Prefetch/Undefined Instruction/Data Abort Errors... cheap refurb as cause?

[jwalker1196]

So I just bought a refurb'd 5.5gen 80gig. Loaded music, started running into those crazy data aborts, undefined instructions, and prefetch errors. So I got thinking, did a little playing around, and arrived at the following.

Idk if there are 2x32MB RAM chips and one is broken, or if the punk amazon storefront I bought it from repaired mine using 5th-gen hardware, but as soon as I add enough songs to my playlist to get the buffer around 28 - 29MB, it does the little blip of loud static and displays one of those errors. (I'm sure the remaining 3MB or so is the directory cache, the program, etc.)

I'm assuming I can install the 32MB build, but that'll defeat 1/2 the purpose of this new ipod (better battery life).

Is my train of thought correct on all above counts? Someone with expertise on the internals and what is actually possibly going on would be sweet. I plan on contacting the seller and demanding one with the correct hardware, so if this is just a rockbox bug that I can fix that'd be sweet to know, before I go postal on an amazonian

(Their refurb sucked anyway... 12 dead pixels, the screen color balance is effed anyway (cheap LCD) and the click wheel is sticky and "squeaky".)


--edit: also idk if this would really go better in hardware... lemme know, feel free to move it, etc etc

--edit 2: went to install the 32MB one, and just out of curiosity, hit autodetect... it gets detected as a 30-gig. Relevant? It works fine, btw. I'm sure with the smaller cache my battery life will be garbage.

[gevaerts]

Refurbished ipods with only 32MB are pretty common, so this is very probably your problem.

Autodetect currently doesn't detect RAM size, soit always installes a 32MB build to be safe.

[GodEater]

On the plus side for you, the difference in battery life between a true 64MB iPod and a 32MB one are neglibile.

The 64MB iPod spins the disc up less often, true, but it has to spin it up and use it for longer to fill the bigger buffer, so I doubt you'll notice much difference.

[jwalker1196]

On the plus side for you, the difference in battery life between a true 64MB iPod and a 32MB one are neglibile.

The 64MB iPod spins the disc up less often, true, but it has to spin it up and use it for longer to fill the bigger buffer, so I doubt you'll notice much difference.

The battery life times between the 5th gen and 5.5 is pretty dramatic... like 12 hours vs. 18. Particularly when playing FLAC and such where it might have to read more than once per song even. And anyway it's not maintaining the 4200rpm that's rough on the battery, it's the acceleration that's the power draw. Having to do that 1/2 as often will help a ton.

If it's not the RAM size that affects this, what is? Since I seem to have a 5th gen main board, I can only assume I'll have 5th-gen battery times. Or is the battery in the 5.5-gen also of larger capacity?

[AlexP]

The RAM doesn't make that much difference.  The 64 MB model has a bigger battery.

[jwalker1196]

The RAM doesn't make that much difference.  The 64 MB model has a bigger battery.

Ah I see. Well then testing will be in order to see just how pissed I really am... certainly it'll make a difference with high-bitrate files, ie FLAC and video files, I guess I will have to see just how much. Unless someone can direct me to some numbers. What's that plugin that monitors runtime? Is it the one that's already included?

I think I'm just gonna start doing my own repair work. About how hard is it to get my hands on a 64MB main board? I've got no idea where a good place to go for that sort of thing... ebay but that's always a bit sketchy for parts of anything... advice?

[LambdaCalculus]

About how hard is it to get my hands on a 64MB main board? I've got no idea where a good place to go for that sort of thing... ebay but that's always a bit sketchy for parts of anything... advice?

There's always iFixit.

[GodEater]

The battery life times between the 5th gen and 5.5 is pretty dramatic... like 12 hours vs. 18.

What is your source for this information?

[jwalker1196]

The battery life times between the 5th gen and 5.5 is pretty dramatic... like 12 hours vs. 18.

What is your source for this information?

Well, Apple's own claimed battery life is as follows:
5th Gen 30GB:  14hr audio, 2hr video
5.5 Gen 80GB:  20hr audio, 6.5hr video

This info available many places, hidden on apple's servers somewhere I've seen, but readily available via Mactracker.

But yeah we all know how the real world vs. claims works, so I pulled 12 and 18 out of my own ass, but I doubt I'm too far off. Note use of the word "like"... lol.

So anyway the question will be of course how much of that is cache-size related, vs. the larger battery. I doubt it's negligible. Note the 350% increase for video, vs. just 143% increase for audio. And since Apple's idea of audio is 128kbps AAC, and I play mostly FLAC and high-bitrate (~500kbps) OGG, my mileage will certainly vary, and not in a good way. So honestly for me, on my old 30GB gen 5, I saw like 6 or 7 hours of audio playback. I was hoping the 20 might turn into like 15, which would be a nice improvement.

