Bons
Posts: 3
Joined: Sat Jan 14, 2012 12:43 pm

Re: Class 10 SD cards on the production boards

Sat Jan 14, 2012 12:51 pm

Hi There,

I know there been many discussions in the past on the Class 10 SD card on the forum.

But does any know if Class 10 SD card will work properly with the production boards?

The wiki says no, but they think this is a Software problem of linux ?

I want to buy this card if Class 10 SD cards Work properly:

http://www.icidu.com/index.php.....a-8gb.html

I have used the forum search, but not able to find an autual discussion about this matter.

I hope to get the answer from the community or the Admins i asume that they have tried many SD card on their own production

alpaca
Posts: 29
Joined: Wed Jan 11, 2012 1:28 pm

Re: Class 10 SD cards on the production boards

Sat Jan 14, 2012 1:37 pm

cdnr1 said:


shut up my bonghole wil ask the question



and what is that supposed to mean?

Also: As far as I recon, it we don't really know. But I guess we will find out when some more people have a board and time to test it (i.e. when the sales begin)

seekerm
Posts: 11
Joined: Tue Dec 27, 2011 12:43 am

Re: Class 10 SD cards on the production boards

Sat Jan 14, 2012 2:46 pm

Bons said:



I want to buy this card if Class 10 SD cards Work properly:

http://www.icidu.com/index.php.....a-8gb.html


You dont need to buy class 10 card in order to achive highspeed R/W.

Some class 4 / 6 cards as fast as class 10 cards.

http://www.amazon.co.uk/Sandis.....038;sr=1-1 << Class 4 card which suppose to work on raspi other models of the same kind do work(ordered that one)..

http://www.amazon.co.uk/Samsun.....roduct_top << clas 6 card, not sure if it works but it suppose too.

Tomo2k
Posts: 127
Joined: Mon Dec 19, 2011 10:00 pm

Re: Class 10 SD cards on the production boards

Sat Jan 14, 2012 3:26 pm

There are a lot of threads about this.

The TL;DR summary:

Class 10 are likely to be slower than Class 4 or 6 when used in this context.

They aren't designed for random access.

Bons
Posts: 3
Joined: Sat Jan 14, 2012 12:43 pm

Re: Class 10 SD cards on the production boards

Sat Jan 14, 2012 3:28 pm

I know that there is no speed difference, but here in the netherlands are the Class 6 SD cards more expensive than class 10.

That why i am asking this question

dukla2000
Posts: 190
Joined: Tue Jan 10, 2012 12:02 am
Location: Reading.UK.EU

Re: Class 10 SD cards on the production boards

Sat Jan 14, 2012 4:32 pm

I don't understand the"problem in Linux" part of Class 10 cards. My no-name brand 16Gb Class 10 from fleaBay has run just fine in my Android tablet (currently kernel 2.6.36.3) for over a year.

In a USB adapter on my desktop (Linux 2.6.34.10-0.4-desktop x86_64) haven't had a problem either:


./f3write /media/disk/
Free space: 14.46 GB
Creating file 0001.fff ... OK!
Creating file 0002.fff ... OK!
Creating file 0003.fff ... OK!
Creating file 0004.fff ... OK!
Creating file 0005.fff ... OK!
Creating file 0006.fff ... OK!
Creating file 0007.fff ... OK!
Creating file 0008.fff ... OK!
Creating file 0009.fff ... OK!
Creating file 0010.fff ... OK!
Creating file 0011.fff ... OK!
Creating file 0012.fff ... OK!
Creating file 0013.fff ... OK!
Creating file 0014.fff ... OK!
Creating file 0015.fff ... OK!
Free space: 0.00 Byte
Average writing speed: 12.33 MB/s

