* Posts by Werner Ruotsalainen

14 publicly visible posts • joined 2 Dec 2007

Nokia N95 update speeds apps with virtual memory

Werner Ruotsalainen

maps

In Maps, Options / Use Network; make sure it's "When needed" and not "Never".

Werner Ruotsalainen

@Jasmine

Yup, no doubt writing to Flash isn't as fast as to RAM, albeit current Flash memories can be written at 5-20 Mbyte/s, which is almost on par with current HDD's (particularly on mobile platforms).

As far as power consumption (when writing) is concerned, writing to Flash is indeed power-hungry. However, continuously and actively writing to Flash memory, IMHO, doesn't (wouldn't) *drastically* reduce the battery life of a device.

Back in 2003, I've made several related tests with an HP iPAQ 2210 running on a 950 mAh and recording an MP3 stream directly to a card. I've run the test till the device has disabled acccessing the card. The constant (!), unbuffered writing to the card has turned out to reduce battery life by 5...10% (depending on the card used) as opposed to writing to RAM. That is, it woulnd't have THAT bad an effect on the overall battery life. Certainly less effect than utilizing a HDD / Microdrive.

Werner Ruotsalainen

@Jasmine

Jasmine, thanks for your post, particularly on the flash memory's wearing out. In the Wiki (see http://en.wikipedia.org/wiki/Flash_memory ), the following is stated:

"Another limitation is that flash memory has a finite number of erase-write cycles (most commercially available flash products are guaranteed to withstand 100,000 write-erase-cycles for block 0, and no guarantees for other blocks). This effect is partially offset by some chip firmware or file system drivers by counting the writes and dynamically remapping the blocks in order to spread the write operations between the sectors; this technique is called wear levelling. Another mechanism is to perform write verification and remapping to spare sectors in case of write failure, which is named bad block management (BBM). With these mechanisms in place, some industry analysts[1] have calculated that flash memory can be written to at full speed continuously for 51 years before exceeding its write endurance, even if such writes frequently cause the entire memory to be overwritten. This figure (51 years) involved a worst-case scenario using specific data parameters and should not be confused with a particular "shelf life" for a flash memory device. The bottom line is that a typical user using a commercial device, such as a camera, with a flash drive will probably not wear out the memory for the effective life of the camera."

Are you sure the Flash memory you refer to didn't use wear leveling and/or BBM? Most (if not all) Flash memories do it nowadays.

(BTW, have you checked out my last posts / questions at http://www.reghardware.co.uk/2007/12/01/power_monitor/comments/ ? ;) )

Werner Ruotsalainen

@Emo on GPS

Emo, make sure you enable assisted GPS. Without that, locking is indeed painfully slow.

Werner Ruotsalainen

@Craig: YouTube support

The just-released Flash Lite 3 beta (which, later, will also be included in the v21 firmware version, which is slated for thismonth) is stated to support full YouTube. There're several related threads at AAS and HoFo on it - check them out if interested.

Werner Ruotsalainen

Virtual memory on mobile OS'es

I've published quite a lot of posts on virtual memory. Let me copy my post from http://www.howardforums.com/showthread.php?t=1277861 :

Unfortunately, the OS itself doesn't support paging and swapping. The reason for this is simple: the creators of Symbian did have to reduce the size of the core. The virtual memory controller does eat up a lot of memory and CPU; this is why it's, generally, not utilized in low-RAM devices like Windows Mobile and Symbian ones.

IIRC, the Linux-based but memory-constrained (64M? 128M?) Sharp Zaurus-line doesn't support virtual memory / swapping for exactly the same reason.

This is the main reason for virtual memory not being implemented, not because of the finite writing cycles of flash ROM. The latter isn't an issue because modern flash controllers make sure they "map out" the cells that have gone wrong. That is, they don't fail; all you perceive is a capacity loss over the years.

Apps can of course, implement swapping on their own. A Windows Mobile example: some picture viewers / editors; most importantly, Conduits Pocket Artist; some other pic viewer apps also do file system caching to free up memory, but this caching is VERY slow and, generally, pretty unreliable.

all in all: there isn't and will NOT be real virtual memory. It's a low-memory, low-CPU operating system, after all.

Werner Ruotsalainen

@William Donelson & image sharpening & compression

According to http://www.allaboutsymbian.com/news/item/6340_The_N95s_camera_quality_It_jus.php , much less sharpening & somewhat lesser degree of compression is taking place.

Nokia app monitors juice

Werner Ruotsalainen

@Jasmine

Thanks for the clarification.