I'd be glad to track my battery runtime over several cycles of "average use" (for me) and post if anyone is curious. And again after (if) I replace the main board with the true 5.5gen one.

[torne]

The original batteries are 400mAh and 600mAh respectively, so a 50% increase in battery life is exactly what you'd expect.

Battery benching my 5.5G with rockbox gives me 21 hours playing fairly high bitrate AAC (forget what exactly) but it depends a lot on codec and settings (some of the gain there is because i have it set to turn the lcd off entirely, which makes quite a bit of difference).

[jwalker1196]

About how hard is it to get my hands on a 64MB main board? I've got no idea where a good place to go for that sort of thing... ebay but that's always a bit sketchy for parts of anything... advice?

There's always iFixit.

Very cool. Good stuff all around, appreciate the link. However, they don't sell a logic board with 64MB RAM. They sell only the 32MB ones for repairing the 60/80GB ipods. So... ebay? Or anyone know of a better option?

edit: if anyone else is has the same problem and wants one... i found one place, etechparts sells it. Turns out I won't need to, see bottom edit.

The original batteries are 400mAh and 600mAh respectively, so a 50% increase in battery life is exactly what you'd expect.

Battery benching my 5.5G with rockbox gives me 21 hours playing fairly high bitrate AAC (forget what exactly) but it depends a lot on codec and settings (some of the gain there is because i have it set to turn the lcd off entirely, which makes quite a bit of difference).

Nice. So when I poke around ifixit I notice they sell 850mAh batts for replacement. So I guess I could get some extra playtime just by snagging one of those as well. And of course, there will still be a world of difference between "high-bitrate" AAC, which is at most 320kb/s, versus a 1Mb/s+ video or FLAC.

And 21 hours is pretty sick... do you have the 850mAh battery then?


UPDATE:
I found somewhere to buy the 64MB logic board, see link a few paragraphs up, but I won't need to. I took a chance and told ipodzens, the e-store I bought it from, straight up that I use rockbox and that I bought the 80GB version in large part for the cache size, and the rep was real cool about it, and we agreed that your average end user it wouldn't matter but for (his words) "a power user" like me it clearly would.

(And I personally feel that if you're too oblivious to know the difference, I don't really care that they cut the corner. I once had an idiot friend who heard me extoll the virtues of the all-wheel drive, turbo 1998 Eclipse GTX as a hidden gem in the tuner world, and next thing I knew he'd bought a regular FWD GS model, and never knew the difference. Be knowledgeable or expect to get screwed alot in this world.)

Anyway, they gave me an return slip, and it'll ship both ways free to me, and they'll swap in a 5.5-gen board free also. So all things considered, not too bad since I only paid like $150 for the thing anyway.


UPDATE 2: Thought I'd mention the runtime. I'm getting about 11 - 12 hours, exactly the increase I'd expect with 32MB RAM... +50% (600mAh batt instead of 400) but minus a bit for the mass of spinning up the extra platter of the thicker 80GB drive. (Little factoid you dudes probably know already... the 80 and 60 have dual platter 1.8" drives, while the 30's are single.)

[Llorean]

You seem to be under the misconception that only 5.5G had 64MB of RAM

The difference was parallel to the HD size - 30GB players all had 32MB while 60 and 80GB players (5Gs and 5.5Gs) had 64MB of RAM.

RAM can make a difference on the battery life, but it's not nearly as significant as you seem to think it will be. The most significant things are codec used (time the CPU is running at full power to decode), backlight usage, and the actual size of your battery.

[jwalker1196]

You seem to be under the misconception that only 5.5G had 64MB of RAM

The difference was parallel to the HD size - 30GB players all had 32MB while 60 and 80GB players (5Gs and 5.5Gs) had 64MB of RAM.

RAM can make a difference on the battery life, but it's not nearly as significant as you seem to think it will be. The most significant things are codec used (time the CPU is running at full power to decode), backlight usage, and the actual size of your battery.

I know. I'm comparing the 5th gen 30 gig to the 80 gig 5.5 for two reasons. First, I've owned and run rockbox for years on both. Also, the 80 was only available for the 5.5 gen.

And, having now spoken personally to a couple members of Apple's own design team, I can say with 100% confidence that the RAM does in fact have a large effect on battery life... if you consider bitrate. Which they did during the design process. They knew the direction the consumer world was making toward higher bitrate, partly because larger storage allowed it, and partly because the information that higher bitrate = higher quality has been slowly filtering to the masses. With higher bitrates becoming more ubiquitous (and Apple's plan for 256kbps AAC in the iTunes Store to match that), the battery life would indeed suffer on HDD-based ipods, so they made a defensive move before they could be charged with being overly deceptive on battery life, and came up with the RAM solution.