./f3read /media/disk/
SECTORS      ok/corrupted/changed/overwritten
Validating file 0001.fff ... 2097152/        0/      0/      0
Validating file 0002.fff ... 2097152/        0/      0/      0
Validating file 0003.fff ... 2097152/        0/      0/      0
Validating file 0004.fff ... 2097152/        0/      0/      0
Validating file 0005.fff ... 2097152/        0/      0/      0
Validating file 0006.fff ... 2097152/        0/      0/      0
Validating file 0007.fff ... 2097152/        0/      0/      0
Validating file 0008.fff ... 2097152/        0/      0/      0
Validating file 0009.fff ... 2097152/        0/      0/      0
Validating file 0010.fff ... 2097152/        0/      0/      0
Validating file 0011.fff ... 2097152/        0/      0/      0
Validating file 0012.fff ... 2097152/        0/      0/      0
Validating file 0013.fff ... 2097152/        0/      0/      0
Validating file 0014.fff ... 2097152/        0/      0/      0
Validating file 0015.fff ...  962560/        0/      0/      0

Data OK: 14.46 GB (30322688 sectors)
Data LOST: 0.00 Byte (0 sectors)
Corrupted: 0.00 Byte (0 sectors)
Slightly changed: 0.00 Byte (0 sectors)
Overwritten: 0.00 Byte (0 sectors)
Average reading speed: 15.72 MB/s



The write speed is significantly better than some 2Gb cards I have played with (obviously no HC, gave 6 or 7 MB/s) or my 4Gb HC Class 4 (4.36MB/s).
Daily driver: Pi3B, 64GB Samsung Evo+ @100MHz, DVB-T, onboard WiFi for internet, BT/USB dongle for KB/mouse, 250GB HDD via USB for media, Raspbian Jessie Lite with Openbox desktop.
Museum: Pi B

Tomo2k
Posts: 127
Joined: Mon Dec 19, 2011 10:00 pm

Re: Class 10 SD cards on the production boards

Sat Jan 14, 2012 4:56 pm

dukla2000 said:

The write speed is significantly better than some 2Gb cards I have played with (obviously no HC, gave 6 or 7 MB/s) or my 4Gb HC Class 4 (4.36MB/s).
Write speed is not the same as latency.

Class 10 will write a 2GB file faster than Class 6, but Class 6 will usually write 2000 1MB files faster than a Class 10.

- This is to be expected, as in a camera (stills or video) you want to write a small number of big files. In a computer, it's a large number of small files.

User avatar
Chromatix
Posts: 430
Joined: Mon Jan 02, 2012 7:00 pm
Location: Helsinki

Re: Class 10 SD cards on the production boards

Sat Jan 14, 2012 5:00 pm

I've got a handful of SanDisk SD cards due to arrive in the next few days, and I can test those in my TrimSlice which has a directly-attached SD card slot just like the R-Pi.  If there is a general Linux problem with these cards, it should show up there.

The SanDisk cards are, however, somewhat above average in reputation.  They are Class 10 technically, but the manufacturer takes genuine pride in their performance and doesn't simply meet the minimum spec.  I therefore expect to get good results from them one way or another.
The key to knowledge is not to rely on people to teach you it.

dukla2000
Posts: 190
Joined: Tue Jan 10, 2012 12:02 am
Location: Reading.UK.EU

Re: Class 10 SD cards on the production boards

Sun Jan 15, 2012 3:27 pm

From JamesH figures on an alpha board http://www.raspberrypi.org/for.....th/#p30617 it seems the ability of the cards is likely pretty academic, the R-Pi interface to the SDcard (or his filesystem or OS or ..) will be the main limitation.
Daily driver: Pi3B, 64GB Samsung Evo+ @100MHz, DVB-T, onboard WiFi for internet, BT/USB dongle for KB/mouse, 250GB HDD via USB for media, Raspbian Jessie Lite with Openbox desktop.
Museum: Pi B

hvc123
Posts: 81
Joined: Fri Nov 25, 2011 11:54 am

Re: Class 10 SD cards on the production boards

Sun Jan 15, 2012 4:12 pm

as we on the subject of SDCARDS. i have stayed away from flash mem for as long as i possibly can with the exception of wiliboard and openwrt butdoes it still have the same issues with a limited number of writes? so would jffs/isofs file structure be the way forward with no swap

seekerm
Posts: 11
Joined: Tue Dec 27, 2011 12:43 am

Re: Class 10 SD cards on the production boards

Fri Jan 20, 2012 2:08 pm

Small test on the above mentioned card "Sandisk Ultra 15MB/s SDHC Card 8GB":

