be avoided by (that is, is the application clever enough NOT to spend any CPU time on the then, invisible visual effects) just shutting down the screen. As can be seen, even some of the top apps (for example, Conduits Pocket Player and, of course, the CPU usage-wise, pretty hopeless WinVibePro; also in this group is VITO AudioPlayer) ignores this and continue to compute for example the spectrum. It's only with withMP3 that the CPU usage decreases in this case.

  Note that I've only filled in this column with a non-n/a value when switching off the screen had the possibility of lowering CPU usage. That is, I have NOT examined plain screen off support in here. This means players with a n/a may also sport screen off capabilities. It's in a later roundup that I'll also elaborate on screen off capabilities not related to CPU usage lowering.

• Finally, the sixth column lists the speed the (XScale) CPU switches to while playing the same 112 kbps MP3 test file (Värttinä: Oi Dai - Oi Dai). The CPU speed setting will also be of interest in addition to the, up to now, 208 MHz-only CPU usage figures. As you'll see, these results are pretty easy to predict based on the results listed in the second (208 MHz, no-EQ/reverb/bass boost CPU usage) column - with a notable exception (WinVibePro).

  Why is this column important? If you don't want to manually force the XScale CPU of your Pocket PC to either 104 or 208 MHz (as opposed to letting the CPU itself switch back to these speeds when there isn't much load), you may want to go right for a program that lets the CPU run at the lowest speed possible; that is, 104 MHz. Currently, there are only two of them: iPlay and, for the most of the time, TCPMP. (Without using the built-in EQ or, with iPlay, Reverb DSP, and, naturally, without additional CPU-intensive tasks like A2DP encoding running in another process. If you do run additional tasks like A2DP encoding - A2DP is VERY resource-intensive, particularly if you use the Microsoft BT stack -, then, the CPU usage of the multimedia player application won't be the only factor for the CPU to decide what clock speed to run itself at.) This is because, say, consuming 22% of CPU cycles at 104 MHz requires LESS power than consuming 11% at 208 MHz, as can also be seen in my past, power consumption-related articles.

  Note that the CPU usage introduced by actively monitoring the CPU speed / load (through Services.exe) with AximSite forum member dogfood’s excellent and highly recommended CPU frequency and load monitor Regtweak, running great on all PXA27x devices, is 1.5% (1.8% with enabling the load monitoring) at 208 MHz and is, therefore, negligible, as far as the test results are concerned. That is, it's highly unlikely the additional, at 312 MHz, ~1.3% CPU usage caused by monitoring is causing the CPU to switch to 416 MHz from 312 and so on.

  Also note that this column is only meaningful for XScale CPU's (it's they that use integer multipliers with the default 104 MHz); Pocket PC's with Samsung (for example, iPAQ 1930/1940/rx1950; HTC TyTN etc.) or the TI (HTC Wizard and a load of other QVGA WM5 PPC Phone Edition devices) CPU's don't really use this kind of automatic speed switching or, if they do, they use vastly different clock frequencies (and steps, if applicable at all).

Other remarks

Note that if you install Platform4 Player 3.0 into the main storage, it'll start playing its default video, MPEG4 by philips.mp4, at starting. This will mess up the screen of the Pocket PC. Therefore, you'll want to manually delete this file from \Program Files\Philips.

It's also worth pointing out that Citsoft's iMusic 2.10 , which is a direct successor to withMP3, seems to be definitely worse, both CPU usage- and capabilities-wise, than its predecessor. For example, it seems it's not possible to switch off the screen from inside the new player, as opposed to withMP3. That is, you may want to stick to the latter, despite its being older.

Note that, as it doesn't have (and I don't particularly like requesting freebies so that I can avoid situations like biting the feeding hand) a trial version, I couldn't test the current version of TCPMP's successor, CorePlayer either. According to the developers (I've talked to them on the matter per e-mail), it has 6...10% less CPU usage than TCPMP 0.72rc1, which may mean it's, now, better than even iPlay. Once again, I haven't tested this.