They made it available on the late-coming 60GB 5th gen and then the 80GB 5.5 because their consumer polls showed the self-professed prospective owners of a high-capacity model would have libraries larger not in number of songs so much as just the space these songs take up. In addition, they realized if there WAS a larger library in filenumber, it would affect battery life too, for the same reason, as the actual control files get stored in the RAM as well, taking up all the room for music caching. Indeed because rockbox doesn't operate that way, that helps diminish the problem; however the bottom line is it still comes down to how many disk spin-ups you have to perform, and 1/2 as many is a very good thing.

Added to the focus on video, which is going to be at least 500kbps, and you had the dismal battery life of the 30GB 5th gens for high-bitrate playback.

Aside from the fact that I have it from a primary source that music caching was the main reason for the larger RAM (and if there was not much performance benefit as you all keep insisting, why would they implement it at all? It's not even an advertised feature, so surely it must serve some benefit), it just follows logic. When analyzing the design of the Macbook Air and it's SSD, I've seen graphs of the power draw of an at-speed vs. upspinning 1.8" drive. It's a BIG difference. It was something like it draws the same amount of power for the 300ms or whatever for spin-up that it does during 15 seconds at speed. (Also SSDs were the clear winner, powerwise, obviously.)

Our own designers considered this effect when they wrote the default spin-down time in rockbox at what it is. I'm a hardware engineer, please don't insult my intelligence and tell me ipods defy both logic and physics. Save that for the religious. Or software guys. The RAM plus the battery is the only difference between the 2 hours of video playback at 1Mbit/s on a 32MB and the 6.5 on a 64, and the battery's only 50% bigger. And during video playback the CPU's pretty much going full-tilt. Put the 600mAh batt on the 32MB one and you'll get about 3. (Which is exactly what the 5th gen 60GB had.) So now the difference between 3 and 6.5 is... the RAM.

Feel free to continue installing the 32MB version on your 64's if you're so convinced. I may also go pull one of my blowers out of my TT. Surely it won't matter too much, right?

Thank you all for the valid info, and the links, etc. If I can get some insider specs on a Classic or Touch, I'll be back in a flash.

edit: de-dickified my post a bit... I've been awake for 36 hours, the guys in my hardware design team just want to argue rather than do work on our barely-on-schedule project, and I shouldn't be taking it out on the forum.

[Llorean]

You do realize that it's possible to do battery benchmarks with half the RAM on any player and demonstrate that it halves playback time (which is what you seem to be claiming it does). You'll notice it's quite possible to get 21h of playback in Rockbox on a 32MB iPod with a stock battery (with a few patches that may not end up committed for other reasons).

Meanwhile on the Archos Players, having 8MB of RAM vs the initially installed 2MB did increase runtime, but it didn't quadruple it as your "3 hours vs 6.5 hours" comparison suggests it should.

So far all real battery benchmarks contributed haven't shown it as being as significant a difference as you think it ought to be. You seem to take me saying "it's not as significant as you seem to think" to mean "it doesn't make a difference." Nobody denies it does, and nobody denies the difference will increase as bitrate increases.

As to your misconception:

Since I seem to have a 5th gen main board, I can only assume I'll have 5th-gen battery times. Or is the battery in the 5.5-gen also of larger capacity?

If you knew that some 5th gens have 64MB of RAM, why are you referring to them as "5th-gen battery times"?

Please, come back with actual battery benches across multiple bitrates showing the differences in 64MB of RAM vs 32MB. But don't complain about your willing misinterpretations of what I've said. I said it clearly would affect things, and only said it wouldn't be as significant.

It's also rather frustrating you've also willingly misinterpreted what I said in another thread to fuel your misplaced anger. I clearly did not tell him his old build was the cause, I only told the him that he should update before posting (which is mentioned in the forum guidelines and a requirement, independently of whether it might be the cause).

If his problems occur in the same files every time, as he said they did, given the history of problems our users have had the most likely problem is a drive with errors.

Instead your first suggestion is that he may be using an audio format we don't actually support (Vorbis doesn't support SBR, so either we wouldn't be able to play the file period, or it would be like MP3+ where we ignore the SBR data).

Your third suggestion is even dumber, frankly, suggesting that his boost got locked "low." This can't persist across reboots (as debug values are not saved). Not to mention 24mhz is barely enough to run the WPS, and given his descriptions of the problem (including mentioning the files are 192kbps Vorbis) he almost definitely wouldn't be able to play any file he's tried realtime.

