[poll; updated] eMMC vs. uSD; modular eMMC

[TrashyMG]

Well if anything I got both Grench and God Ginrai to agree on something.

 

[God Ginrai]

Well if anything I got both Grench and God Ginrai to agree on something.

We actually agree on things about half the time, and disagree the other half.

-God Ginrai

 

[SNESFAN]

The GCW zero has an internal microsd instead of a nand and I've taken it out and replaced it a total of once in favor of a 64GB card. There is no point in constantly removing it because you have to open the case, and without it the device won't boot. The speed is fine because the zero boots in like 3 seconds, it runs a minimal Linux rather than a full desktop but I doubt it makes that much of a difference. I notice literally zero loading time when loading apps and Roms from it. The concern over the access and read/write speed over emmc is probably a situation of over analysis of numbers on paper. The addition of removable internal on the zero is a strong selling point in my opinion, most people probably don't care it can do it. Those that do care would be upset its not there. The difference in real life performance I promise would be negligible at best.

 

[_wb_]

Probably almost no one would go through the trouble of upgrading the internal microSD card to get more storage or speed when they could upgrade one of their external full SD cards instead. Sure, you could max out all of them, but the max storage on the microSD card will be at best half of what the full SD can give so you'd be missing out on being able to increase your total storage by under 25%. That's pretty whatever.

The point of the external SD slots is that you are free to eject the cards in them at any time. The point of internal storage is that it contains all the software and data that you want to have available all the time. External storage is not a real replacement for internal storage, unless you sacrifice one of the external slots to have a non-ejectable card. Having two free-to-eject external SD slots is a really nice feature, it allows you to copy things from one SD card to another SD card directly. Non-upgradeable internal storage effectively forces people to give up one ejectable external SD slot at some point in time.

 

[Natsu]

Generally it's placed in places like the battery compartment that doesn't need a screwdriver.

I really hope that's the case with the Pyra. I really don't want screws on the battery cover.

 

[comradekingu]

Im contemplating internal uSD a bit, i dont know if the hardkernel option takes a performance hit with the adaptor, but in any event that is possible, and probably going to come down to size /possible to build.

Meanwhile 32GB toshiba uSD is a workable solution, it is non proprietary, and uSD will stay around for ages. Also its extendable and means no comprimises have to be made on the software front.

To offset the inherrent slowness of uSD how about going for more memory? The idea of more memory upfront makes more sense since its soldered in. While more memory means a tiny bit more power use, it is negligible, also i think uSD uses less than emmc would.

If 8Gb density chips are used, with four of them thats 4GB of memory. At that point one can load the whole OS into memory and still have over half left over available to use.

8GB is possible on the TI side, but i couldnt find any chips for it.

http://www.skhynix.com/products/computing/computing.jsp?info.ramCategory=computing&info.ramKind=19&info.eol=NOT&posMap=computingDDR3

http://www.samsung.com/global/business/semiconductor/product/computing-dram/catalogue

http://www.micron.com/products/dram/ddr3-sdram#fullPart&236=3

http://www.elpida.com/en/products/ddr3.html

Edit:

http://www.etron.com/en/products/specialtydram_detial.php?Product_ID=3

http://www.esmt.com.tw/english/main_products.asp 

http://www.issi.com/US/product-dram-ddr3.shtml

 

[destraht]

I don't understand how you can eliminate modular eMMC in your logic.  Somehow only being  able to upgrade from one source has issues so then there must be no upgrade?!  How does that work.

That one source stops making it, you're boned.

-God Ginrai

Also, I don't see how ED would be able to sell everything between 4GB and 128GB modular eMMC, at various speed/price points for each size, for a good price. We just don't have the scale for that, so in practice, the number of options would be rather small.

I think that having it at a good price is not going to be that important in a few years.  If you think of it as expensive eMMC then it sucks but if you think of it as ungrading your system then then price will not be all that expensive.

He could offer 4GB and 64GB to start.

 

[destraht]

And your response to how that is the slowest option available?

Is it slow with IOPS because of the protocol or is something to do with the way the current chips are made?  It is possible that future chips could fix the [big] IOPS issue if the market demanded that.  Unfortunately that is a gamble.

