PDAgold.com on Memory in Pocket PCs

[Janak Parekh]

<div class='os_post_top_link'><a href='http://www.pdagold.com/articles/detail.asp?a=243' target='_blank'>http://www.pdagold.com/articles/detail.asp?a=243</a><br /><br /></div><i>"A vast majority of PDA owners have some experience with personal computers and know at least something about the basic elements, such as the memory. However, if such a user tries to apply the knowledge on pocket computers, he may be rather confused. This article explains the basic terms, describes different types of memory used in pocket computers and main differences between them."</i><br /><br />Veteran users probably know all there is in the article, and Pavel doesn't go into the details of WM5's persistent storage features. However, for those of you who have recently obtained a new Pocket PC, this is an excellent overview of both the hardware and software aspects of both RAM and ROM and what it means for you.

[ipaq_wannabe]

so, what happens to the memory if you have exceeded the number of writes/rewrites?

[Gerard]

I suspect that when the write limit of flash memory is reached, or approached and errors start becoming more than irritating, it becomes time to 'upgrade' to a new device.

I've been waiting for someone, at Microsoft or elsewhere, to come out and offer some reassurance that ROM being the main area for saving temporary and more permanent (if perhaps often updated and re-saved) documents is 'safe' or at least reliable enough for, say, 5 years' use. I realise that 5 years is a long, long time for most of us. My X5 is getting long in the tooth, and it's still far, far less than that old. My Casio EG-800 is still somewhat functional and my kid uses it several times a week, and it's not quite 5 years old.

But I do run into people who are still using the first series of iPAQs, and the odd Jornada or Casio from 2000 as well. These kinds of users are the more average sort, not the limit-pushers and early-adopters many here are. And their needs should be honoured.

So, is flash ROM a safe place for daily use for 5 years, including web browsing temp files, email, documents, installed software, and all the many other bits and pieces which make up the write-life of an average PPC? Or will it fall apart, leaving us with paperweights in shiny, almost-new shells? If the latter, I guess a big wallet will be the order of the day, and screen protectors will hardly be any more relevant than backup software.

[ipaq_wannabe]

so, how long could one characterize a 100,000 rewrites memory?

my handheld is an Axim x50v - and they say that the BIS has about that approximate amount of rewrites -

so, assuming that you work on 10 files a day (stored in BIS)...

assume that you save the file 10 times a day per use (those periodic saves)...

that would give us about 2.7 years of a lifetime, right?

which is basically the average time most users change handhelds - i usually upgrade in about between a year and a year-and-a-half...

but dont you think manufacturers should give us more value for our money? most HDDs have about years/decades of MTBFs...

[Janak Parekh]

No, it's not so bad. Most flash memory controllers optimize the writes to be over the entire space of the media, so that if you rewrite a file, it may actually go somewhere else so that no one space on the flash is rewritten too much. The last time we did calculations on the board, we found that for most people it would exceed 5+ years of active use (IIRC; they weren't scientific, though).

--janak

[surur]

Memory card corruption is an increasingly common problem. How many post down we see that goes "Everything on my card has disappeared!".

We all blame the card manufacturer and the OS, but I wonder whether we are not running into this fundamental problem more often than we think.

Surur

[ipaq_wannabe]

No, it's not so bad. Most flash memory controllers optimize the writes to be over the entire space of the media, so that if you rewrite a file, it may actually go somewhere else so that no one space on the flash is rewritten too much. The last time we did calculations on the board, we found that for most people it would exceed 5+ years of active use (IIRC; they weren't scientific, though).

--janak

could this be seen somewhere? URL maybe?

[Menneisyys]

So, is flash ROM a safe place for daily use for 5 years, including web browsing temp files, email, documents, installed software, and all the many other bits and pieces which make up the write-life of an average PPC?

Yes and no. With devices like the iPAQ rx3xxx, or, to a much lesser degree, the hx4700 (see for example this thread (http://discussion.brighthand.com/showthread.php?s=&amp;threadid=113388)), PIE (and other) cache shouldn't be put in the File Store. Not only wiull it become very-very-very-very-very slow (http://www.pocketpcthoughts.com/forums/viewtopic.php?t=36376), but also these devices, especially the Samsung flash controller in the iPAQ rx3xxx, will be damaged from even several hundred (!) write cycles.

