Memory Poll on Pyra

[Natsu]

 
Mmm... I think we know or can extrapolate more than that. 'Nothing is known yet' is a bit of a misstatement.


A 8GB 95MB/s Toshiba Exceria microSD card retails for under $12 with shipping included.

http://www.amazon.com/Toshiba-micro-Exceria-Memory-Class10/dp/B00BJW6CL4


A 16GB 95MB/s Toshiba Exceria microSD card retails for under $18 with shipping included.

http://www.amazon.com/Toshiba-micro-Exceria-Memory-Class10/dp/B00AIH3CUA


ED should be able to acquire these much cheaper in volume with consolidated shipping (Wholesale).


There is already a SIM card slot planned on the board ED pictured. Combo sockets are available that can house both a microSD card and SIM. They are about $2.50 per unit. http://www.mouser.com/ProductDetail/JAE-Electronics/ST7S014V4BR800/?qs=sGAEpiMZZMuJakaoiLiBpunsJ11fq%2b7VPEylYfUR2n4%3d


This is a price increase of $2 over the cheapest SIM only connectors.


Comparing at 16GB eMMC ($19) vs 16GB microSD (<=$18) + socket difference ($2) the net cost difference for the hardware is <= $1. Likely the microSD option would be LESS expensive as the above retail prices include individual piece US shipping.


Power drain of eMMC Vs microSD is very likely a wash or nearly so. microSD cards are designed for use in portable devices. They likely will consume power at a similar or less rate than full sized SDXC cards of similar specs. I haven't heard any reports of Pandora owners who boot from SD cards stating that their battery gets eaten faster.

Apparently, Sandisk's best beat this in write speeds. 

Is there really any chance in using µSSD?

 

[klapse]

32GB... so it seems you want the Pyra to cost 900 USD in the end?


It only sounds like little things you all want to add, but these add up in the end.

It is only 22 USD more for 32GB...

 

[Natsu]

While I pick 4GB of Ram, 3GB is fine with me. While speed is a concern, I'm sure he'll pick a good speed.

I agree 16GB price vs speed is fair in general.

The 2GB lead isn't really fair since someone is pushing for it which could be messing with the results.

 

[_wb_]

I'd rather have more RAM than faster internal storage. You can use RAM for lots of things, including caching. Linux is good at caching. If things can be kept in RAM, then the speed difference between SD (100MB/s theoretical max), eMMC (200MB/s theoretical max) and SATA SSD (300MB/s theoretical max) is completely moot because RAM can be 8.5GB/s.

 

[Exophase]

While I pick 4GB of Ram, 3GB is fine with me. While speed is a concern, I'm sure he'll pick a good speed.

3GB is probably not possible, at least not with OMAP5432. Not unless someone makes DRAM chips with the correct non-power-of-two size.

I'd rather have more RAM than faster internal storage. You can use RAM for lots of things, including caching. Linux is good at caching. If things can be kept in RAM, then the speed difference between SD (100MB/s theoretical max), eMMC (200MB/s theoretical max) and SATA SSD (300MB/s theoretical max) is completely moot because RAM can be 8.5GB/s.

I just don't see fitting 8 DDR3L chips on the board as being very realistic, this and internal storage are things that needed zero additional space on Pandora yet it was already strapped for space..

 

[Natsu]

While I pick 4GB of Ram, 3GB is fine with me. While speed is a concern, I'm sure he'll pick a good speed.

3GB is probably not an option, at least not with OMAP5432.

Ok, 4GB it is then.

 

[Exophase]

Okay, apparently the OMAP5432 datasheet is way out of date on this, you can actually currently buy 8Gbit DRAMs with 16-bit interfaces, at least from Hynix. So 4GB w/4 DRAM chips is at least feasible.

 

[Grench]

While I pick 4GB of Ram, 3GB is fine with me. While speed is a concern, I'm sure he'll pick a good speed.

3GB is probably not possible, at least not with OMAP5432. Not unless someone makes DRAM chips with the correct non-power-of-two size.