So rather than offering truly useful suggestions to him, what you did is propose two ideas that add confusion into the mix, while mentioning my suggestion is worth checking out (then complaining at me for the same suggestion elsewhere). Of course you did throw in a mention that defragmenting might be useful, once more demonstrating you lack of understanding. See, fragmentation would affect it as it's loaded, but not during playback. Heavy disk fragmentation could make the buffer take longer, sure. But once it was buffered it would be defragmented in RAM - this means that what songs skipping would occur on would depend entirely on playlist order. The user said it's always the same songs, so if his statement is truthful it wouldn't be an issue of fragmentation.

I do appreciate you don't like me, but at the very least, please try not to give stupid advice that will just confuse users.

[jwalker1196]

You do realize that it's possible to do battery benchmarks with half the RAM on any player and demonstrate that it halves playback time (which is what you seem to be claiming it does). You'll notice it's quite possible to get 21h of playback in Rockbox on a 32MB iPod with a stock battery (with a few patches that may not end up committed for other reasons).

Meanwhile on the Archos Players, having 8MB of RAM vs the initially installed 2MB did increase runtime, but it didn't quadruple it as your "3 hours vs 6.5 hours" comparison suggests it should.

So far all real battery benchmarks contributed haven't shown it as being as significant a difference as you think it ought to be. You seem to take me saying "it's not as significant as you seem to think" to mean "it doesn't make a difference." Nobody denies it does, and nobody denies the difference will increase as bitrate increases.

As to your misconception:

Since I seem to have a 5th gen main board, I can only assume I'll have 5th-gen battery times. Or is the battery in the 5.5-gen also of larger capacity?

If you knew that some 5th gens have 64MB of RAM, why are you referring to them as "5th-gen battery times"?

Please, come back with actual battery benches across multiple bitrates showing the differences in 64MB of RAM vs 32MB. But don't complain about your willing misinterpretations of what I've said. I said it clearly would affect things, and only said it wouldn't be as significant.

It's also rather frustrating you've also willingly misinterpreted what I said in another thread to fuel your misplaced anger. I clearly did not tell him his old build was the cause, I only told the him that he should update before posting (which is mentioned in the forum guidelines and a requirement, independently of whether it might be the cause).

If his problems occur in the same files every time, as he said they did, given the history of problems our users have had the most likely problem is a drive with errors.

Instead your first suggestion is that he may be using an audio format we don't actually support (Vorbis doesn't support SBR, so either we wouldn't be able to play the file period, or it would be like MP3+ where we ignore the SBR data).

Your third suggestion is even dumber, frankly, suggesting that his boost got locked "low." This can't persist across reboots (as debug values are not saved). Not to mention 24mhz is barely enough to run the WPS, and given his descriptions of the problem (including mentioning the files are 192kbps Vorbis) he almost definitely wouldn't be able to play any file he's tried realtime.

So rather than offering truly useful suggestions to him, what you did is propose two ideas that add confusion into the mix, while mentioning my suggestion is worth checking out (then complaining at me for the same suggestion elsewhere). Of course you did throw in a mention that defragmenting might be useful, once more demonstrating you lack of understanding. See, fragmentation would affect it as it's loaded, but not during playback. Heavy disk fragmentation could make the buffer take longer, sure. But once it was buffered it would be defragmented in RAM - this means that what songs skipping would occur on would depend entirely on playlist order. The user said it's always the same songs, so if his statement is truthful it wouldn't be an issue of fragmentation.

I do appreciate you don't like me, but at the very least, please try not to give stupid advice that will just confuse users.

There are some assumptions you're making about the way some things work that are wholly incorrect, in this thread and what you just said about the other one, but I'm not about to start a forum war to try to prove it. (For example, I've personally seen the fragmentation cause this very problem; also a corruption could indeed lock the CPU at 30 permanently, though not 24.)

Also, like I took your statement further than you meant it, I will clarify that I don't for a minute believe that double RAM = double battery life or anything near it. Different things are the "weakest link" depending on how you run your ipod. If you ran low-bitrate AAC+ at the point it would just barely decode realtime, that link would be the CPU. It would eat the battery long before the RAM had any real effect. Likewise, FLAC is pretty easy on the CPU, but rapes the battery with disk accessing. And even playback volume has a very large effect, like backlight intensity. But for the way I use mine, the RAM is a big chokepoint.

You'll also notice I did back down and acknowledge I was just being a dick and need to sleep... probably while you were typing this already I imagine, as you referenced things I edited out.

FTR, the original 5th gen board was 32MB, regardless of disk size. They changed that later, and did a couple other things, and 5.5 was born. I got that straight from their design team, as I was confused myself.