I still dont understand how AAC playback can only consume 200 mW - it's way lower than I've measured on the 95-1 - or, for that matter, others measured on their 95-3 using the traditional "let it continuously play back some AAC files until the device shuts down" method (that is, no software metering had taken place). After all, consuing 200 mW would result in the impossible 18 hours of playback, which is certainly not achievable, not even with the N95-3.

In addition, many ave reported (see the benchmarks at http://phone-rush.typepad.com/phonerush/2007/11/n95-generations.html ) that the N95-3 has slightly (about by 7%) higher (!) power consumption than the N95-1. (The figures are really worth checking out.) Any comments on this?

Werner Ruotsalainen

@kevin pownceby

Unfortunately, you just cant expect more of a 120g, snappy (!) all-in-one device - unless you don't use its advanced features, that is. Otherwise, name any 120g 3.5G (HSDPA) device offering so much versatility and having a considerably better battery life.

Nevertheless, the new, v20 firmware has a definitely better power management - power saving can be as high as 45%. You might want to read my above-linked article(make sur eyou follow the links to my earlier ones too), where I've explained this all.

Werner Ruotsalainen

@Stewart Atkins: 5W while playing back video? Ouch!

5W means about 50 minutes total playback... What format was it in? What did you use for playback? I've thrown several, even high-res (VGA native MP4 / DivX) video files at both RealPlayer and the current version of CorePlayer. I've never measured power consumption OVER 1 Watts with the lowest backligt level and moderate speaker volume.

Are you sure the handset wasn't on a charger while doing the tests?

Werner Ruotsalainen

nope, no iPhone

AndyB, nope, it doesn't - it's a Symbian-only app.

Werner Ruotsalainen

real-world tests

Some VERY thorough tests posted to... (these are cross-posts to different forums!):

http://www.allaboutsymbian.com/forum/showthread.php?p=350842

http://www.howardforums.com/showthread.php?t=1286676

http://www.pocketpcmag.com/blogs/index.php?blog=3&p=2338&more=1&c=1&tb=1&pb=1

You’re welcome to comment on them.

Werner Ruotsalainen

@Jasmine

[i]<i>“a) the energy monitor itself burns about a third of a watt (which is an order of magnitude more than the normal quiescent power of the N95 as measured with my handy dandy milliwatt meter.”</i>[/i]

That’s quite a lot. BTW, as I’ve also explained in the article (and, for example, at http://www.allaboutsymbian.com/forum/showpost.php?p=350842&postcount=15 ), the numeric battery life estimation results it provided were all VERY close to the results I’ve measured on my relatively (~2 months old; that is, it’s not worn-out in any way) battery with my “let’s completely deplete the battery to find out how long it lives” approach. That is, even if it indeed consumes a lot of battery, in my N95-1 v20 tests, its results have turned out to be very reliable.

[i]<i> “Typical power dissipation of an N95-3 playing AAC audio at 320 kbits/sec into 32 ohm headphones at maximum volume is around 300 mW with display illuminated and 200 mW with it dark”</i>[/i]

200 mW is VERY low and would result in about 18 hours of continuous playback, which is certainly not the case, not even with the v20 firmware (with AAC, the typical playback time is about 8 hour under the circumstances you’ve described).

I’ve measured the total battery life (with the deplete-it-completely approach) and Nokia’s Profiler several times and I’m absolutely sure the battery life is 8 hours when playing back AAC / MP3 / WMA over the factory headphones at 70% volume (NOT 100% as is in your tests – that is, my setup should consume even less power than yours).

Are you ABSOLUTELY sure your figures are OK?

[i]<i>“while on a rather weak UMTS HSDPA cell (-= 30 mW for GSM/Edge)”</i>[/i]

I recommend my related 3G vs. 2G measurements on both strong and weak signal. With the former, it’s “only” the average power consumption that is much higher than with 2G networks, there’re two entirely separate cases: either the base power consumption (which is exactly the same as with 2G) or the very high peaks (which, on the other hand, are way more than +30 mW). That is, is your +30 mW a time-averaged value (then, it seems to collerate with my results), or, just one in a definite time instance?

Nokia N81 8GB smartphone

Werner Ruotsalainen

Additions

It should also be pointed out that, unlike some of the more expensive models (N93(i), N95), it doesn't have hardware 3D acceleration, which would be pretty much welcome on a gaming handheld.

Second, you state "it has a UPnP media streaming facility for transmitting tracks to a compatible hi-fi." UPnP is able to do a lot more - for example, streaming to YOUR handset from a UPnP-enabled NAS, which is the most widely used usage area of UPnP.