I'd rather have more RAM than faster internal storage. You can use RAM for lots of things, including caching. Linux is good at caching. If things can be kept in RAM, then the speed difference between SD (100MB/s theoretical max), eMMC (200MB/s theoretical max) and SATA SSD (300MB/s theoretical max) is completely moot because RAM can be 8.5GB/s.

I just don't see fitting 8 DDR3L chips on the board as being very realistic, this and internal storage are things that needed zero additional space on Pandora yet it was already strapped for space..

If only the SoC could use LPDDR3...

http://www.engadget.com/2013/07/24/samsung-starts-making-3gb-low-power-memory-for-smartphones/

Probably very expensive anyway.

 

[Grench]

I'd rather have more RAM than faster internal storage. You can use RAM for lots of things, including caching. Linux is good at caching. If things can be kept in RAM, then the speed difference between SD (100MB/s theoretical max), eMMC (200MB/s theoretical max) and SATA SSD (300MB/s theoretical max) is completely moot because RAM can be 8.5GB/s.

Or 17GB/s on the z3770 using LPDDR3, but that seems to be moot too.

 

[TrashyMG]

Some tests when dealing with some large files. Ran each test 3 times to ensure consistency. I'm using a class 10 uSD card.

Read Tests, (Read a large video file to the null device, cleared cache before each read).

Spoiler

Pass1 eMMC:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 14.4559 s, 39.3 MB/s

Pass1 SD:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 28.9444 s, 19.6 MB/s

Pass2 eMMC:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 14.4484 s, 39.3 MB/s

Pass2 SD:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 28.9525 s, 19.6 MB/s

Pass3 eMMC:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 14.4539 s, 39.3 MB/s

Pass3 SD:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 28.9571 s, 19.6 MB/s

Write test, enabled syncing to give a more realistic view.

Spoiler

Pass1 eMMC:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/mnt/eMMC/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 34.1494 s, 11.8 MB/s

Pass1 SD:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/root/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 60.0473 s, 6.7 MB/s


Pass2 eMMC:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/mnt/eMMC/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 33.2429 s, 12.1 MB/s

Pass2 SD:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/root/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 87.9286 s, 4.6 MB/s


Pass3 eMMC:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/mnt/eMMC/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 33.4661 s, 12.0 MB/s

Pass3 SD:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/root/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 61.3281 s, 6.6 MB/s

Copy Tests (copied a large video to the same drive), enabled syncing to give a more realistic view.

Spoiler

Pass1 eMMC:
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/mnt/eMMC/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 59.7541 s, 9.5 MB/s

Pass1 SD:
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/home/trashy/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 111.416 s, 5.1 MB/s


Pass2 eMMC:
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/mnt/eMMC/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 60.0551 s, 9.5 MB/s

Pass2 SD:
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/home/trashy/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 114.262 s, 5.0 MB/s

Pass3 eMMC:
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/mnt/eMMC/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 59.8378 s, 9.5 MB/s

Pass3 SD:
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/home/trashy/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 110.253 s, 5.1 MB/s

 

[Grench]

Some tests when dealing with some large files. Ran each test 3 times to ensure consistency. I'm using a class 10 uSD card.

Read Tests, (Read a large video file to the null device, cleared cache before each read).

Spoiler

Pass1 eMMC:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 14.4559 s, 39.3 MB/s

Pass1 SD:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 28.9444 s, 19.6 MB/s

Pass2 eMMC:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 14.4484 s, 39.3 MB/s

Pass2 SD:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 28.9525 s, 19.6 MB/s

Pass3 eMMC:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 14.4539 s, 39.3 MB/s

Pass3 SD:
root@gta04:/mnt/eMMC# echo 3 > /proc/sys/vm/drop_caches
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/dev/null bs=384k
1443+1 records in
1443+1 records out
567580702 bytes (568 MB) copied, 28.9571 s, 19.6 MB/s

Spoiler

Write test, enabled syncing to give a more realistic view.