./f3write /media/disk/
Free space: 7.39 GB
Creating file 0001.fff ... OK!
Creating file 0002.fff ... OK!
Creating file 0003.fff ... OK!
Creating file 0004.fff ... OK!
Creating file 0005.fff ... OK!
Creating file 0006.fff ... OK!
Creating file 0007.fff ... OK!
Creating file 0008.fff ... OK!
Free space: 0.00 Byte
Writing speed: 9.82 MB/s

./f3read /media/disk/
SECTORS ok/corrupted/changed/overwritten
Validating file 0001.fff ... 2097152/0/0/0
Validating file 0002.fff ... 2097152/0/0/0
Validating file 0003.fff ... 2097152/0/0/0
Validating file 0004.fff ... 2097152/0/0/0
Validating file 0005.fff ... 2097152/0/0/0
Validating file 0006.fff ... 2097152/0/0/0
Validating file 0007.fff ... 2097152/0/0/0
Validating file 0008.fff ... 827328/0/0/0

Data OK: 7.39 GB (15507392 sectors)
Data LOST: 0.00 Byte (0 sectors)
Corrupted: 0.00 Byte (0 sectors)
Slightly changed: 0.00 Byte (0 sectors)
Overwritten: 0.00 Byte (0 sectors)
Reading speed: 18.88 MB/s

User avatar
psergiu
Posts: 223
Joined: Mon Nov 07, 2011 8:36 am
Location: TX, U.S.A. (was: RO, E.U.)
Contact: Website

Re: Class 10 SD cards on the production boards

Fri Jan 20, 2012 5:47 pm

Read the Wikipedia entry in Secure Digital. http://en.wikipedia.org/wiki/Secure_Digital

Class 10 cards are using "4-bit SD bus", the slower ones are using SPI & 1-bit.

If the Broadcom bootloader cannot talk "4-bit" it won't be able to boot from those cards as some newer/high capacity cards no longer support the simpler protocols.

seekerm
Posts: 11
Joined: Tue Dec 27, 2011 12:43 am

Re: Class 10 SD cards on the production boards

Fri Jan 20, 2012 8:37 pm

psergiu said:


Read the Wikipedia entry in Secure Digital. http://en.wikipedia.org/wiki/Secure_Digital

Class 10 cards are using "4-bit SD bus", the slower ones are using SPI & 1-bit.

If the Broadcom bootloader cannot talk "4-bit" it won't be able to boot from those cards as some newer/high capacity cards no longer support the simpler protocols.



Thats why some fine Class 4 & 6 cards are suggested here

User avatar
Chromatix
Posts: 430
Joined: Mon Jan 02, 2012 7:00 pm
Location: Helsinki

Re: Class 10 SD cards on the production boards

Fri Jan 20, 2012 11:22 pm

psergiu said:


Class 10 cards are using "4-bit SD bus", the slower ones are using SPI & 1-bit.



That's interesting.  Is there an easy way to tell whether a given, inserted card is using 1-bit or 4-bit mode in Linux?  I could use that to investigate why a SanDisk Ultra II isn't reaching it's rated speeds in my TrimSlice.
The key to knowledge is not to rely on people to teach you it.

User avatar
Chromatix
Posts: 430
Joined: Mon Jan 02, 2012 7:00 pm
Location: Helsinki

Re: Class 10 SD cards on the production boards

Fri Jan 20, 2012 11:54 pm

psergiu said:

If the Broadcom bootloader cannot talk "4-bit" it won't be able to boot from those cards as some newer/high capacity cards no longer support the simpler protocols.

