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

[Grench]

Hrm... a 16GB uSD card... my quick search turned up cards for $50, but for only $18, that's the same range as the eMMC.

It's still half(ish) the speed, but without the huge cost difference that may be an acceptable compromise to be able to upgrade it later. Gonna need to think a bit.

So what was gta04 talking about?

No idea, I don't remember seeing what you're talking about. Got a link?

Not half-ish speed either. More like 95/128 ~ 74%.http://www.tomshardware.com/charts/microsd-cards-2014/-1-Sequential-Read-MB-s,3480.html

The eMMC port IS an SD port. There's no loss going from one to the other, it'd be the same port internally regardless.

And the OMAP5 has 4 SD ports: one eMMC, 2 SD, and one WLSDIO port.

So what was gta04 talking about? I'm sure he was talking about troubles handling 2 SD and the NAND directly with the OMAP...

 I think I remember the post you are referring to, but IIRC  I think he was actually referring to one of the other SOC's being proposed and comparing it to to OMAP.

- Neelix

If I'm recalling the right post - he was talking about the z3770 and how he initially thought it only had an eMMC and one SD. It actually has an eMMC, an SD and an SDIO. Semi-irrelevant for this conversation though.

 

[Reyhn]

Not half-ish speed either. More like 95/128 ~ 74%.http://www.tomshardware.com/charts/microsd-cards-2014/-1-Sequential-Read-MB-s,3480.html

What about random access and latency ? This should be taken into account too.

Plus, how easy would it be to replace an internal µSD ?

 

[God Ginrai]

Plus, how easy would it be to replace an internal µSD ?

Depending on how it's designed, anywhere from extremely easy to still easier than replacing eMMC.

-God Ginrai

 

[vcoleiro1]

If it requires the case to be opened, that is not necessarily easy.  Easy for me, but I imagine some people are not going to be comfortable doing that , which may lead to more warranty returns should the uSD go bad

 

[Grench]

If there is going to be a 3g module in it - then we'll need end users to be able to access the SIM card. A microSD shouldn't need to be any more difficult.

 

[vcoleiro1]

If the concept is for the uSD to be externally accessible like the SIM, then fine. If not, it could be problematic. 

 

[God Ginrai]

If the concept is for the uSD to be externally accessible like the SIM, then fine. If not, it could be problematic.

Yes, but my point is. Worst Case Scenario: It is STILL easier than replacing the eMMC. Best Case Scenario: It's as easy as popping in the card through the side.

-God Ginrai

 

[vcoleiro1]

What are the failure rates  for eMMC and uSD?

 

[God Ginrai]

What are the failure rates  for eMMC and uSD?

Failure rate for microSD is irrelevant. If it fails, you can replace it.

-God Ginrai

 

[vcoleiro1]

IF the failure rate of eMMC is very low, it might not be worth worrying about. Which is why I would like to know the failure rate.

 

[God Ginrai]

IF the failure rate of eMMC is very low, it might not be worth worrying about. Which is why I would like to know the failure rate.

It's still worth worrying about if you care about device longevity.

And even not considering that, microSD gives you the advantage to increase your OS storage size, easily move your OS into a new Pyra, (if your Pyra dies for some other reason, for example) or a way to easily switch your base OS on your device. (have multiple micro SDs for debian, arch, android, etc.) Not to mention, instead of having to spend time flashing the device before shipping, EvilDragon can have a premade image that he copies onto his microSDs that he just inserts into the device as soon as he gets it and ships it out.

-God Ginrai

 

[vcoleiro1]

That depends on the failure rate, if for example 1 in 10,000 eMMC failed over 20 years, it definitely would not be worth worrying about.

 

[Grench]

And for a completely different idea... If we want real speed...

The OMAP has a SATA interface. SD cards are already available that are 2.5 times faster than the SD bus on the OMAP.

Using the on-board equivalent of:

http://www.amazon.com/Micro-SATA-Cables-TM-Converter/dp/B0078PVL8U

Ship with a 'good' -fast- SD card.

http://www.amazon.com/SanDisk-Extreme-Class-Memory-SDSDXPA-008G-AFFP/dp/B007NDL542

If users choose to, then they can slap one of these in it:

http://www.engadget.com/2014/02/12/sandisk-u3-sdxc-card/

Then have two exposed, one internal SD instead of a microSD slot - with the internal slot running off of the SATA connections on the SoC.

 

[God Ginrai]

That depends on the failure rate, if for example 1 in 10,000 eMMC failed over 20 years, it definitely would not be worth worrying about.

It is impossible to gather data over 20 years for current eMMC. Your point makes no sense.

-God Ginrai

 

[comradekingu]

Its potentially every single one, or a large amount, which is worrisome. And they all fail, invariably at some point, at which i dont want it to be soldered to the board. Sure would like to put a 64GB eMMC in my pandora now that they are cheap.

 

[vcoleiro1]

@God Ginrai - Seriously!,  it was a hypothetical and wasn't calling for 20 years of data.  All I have said is that I'd be keen to see the e-MMC failure rate. IF its very low, than the point re: being able to swap a uSD and not an e-MMC in failure cirumstances is pretty much moot. That's not discounting other benefits of a swappable uSD .

Also, not that it's important, but you do realise failure stats can be provided for long periods of time.  They write/read from the component a certain amount of times that equates to a number of years usage.  Ie the test don't actually run for years .

 

[Klumpen]

Stay friends, guys.

The atmosphere here is a bit toxic atm.

 

[comradekingu]

Over time the failure rate is 100%, whatever your span may be. And in any event, you have the benefit of price coming down, available size going up. Also its very servicable. Moreso if you dont have to unscrew the whole case to replace it.

Manufacture burst rate speed ratings on sd cards are next to useless. you need Random IOPS, which is better on eMMC and nand

Since there is no space for mpcie, and eMMC makes the most sense for OS, at which point you can wish for extra speed, which in turn makes that a size argument, bigger sizes being faster, then you run into that being different models and hard to maintain, which is solved by going removable eMMC, which has no drawbacks beside being a little bigger than soldered nand.

 

[vcoleiro1]

Over time , the failure rate for all your components in the Pyra will be 100%.  As much as you would like it to , it won't live forever. That's why some reasonable assessments need to be made.  

 

[WizardStan]

Not half-ish speed either. More like 95/128 ~ 74%

That's optimal, not necessarily average, and not even a guaranteed maximum if the Pyra just can't handle it: the bus itself has a theoretical maximum of 95MB/s for SD cards and theoretical maximums are often hard to hit let alone sustain, so who knows what practical is going to be. The 128MB/s is out of a theoretical max of 200MB/s, and on the Odroid: it's impossible to know just by looking at the numbers whether that's because of the chip itself or the Odroid or something else entirely.If you're going to treat the 95MB as a given then you'll need to treat the eMMC's maximum rating of 192MB/s (the theoretical maximum of the OMAP5) as a given as well, at least until we can benchmark both on the same system, that's why I said half.

To put another way, the 128MB/s is worst case, the 95MB/s is best case, with the truth being somewhere in between.