Pass1 eMMC:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/mnt/eMMC/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 34.1494 s, 11.8 MB/s

Pass1 SD:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/root/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 60.0473 s, 6.7 MB/s


Pass2 eMMC:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/mnt/eMMC/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 33.2429 s, 12.1 MB/s

Pass2 SD:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/root/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 87.9286 s, 4.6 MB/s


Pass3 eMMC:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/mnt/eMMC/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 33.4661 s, 12.0 MB/s

Pass3 SD:
root@gta04:/mnt/eMMC# dd if=/dev/zero of=/root/test_file conv=fdatasync bs=384k count=1k
1024+0 records in
1024+0 records out
402653184 bytes (403 MB) copied, 61.3281 s, 6.6 MB/s

Copy Tests (copied a large video to the same drive), enabled syncing to give a more realistic view.

Spoiler

Pass1 eMMC:
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/mnt/eMMC/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 59.7541 s, 9.5 MB/s

Pass1 SD:
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/home/trashy/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 111.416 s, 5.1 MB/s


Pass2 eMMC:
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/mnt/eMMC/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 60.0551 s, 9.5 MB/s

Pass2 SD:
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/home/trashy/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 114.262 s, 5.0 MB/s

Pass3 eMMC:
root@gta04:/mnt/eMMC# dd if=/mnt/eMMC/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/mnt/eMMC/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 59.8378 s, 9.5 MB/s

Pass3 SD:
root@gta04:/mnt/eMMC# dd if=/home/trashy/\[HorribleSubs\]\ Sword\ Art\ Online\ -\ 11\ \[1080p\].mkv of=/home/trashy/video_copy.mkv conv=fdatasync
1108556+1 records in
1108556+1 records out
567580702 bytes (568 MB) copied, 110.253 s, 5.1 MB/s

Which eMMC (brand and type)?

Which SD card (brand and type)?

 

[TrashyMG]

eMMC no clue.. need to look up the part number when I get home.
SD card.. It's labeled a class 10. forgot manufacturer, will have to confirm it when I get home.

 

[WizardStan]

eMMC is approximately double that of the SD card. That is actually really disturbing, it makes me think there's something holding it back, like it's clocked the entire thing at just 40Mhz and the only reason the eMMC is faster is because of the double wide bus.

 

[Grench]

eMMC no clue.. need to look up the part number when I get home.

SD card.. It's labeled a class 10. forgot manufacturer, will have to confirm it when I get home.

Those numbers look like you're maxing out the eMMC and SD card - but not the device? I.e. with faster eMMC or faster SD card the differences could be substantial. Class 10 only means that the card should be able to sustain write speeds of 10MB/s.

You also might want to check the settings on the interfaces.

http://www.toshiba.co.jp/p-media/wwsite/microsd.htm#p1

Click on specifications.

To max out the SD side, you'll want a fast UHS-I card capable of using the SDR104 mode.

UHS-I has two speed specs inside it. 50MB/s and 104MB/s.

 

[Grench]

It's too bad the SoC is limited and can't do UHS-II speeds.

http://www.toshiba-memory.com/cms/en/products/sd-cards/exceria/

260MB/s reads and 240MB/s writes, but I'll bet that media is outlandishly expensive right now.

We could theoretically do that through the SATA interface with an on-board adapter. More complications though.

 

[TrashyMG]

Until I can boot off the eMMC or maybe a USB drive, I can't try another SD card..

 

 

[Grench]

eMMC is approximately double that of the SD card. That is actually really disturbing, it makes me think there's something holding it back, like it's clocked the entire thing at just 40Mhz and the only reason the eMMC is faster is because of the double wide bus.

Very good point. Too perfect of a multiplier to be just the memory itself.

 

[Exophase]

If TrashyMG has a probe that can measure frequency he could look at the clock pin going to the SD card slot..

 

[Natsu]

We could theoretically do that through the SATA interface with an on-board adapter. More complications though.

Well, that's a interesting way to do eMMC+µSD.

 

[EvilDragon]

Instead of copying maybe a dd if=/dev/null of=/fileoncardoremmc might be an idea as well.