 

[Askarus]

Probably almost no one would go through the trouble of upgrading the internal microSD card to get more storage or speed when they could upgrade one of their external full SD cards instead. Sure, you could max out all of them, but the max storage on the microSD card will be at best half of what the full SD can give so you'd be missing out on being able to increase your total storage by under 25%. That's pretty whatever.


In my mind, the ability to more cheaply replace a broken microSD card is the only compelling advantage for the user after the device has been purchased (so ignoring any advantage in manufacturing regarding cost or ease of programming). But it's only an advantage if the eMMC actually dies before something else does and the unit is no longer under warranty, and you haven't yet replaced the device with something newer and better. Given how many other failure points there are on the device it doesn't seem worth that much to mitigate one of them unless it's known to be less reliable, which it isn't.

I will go through the trouble of upgrading the internal micro SD as soon as I can upgrade for a reasonable price.

Don't see any reason why I wouldn't.

Assuming we'll use a combo port for SIM and micro SD then it's just a lid to open.

Everyone can do that.

Also I wouldn't want to sacrifice storage on one of the big SD-cards for the OS.

I want to be able to take both out and still boot the Pyra.

We also gain the advantage that we can make a copy of our OS Micro SD card and pot it in an other Pyra.

Need a backup?

Just mirror the Micro SD and swap it if needed.

 

[Ligachamp]

What if we work with a removable adaptor? We make a slot for removable eMMC and offer a eMMC to µSD adaptor, which you plug into the eMMC slot. So there would be the possibility to use removable eMMC or µSD just how you prefer! I thougt about something like the adaptor on this page http://www.kingston.com/us/flash/emmc

I think soldered eMMC wouldn't be as future proof as changable storage devices.

 

[_wb_]

IOPS of the microSD and eMMC itself should be roughly similar, since the underlying technology is the same. On decent SD cards, for random small reads ~1000 IOPS is typical, for random small writes ~100 IOPS is typical (writing is always slower because of the underlying technology, i.e. block-based NAND flash). For comparison, a spinning HDD typically does ~100 IOPS (reading or writing makes no difference), while a good SSD can usually do 10,000-100,000 IOPS.

IOPS is not an issue, and the difference between eMMC and µSD is not significant in that respect.

 

[Granitehead]

So, after looking at this thread and thinking about my questions again, I think the right one to ask would have been:

Regarding the speed of internal primary storage...

-a maximum speed faster (in all aspects) than your average hard drive (and Pandora NAND) is enough

-I need even more speed

Regarding it's size...

-I'm okay with 16GB (32GB at most maybe, according to ED) FOREVER

-I might want to (cheaply) upgrade it at some point in the future, to any (more or less) size I want. Or maybe I just want to be sure to still have working internal storage in 20 or 30 years

With those answers, ED could decide. And I think these answers would be less biased than the ones for the current questions.

Any feedback? I might reset the poll with those new question if there is none/positive.

I don't think the "additional point of breakage" argument against uSD is valid. First, people wouldn't often change it (there are two other SDs, you can easily connect it over network, there is another /bootable/ SD, etc). Second, when it breaks: It's certainly easier to resolder an uSD cage than a eMMC chip (the uSD cage wouldn't be that heat sensitive, it's pins would probably be easier to access than {I think? maybe forget that one}, if you screw up resoldering it once, who cares?, you just buy a second one for $5 and try again, if there are any returns because of it it costs ED way less to replace, etc.). Third, it could even save PCB space when combining it with the SIM slot.

 

[zonova]

How prone is eMMC to breaking compared to uSD, maybe from things like static build up or whatever else.

 

[Granitehead]

How prone is eMMC to breaking compared to uSD, maybe from things like static build up or whatever else.

eMMC is pretty unlikely to break from anything but old age/wear. But it is /sure/ to break after you use it long enough.

The uSD cage should only break if you aren't careful enough when inserting/ejecting or after 1000/5000/10000 insertion cycles (that's a value which you know when you buy them). That is, it breaks not depending on how much you use the memory, but depending on how often you change uSD. That moves the expected lifetime up a whole order of scale.