Some even say (see lorettaboy's posts here (http://www.pocketpcthoughts.com/forums/viewtopic.php?t=39971&amp;postdays=0&amp;postorder=asc&amp;highlight=rx3715&amp;start=10)) that HP doesn't release the much more write-to-the-filestore-instensive WM5 upgrade for the rx3700 series because of this:

"There is a very good reason for HP not to upgrade the rx3700 series to WM 5.0. Have you all forgotten about the filestore problem? A decent percentage of all rx3700's have this problem, I should know, I had 3 rx's in a row die because of this. And since WM 5.0 will be using the filestore as the primary data store, having it fail would be detrimental. I honestly don't know why HP has not pulled this device yet.
So, I think it is easy to see why it is in HP's best interest not to release an upgrade for this particular model."

[Menneisyys]

No, it's not so bad. Most flash memory controllers optimize the writes to be over the entire space of the media, so that if you rewrite a file, it may actually go somewhere else so that no one space on the flash is rewritten too much. The last time we did calculations on the board, we found that for most people it would exceed 5+ years of active use (IIRC; they weren't scientific, though).

--janak

could this be seen somewhere? URL maybe?

I can second this: I've also worked on SmartCard (which are based on EPROM) projects as a developer/planner. Even in the (comparatively to a Flash ROM memory card) very dumb SmartCards it's possible to 'map out' memory areas that, over time, became faulty. Flash cards (and file stores - not those of Samsung, seemingly :devilboy: ) work the same way.

I've been using flash ROM cards actively (writing all the time - for example, constant MP3 recording) for long years for my digicams and PDA's. None of them have broken on me. This means their protection and dynamic remap circuitry is working OK.

[Gerard]

Thanks a lot for the clarifications Menneisyys. This confirms my feeling about waiting, not buying an X50v nor a Loox 720, as it is yet unknown whether these devices' flash ROM will thrive under WM5.0. I'd rather sit it out, see what the newer models behave like and also how well those two upgrade. Not being anything remotely 'wealthy' I just can't afford a bad purchase.

[davea0511]

If you're working every single day on enough files on your PDA to result in 100 automatic file backups per day, then you seriously need to get a life. If you don't then what do you have to live for?

Even the agressive PPC user wil probably average no more than 10-20 automatic file backups / day.

Also 100,000 writes is the guaranteed minimum writes. The actual figures are closer to 300,000 on average for most flash drives.

Given these two figures you're looking at close to 50 - 100 years before you'll have reliability problems due to gate oxide breakdown (the failure mechanism which rewrites cause).

It's true that flash drives have a history of being glitchy, but that has more more to do with living on the bleeding edge of poor design and manufacturing process variations than anything else. Gate Oxide breakdown in flash drives is a very rare occurence for chips that receive normal processing.

[Gerard]

I have visited 5 forum pages since last clearing my Pocket IE cache, including this reply form. All have been very short pages, three being very simple ones on pocketpcfaq. Checking my temp folders I find a total of 176 new files, with 14 older cookie TXT files from my prior browsing sessions - I rigorously delete unwanted cookies, manually, from all but frequently visited sites where I just don't want to waste time logging in.

So multiply that by... approximately 50 or so, and you've got my average daily browsing of around 200+ pages, in various forums, news sites, etc. Used to be more but I use an RSS reader now. That's likely to be close to 1000 files written to RAM every day, just browsing in Pocket IE, never mind all the other stuff I do on this device like taking pictures with CECam and my FlyCam, dealing with email in nPOPw, saving and working on new documents...

As you can see, I use my PPC as my primary computer. I rarely boot my PC. I think the last time was 8 or 9 days ago. So my worry is genuine regarding this new location protocol for file storage under WM5.0. If they have not taken users such as myself into consideration, I suppose we'll have to start thinking of other OS types, maybe something along the lines of a micro-PC or the Sharp Zaurus line. I don't want to have to give up on the PPC. Buggy as they are, I like the things. But if they have severely limited lifespan, with no user-swappable ROM chip in case of over-use failure, than there's not a lot of point in investing in such ridiculously short-sighted technology.