Also note that the current roundup ONLY tests MP3 playback. For example OGG, WMA or Flac playback is NOT benchmarked in here. In some of my older benchmarks, I've published Ogg-related information too and I'll definitely elaborate on these questions a LOT in my forthcoming, multimedia player-related articles. Stay tuned :)

Finally, the latest downloadable build of the WinCE port of well-known (also see this Wiki entry) VLC wasn't able to play any MP3 files at all; this is why I haven't included it in the chart.

Other recommendations on decreasing CPU usage

As can be seen, visualizations (most importantly, RTA and, with some titles - for example, TodayPlayer and the two VITO players- even peak meters and title display) are one of the worst enemies of battery life. The most CPU-intensive task is displaying equalizer spectrums. Some apps (for example, WinVibePro) don't even notice the spectrum not being visible and still consume a lot of additional CPU cycles - in vain. Therefore, make sure you scrutinize the peak meter / spectrum-related remarks in the chart and either disable them or make sure they are hidden while playing.

The comparison chart

(Note that, should you have problems with the local rendition of the chart, you can also access it HERE):

PlayerCPU usage at a fixed 208 MHzThe same, with (only) enabled equalizer and / or bass boost when presentThe same, with (only) enabled (maximal) reverb when presentDoes screen off / minimizing result in CPU usage decrease related to visualization? What kind of screenoff is available?auto CPU frequency, the player runs the CPU at the following speed (in MHz)
iMusic 2.1026% without and 53% with spectrum (the latter is the default) and visible. With enabled spectrum but invisible state, there is about +5% (31%) CPU usage.Eq: 68% (w/o spectrum)n/aNo built-in screenoff capabilities, unlike in its predecessor208
Platform4 Player 3.037%n/an/an/a (furthermore, there is not even built-in screenoff!) 312
PocketMind PocketMusic Bundle 4.3.4 26.5%EQ: 74% (HUH!); Bass: 35% n/an/a312
WinamPAQ v1.0 Beta build 0509 40% (pretty unreliable and crashes a lot on all my WM5 test devices, except for the Universal. On the Universal, on the other hand, it used 52% CPU time at 520 MHz)Eq: samen/an/a312
TodayPlayer 3.05.00 21% when invisible, 53% when visible. With disabled peak meters, the latter can be decreased to 25%Eq: 22% when invisible; max. bass boost: 28% (-> use the equalizer whenever possible)n/aBad: Nope. (Built-in, assignable to a HW button.) That is, either disable peaks entirely or switch to another app.208 when invisible, 416 when visible. 108 when visible but inactive.
withMP3 1.6024% without and 52% with spectrum (the latter is the default) and visible. With enabled spectrum but invisible state, there is no excess CPU usage.47% (w/o spectrum)n/aGood: Built-in, key-assignable screenoff does decrease CPU usage with enabled spectrum.208 without and 312 with spectrum
Nero Mobile 1.2.0.1231% when invisible, 70% when visible (spectrum can't be disabled / fine-tuned)43% (when invisible)n/aNo built-in screenoff capabilities: in Menu / Settings / General Settings, you can only redefine the home / go / menu and jukebox keys.312 when invisible, 520 when visible
Built-in WMP as of version 1021.4%n/an/an/a208
WinVibePro 4.9.5~24% with enabled ”Turn off spectrum and VU-meter” in Tools / Settings / Visualization (default); after disabling it, 39% even when the spectrum isn’t displayed. With displayed spectrum, 59%, using the default.No equalizer; bass boost is always enabled but can’t be changed29% (the same quality as with Mort / GS, much worse than iPlay)Absolutely bad: Not even (built-in) screen dimming does reduce the CPU usage when spectrum is enabled.416, even in background
iPlay 4.8 (the last version as of October 2006; iPlay 5.x is no more WM2003(SE)-compliant)11%n/tn/tn/a104
iPlay 5.0 (the last, WM5+-only version as of January 2007)11%Eq: 20%37%; at “more” (as opposed to “massive"), the samen/a104
Conduits Pocket Player 3.0124% with fully disabled visualization; 27% with the default one and 28% with spectrum (switchable in More / Visualization) even when not visible(with default visualization) Eq: 29-30%; bass boost (with the default settings): 35%. That is, stick to the equalizer and leave the bass boost DSP untouched.No reverb; echo sounds pretty awful; it consumes 30%Built-in screenoff; bad: doesn't decrease CPU usage208
LGC Jukebox 2.1020%Eq: 25.4%n/an/a208
TCPMP 0.72rc112%Eq: 17%n/an/a104 most of the time, sometimes switches to 208
GSPlayer 2.25 21%with both enabled, 27.5%; bass boost only: 26.7%; equalizer only: 21.5% (that is, prefer using the equalizer to bass boost!)(it has appr. the same quality as MortPlayer) 26%n/a208
MortPlayer 3.31RC622% even when minimized Bass boost: 27.4%; eq: 22.4%(not very good but still better than that of Pocket Player) 27%n/a208
VITO SoundExplorer 1.8 27.5%; Today plug-in is as CPU cycle-consuming as before: it uses about 25% even when inactive and keeps the CPU at 520 MHz. Eq: 69% (with three points); 41% (with one point)n/an/a416 when in fore- and 312 in background (because of animation)
VITO AudioPlayer 1.15 46% when visible (including, unfortunately, the shut down screen state); 21% when not visible (AudioPlayer minimized / sent into background); animation can’t be disabledSuperbass: 42% (it has no EQ, as opposed to its big brother, VITO SoundExplorer)n/aNope, unfortunately: the visualization-related excess CPU usage is present even when the screen is (auto)shut down Jumping between 416 and 520 when visible, jumping between 208 and 416 when invisible (minimized)
Resco Audio Recorder 3.21 12.5% (10 (main exe) +2.5 (device.exe))n/an/an/a312 when in fore- and mostly 208 (sometimes 104) in background (because of fine track title scrolling)
NoteM 1.21 22%; heavily skips if there is anything else running in the background or you switch tasks / bring up the Start menun/an/an/a208