 

[vcoleiro1]

Don't forget the "I understand I can only get half the speed of eMMC with uSD - and I'm happy to live with that forever"  option

 

[SNESFAN]

I still can't fathom a situation where anyone would be removing the internal microsd so many times to the point where it breaks. I also can't see nand being written to so much that it fails before the device is obsolete. I'm sure its possible, but I bet android devices write to nand a hell of a lot more than the Pandora does and aside from the cheap Chinese SoC's I have yet to have any actual issues with it. So I think the lifespan of either isn't an issue. The speed I don't think is a real issue, more of a benchmark issue that I doubt anyone is going to notice. In my opinion what it comes down to is price and space. If nobody can see using more than the 8,16,or 32gb (whatever is decided on) for their internal storage then no problem. But in 2009 when the Pandora came out 16GB microsd cards were the biggest you could get and 32gb was right around the corner then. Now we have 64gb with 128gb right around the corner. Even if we maintain only the 2 jumps in 5 years we are talking about the possibility of having 256gb of internal with 512gb around the corner. Deciding on nand (if not removeable) will lock you in time and make you look old as crap in 5 years. 512mb is laughable by today's standards. 8gb (if chosen) will be laughable by future standards. Removable emmc doesn't have a universal standard and will put you in a situation of expensive, proprietary, and hard to find, that is if something better doesn't come out and make it actually impossible to find/uprade to. Look 5 years down the road (my estimate on this model) what will future you say about the decision now?

 

[comradekingu]

vcoleiro: Except that is a premise you cant make, because we dont know that

SNESFAN: Nand fails, SLC being a little better than MLC, moreso if your nand is of low quality, because it mathematically will, in any event eventually, such is also the nature of forever shrinking the process at which it is made, making it more prone to failure, which a lot of nand is, _especially_ if you are a small company sourcing it yourself. At which point we are all boned.

If you are a user in the time doman, and value your time, then you notice the difference.

A slower device can get away with slower storage,

A device with more ram can also get away with slower storage. Like modular eMMC makes sense in a holistic way, having more ram if the device is going to use uSD for OS files makes sense.

At the time of making the pandora, microSD was not an option, for much the same reason sourcing a good uSD is even hard now, maybe its a barable option now, and certainly the prospect of how much improvement there has been since the pandora days, makes the foreseeable upgradability an alluring prospect.

Laws do apply though, and the forever shrinking to forever upgrading controllers wont change. Quality always costs, but at a equally falling rate.

If a modular emmc isnt available then im still partial towards uSD, because its less risk to have a more modular approach to the things prone to failure by wear/use (battery, nand). Also that connector is very standard and a custom future solution is all the more viable if its only custom at one end.

TL;DR speed is a variation over the theme, how advanced do you want your use to be/what can you defend relating to cost at a diminishing return

More important to me is less risk, and the more use over time aspect of modularity.

EDIT: there are 4 buses that can be used for SD cards from what i understand. How about 2 uSD slots for OS storage?

 

[vcoleiro1]

@comradekingu :-  I cant make the premise that the OMAP5 bus limit is 200MB/s for eMMC and 100MB/s for uSD as Exophase posted previously?.   Those are hard limits regardless of the improvements of uSD.  ->  http://boards.openpandora.org/topic/15651-poll-updated-emmc-vs-usd-modular-emmc/?p=314516

Look at it like this , which is the lesser of the 2 evils.   

1) Have 2 card slots and eMMC :-   gives everyone 2 x speed of uSD forever , has 2 card slots which give everyone else the option and flexibility to do what they want, including using bigger cards in the future

2) Have 3 card slots (1 uSD) :-  gives everyone 1/2 speed of eMMC forever, has 3 card slots for everyone to use as they want,  including using bigger cards in the future

 IMO 16GB eMMC will be enough for most people - you could fit several OS on that , perhaps we can even get 32GB.  With option 2, you are saying that the minority of users who would upgrade the main storage in the future should impose slower speeds on the majority who wouldn't.  Indeed , slower speeds on everyone.  With option 1, you still have 2 slots to use as you want and the ability to get larger cards , so you still have that benefit. 