Of course, I'm not saying it is doomed. I am only asking. And apparently the question is not all that clear? What I ask is; can a user who needs stable device function for at least a couple or a few years expect to be able to do 1000+ file writes per day under WM5.0, on whatever device? Or not? It seems many here do not browse from a PPC with NetFront or PIE or ftxPBrowser or whatever, and so those users are causing dramatically fewer files to be written on their devices. I am not asking about such limited use. I am asking about using the PPC as a primary communications/capture/data processing device.

[Janak Parekh]

I can second this
Thanks, Menneisyys. :D This is not my area of expertise, I was just basing it on what I overheard.

--janak

[davea0511]

I have visited 5 forum pages since last clearing my Pocket IE cache, including this reply form. All have been very short pages, three being very simple ones on pocketpcfaq. Checking my temp folders I find a total of 176 new files, with 14 older cookie TXT files from my prior browsing sessions - I rigorously delete unwanted cookies, manually, from all but frequently visited sites where I just don't want to waste time logging in.

So multiply that by... approximately 50 or so, and you've got my average daily browsing of around 200+ pages, in various forums, news sites, etc. Used to be more but I use an RSS reader now. That's likely to be close to 1000 files written to RAM every day, just browsing in Pocket IE, never mind all the other stuff I do on this device like taking pictures with CECam and my FlyCam, dealing with email in nPOPw, saving and working on new documents...

As you can see, I use my PPC as my primary computer. I rarely boot my PC. I think the last time was 8 or 9 days ago. So my worry is genuine regarding this new location protocol for file storage under WM5.0. If they have not taken users such as myself into consideration, I suppose we'll have to start thinking of other OS types, maybe something along the lines of a micro-PC or the Sharp Zaurus line. I don't want to have to give up on the PPC. Buggy as they are, I like the things. But if they have severely limited lifespan, with no user-swappable ROM chip in case of over-use failure, than there's not a lot of point in investing in such ridiculously short-sighted technology.

Of course, I'm not saying it is doomed. I am only asking. And apparently the question is not all that clear? What I ask is; can a user who needs stable device function for at least a couple or a few years expect to be able to do 1000+ file writes per day under WM5.0, on whatever device? Or not? It seems many here do not browse from a PPC with NetFront or PIE or ftxPBrowser or whatever, and so those users are causing dramatically fewer files to be written on their devices. I am not asking about such limited use. I am asking about using the PPC as a primary communications/capture/data processing device.

Those temp folders are really in your RAM though aren't they? Isn't your PocketIE Cache only in RAM? I don't think it's ever going into your ROM unless you have those pages all set up for offline browsing.

Also, it seems like you can set up most programs so that they'll save stuff like pictures and all files (temporary and permanent) on an external flash card with the RAM acting only as a cache - but your ROM again should be untouched.

So I think even in your scenario the ROM will still only get a few writes per day. Here's how you can see if your ROM is getting written when you simply do browsing and taking pictures without saving them (example: send in an email instead): do a hard reset. If that stuff is still there then you're right - it is getting saved to Flash ROM. I can pretty much guarantee you though that it won't be there because it's only in RAM.

[Menneisyys]

Those temp folders are really in your RAM though aren't they? Isn't your PocketIE Cache only in RAM? I don't think it's ever going into your ROM unless you have those pages all set up for offline browsing.


alex_kac here (http://www.pocketpcthoughts.com/forums/viewtopic.php?t=39977&amp;postdays=0&amp;postorder=asc&amp;start=80) (see his post at Wed May 11, 2005 8:57 pm) have stated the OS in WM5 has been greatly optimized to read/write data to the filestore. So, it may have a completely new way of writing even PIE cache files into there.

[surur]

In a podcast at Msmobiles.com (75 or 72 I think) they talk about how much slower writing to ROM is, and how they had to work around this. The solution was to buffer writes, and then do them all at the same time, i.e. instead of doing 4 individual writes, they queue them up and do 4 at once, for a great improvement in speed. It appears they had to do a lot of optimisation to get it to work acceptably. It does cause me some concern however, as I can see that a device that locks up can lead to the loss om many retained settings and a corruption of still open files. I assume win ce does not used a journaling file system, so I'm a bit worried.

At least it sounds as if the beta program was quite large (1900 users) and that they were quite happy with the OS, so I hope end users will not have many issues.

Surur

[Menneisyys]

The caching-related discussion has been continued here (http://www.pocketpcthoughts.com/forums/viewtopic.php?p=354822).