Actually, reading the Wikipedia page, you seem to be incorrect here.  All SD/MMC cards (including SDHC and SDXC) boot up in 1-bit or SPI mode, and the host has to manually switch them to 4-bit mode if supported.  The UHS performance standard (*not* Class 10) mandates support for the 4-bit mode, but these cards *still* boot in 1-bit mode for compatibility.

Electrically, supporting 4-bit mode seems to be as simple as connecting three lines which are not used in 1-bit mode.  It should be fairly simple to determine from the cirvuit board design whether these lines are present.
The key to knowledge is not to rely on people to teach you it.

User avatar
Chromatix
Posts: 430
Joined: Mon Jan 02, 2012 7:00 pm
Location: Helsinki

Re: Class 10 SD cards on the production boards

Sat Jan 21, 2012 9:12 am

Having tweaked the kernel on my TrimSlice to report the host and card capabilities explicitly, it looks like the 1-bit vs. 4-bit distinction is *really* not the problem here.  Every single SD card currently in my collection supports 4-bit mode, including a tiny 256MB and a really bog-standard 2GB card.

The only distinction I have found is that the above-mentioned cards operate at 25MHz while the newer and larger cards support 50MHz (although the TrimSlice itself only supports 48MHz).  None of these cards get anywhere near the speeds theoretically available from their interface, even when reading.

The Foundation chaps report that SanDisk Ultra series cards work fine.  I unearthed a SanDisk Ultra from one of my cameras, which is among the 50MHz, 4-bit interfaced examples.

However my personal collection includes cards only up to Class 6 so far - the Class 10 cards I have previously tested belonged to the office.  This should hopefully be corrected when the cards I have ordered finally arrive, as they include a SanDisk Extreme series card which claims a UHS-1 performance rating (which has Class 10 as a prerequisite).

It seems plausible that problematic cards in this category might use DDR signalling when in high-speed mode.  If it is a generic Linux problem, it should be revealed by my TrimSlice with the SanDisk Extreme card.  If it is a host-hardware problem however, it should be possible to "quirk" it out in the kernel driver so that the DDR mode is not used on the R-Pi.
The key to knowledge is not to rely on people to teach you it.

foo
Posts: 52
Joined: Thu Dec 29, 2011 12:49 am

Re: Class 10 SD cards on the production boards

Sun Jan 22, 2012 1:23 am

hvc123 said:


as we on the subject of SDCARDS. i have stayed away from flash mem for as long as i possibly can with the exception of wiliboard and openwrt butdoes it still have the same issues with a limited number of writes? so would jffs/isofs file structure be the way forward with no swap


Today's tech as an increased number of writes and automatic wear-leveling, effectively making the limit a non-issue.

mightygoose
Posts: 85
Joined: Mon Jan 02, 2012 1:53 pm
Contact: Website

Re: Class 10 SD cards on the production boards

Sun Jan 22, 2012 12:19 pm

FWIW i have a 8gb samsung class10 micro SD waiting to test as soon as my pi arrives as well as a 8gb class 4 as a backup.

User avatar
Chromatix
Posts: 430
Joined: Mon Jan 02, 2012 7:00 pm
Location: Helsinki

Re: Class 10 SD cards on the production boards

Tue Jan 24, 2012 11:01 pm

I've just got hold of Class 10 4GB cards from both SanDisk (Extreme series) and Transcend (black label).  I've also used some judicious kernel patching and benchmarking to give me more technical information about them.

The short version is: I can't find any difference whatsoever which would prevent them working in any slot in which a comparable Class 4 or Class 6 card would work.  The interface spec is identical, the performance is on-par or better even when I make them perform acrobatics, and they work just fine with the directly-attached slot in my TrimSlice, as well as with a USB adapter on my old netbook.

I would therefore be rather interested to hear more about any Class 10 cards which supposedly *don't* work in the R-Pi.  There has got to be some difference between those and the ones I can actually get hold of around here.
The key to knowledge is not to rely on people to teach you it.