Verdict

My personal pick / recommendation is iPlay. I only wish I could find ANY way of purchasing it. I’ve been trying very hard to find a proxy server so that I can visit the homepage of the developer; so far, without any success. The developer is also ignoring my e-mails and public messages to contact me back. It seems the developer considers me because I'm unable to pass to his homepage a "dark alley guy", as he often puts it about everyone that complains about being banned out of the website of 40th Floor. A nice way of handling would-be customers, eh? ;) (I also recommend THIS and THIS AximSite threads on this.) Nevertheless, iPlay is still really worth checking out and is highly recommended - assuming, again, that you can get even the trial version.

Cross-posted to: PPCT, MobilitySite, AximSite, XDA-Developers, FirstLoox, BrightHand, HowardForums, PDAMania

UPDATE (01/31/2007): in the meantime, I've tested x50mix on some of my other Pocket PC's. It turned out to support the Fujitsu-Siemens Pocket Loox 720 flawlessly; this may mean it also supports other Fujitsu-Siemens models. That is, it's not only for the newer Dell Axim devices.

It, unfortunately, doesn't support the HTC Universal.

Give it a DEFINITE try on your particular model and let me know whether it works on it or not!

UPDATE (02/03/2007): PPCT frontpage; well-known AximSite e-books admin akheron has announced he'll add this article to the AximSite e-books section.

iPlay homepage accessibility-wise, I've also tested Skweezer, MobileLeap, Google Mobile and WebWarper (see THIS for links) - a no-go. 40th Floor seems to long have been disabled access via these services - as they have done with http://anonymizer.nntime.com/ posted some days ago by one of my blog readers, CaUcHy. It seems the 40th folks are monitoring my articles and also readers' comments - but, of course, they don't try to contact me. Oh well.

The page is accessible via http://anonymizer.su/ but you can't download anything if you use this service.

UPDATE (03/30/2007): some additional iPlay-related debates.

UPDATE (04/09/2007): and yet others

UPDATE (10/31/2007):

1. HTC has also released a hardware equalizer compatible with most (but NOT all!) Pocket PC Phones
2. I've started working on the long-awaited Multimedia Bible! It'll kick some serious butts, I promise :)

UPDATE (11/17/2007): x50mix is available HERE, as opposed to the above link, which has gone dead in the meantime