Which is the fairer option for everyone.?  IMO it's number 1.

 

[comradekingu]

Problem is people who think of software as cartridges on a bazaar linux|gnu system, in 2014. I suppose those same people would probably want removable uSD so they can have different OS-es on different cards though, so how can we rub everyone the right way? Just voting on different options with a subset of options, each having implications not understood fully by any of the voters, is a crapshoot. Lets try to reason it down to something that results in something that makes sense overall.

The big problem is shipping a sane default, one that doesnt cost through the nose. Not in purchase price, nor in potential damage repair costs. While those toshiba cards are nice for the one metric people seem to care about, they cost 60+ dollar, which is just too damn high. The samsung plus, crushed it on the charts, and it costs 23... its also got a lot better results than the pro variant...

I got a samsung 32GB MB-MSBGA class 10 32GB microsdhc card, bought it June 2012, it was just lauched then i think, and it is wonderful. Some review sites as always have negative reviews, but those are the people to leave reviews, and you factor that into sending out extra cards, users testing themselves instead of going the full refund the pyra route (even in vain) The limited 5 year guarantee means samsung is too playing that refund/reward game, the right way by the looks of it.

Samsung plus, doing then consistently better than even the new sandisk emmc where it matters...

 http://www.anandtech.com/show/7347/sandisk-announces-optimized-inand-extreme-emmc-for-bay-trail

also see http://www.kingston.com/us/flash/emmc

It seems eMMC doesnt have the performance highground it once did, but something about it all seems too good to be true. A real world test on real metal is what needs to happen. Testing with uSD and if thats as good as the http://www.tomshardware.com/charts/microsd-cards-2014/benchmarks,169.html bench, then so be it, thats the only thing i do trust. And then some wear levling estimates based on torture tests.

I think we need to look at the bigger picture, i think option a being no uSD and option b being 32gb samsung plus is the two better options if both of them have more ram to boot.

So much cool stuff can be done with more memory, and at 4GB the price well defends the abilities and speed, it also, which is the kicker here, takes the performance downfall from SD cards away. _And_ its better for wear and tear.

Left is then the nice parts about SD, which almost perfectly solves the decision on what hard limit soldered option one should go for, once you set a lower hard limit, you are stuck supporting that, which is how the option thats cheap ruins the day for those who can pay more.

no hard limits is in turn is how software issues get solved. No need to carry on something that is a compromise (pnd) when the reason for its existence is limits that are no longer there.

That last part im 100% on, because i have taught many novice and computer illiterate people to use linux|gnu systems over the years, and explaining package management twice just does not fly. Give them something that works, in an uniform way, for their needs, and stick to it.

Edit: keeping parts the same between neo900 and pyra would be great. That dual sim/uSD slot tested in the GTA04 seems like a good option, but preferably somewhere it can be reached without screws. Sim is going to be reachable, so why not.

 

[_wb_]

Look at it like this , which is the lesser of the 2 evils.   

1) Have 2 card slots and eMMC :-   gives everyone 2 x speed of uSD forever , has 2 card slots which give everyone else the option and flexibility to do what they want, including using bigger cards in the future

2) Have 3 card slots (1 uSD) :-  gives everyone 1/2 speed of eMMC forever, has 3 card slots for everyone to use as they want,  including using bigger cards in the future

Which is the fairer option for everyone.?  IMO it's number 1.

Only if we actually get an eMMC _now_ that maxes out the bus speed (and not just on easy sequantial reads, but on also on random small writes), is big enough to not be laughable in a few years, never wears out, and is cheap enough to not drive up the total device cost significantly. I don't know if those requirements can be met even when ignoring price.

If we end up with a sensible size/speed/price compromise eMMC, then it may well be slower and smaller than the best microSD.

If you want speed, just use a ram disk. With proper caching (no need to force immediate writebacks on an internal microSD card!), that is what will essentially happen anyway, most of the time. Frequently accessed files will usually never really leave RAM, so speed is not that much of an issue.