User avatar
Chromatix
Posts: 430
Joined: Mon Jan 02, 2012 7:00 pm
Location: Helsinki

Re: Class 10 SD cards on the production boards

Thu Jan 26, 2012 7:52 pm

After exploring the nether regions of my friendly local computer megahyperstore, I now have an extra selection of SD (and MicroSD) cards from two additional vendors (Kingston and Verbatim) to play with.  Most of them are Class 10s.

A quick plug into the tweaked TrimSlice reveals no interface differences compared to other current-generation cards.

Benchmarks and torture tests are yet to be run – I expect fully craptacular results from the Kingstons, but the Verbatims are an unknown quantity.

On a side note, Transcend wins hands down on ease of opening the packaging.  Verbatim loses bigtime here, though SanDisk aren't that much better.  Kingston's packets are tolerable if you know the trick to them.
The key to knowledge is not to rely on people to teach you it.

dukla2000
Posts: 190
Joined: Tue Jan 10, 2012 12:02 am
Location: Reading.UK.EU

Re: Class 10 SD cards on the production boards

Thu Jan 26, 2012 9:22 pm

Chromatix said:

On a side note, Transcend wins hands down on ease of opening the packaging.  Verbatim loses bigtime here, though SanDisk aren't that much better.  Kingston's packets are tolerable if you know the trick to them.
Suspect this is going to be the most significant benchmark of cards for the Pi! As per JamesH numbers the card performance itself is not an issue in a Pi. (or at least alpha-Pi).
Daily driver: Pi3B, 64GB Samsung Evo+ @100MHz, DVB-T, onboard WiFi for internet, BT/USB dongle for KB/mouse, 250GB HDD via USB for media, Raspbian Jessie Lite with Openbox desktop.
Museum: Pi B

Themroc
Posts: 44
Joined: Fri Jan 06, 2012 9:35 am

Re: Class 10 SD cards on the production boards

Thu Jan 26, 2012 11:06 pm

This seems to evolve into a SD card benchmark thread... So, these are the cards we (will) have:

Bons said:

I want to buy this card if Class 10 SD cards Work properly:
http://www.icidu.com/index.php.....a-8gb.html


That would be a "ICIDU 8GB, Class10, 20MB/s"

seekerm said:


Small test on the above mentioned card "Sandisk Ultra 15MB/s SDHC Card 8GB":

Writing speed: 9.82 MB/s

Reading speed: 18.88 MB/s



Chromatix said:

The Foundation chaps report that SanDisk Ultra series cards work fine.  I unearthed a SanDisk Ultra from one of my cameras, which is among the 50MHz, 4-bit interfaced examples.
[...]

This should hopefully be corrected when the cards I have ordered finally arrive, as they include a SanDisk Extreme series card which claims a UHS-1 performance rating



mightygoose said:


FWIW i have a 8gb samsung class10 micro SD waiting to test as soon as my pi arrives as well as a 8gb class 4 as a backup.



Chromatix said:


I've just got hold of Class 10 4GB cards from both SanDisk (Extreme series) and Transcend (black label).




Chromatix said:


After exploring the nether regions of my friendly local computer megahyperstore, I now have an extra selection of SD (and MicroSD) cards from two additional vendors (Kingston and Verbatim) to play with.  Most of them are Class 10s.



Now, i'm gonna order a Panasonic 8 GB SDHC Gold (UHS-1 / Class 10). They claim it can read at 90MB/s, and i'd guess the read speed is very important for our intended use as primary linux filesystem. Shall we use  f3write to compare results once the r-pi arrives?

User avatar
Chromatix
Posts: 430
Joined: Mon Jan 02, 2012 7:00 pm
Location: Helsinki

Re: Class 10 SD cards on the production boards

Fri Jan 27, 2012 12:02 am


The straight-line performance, while easy to measure, is really *not* so important here.  Most SD cards are like muscle cars - cheap and good at drag racing but hopeless at going around corners.  Linux filesystems are mountain roads, not the drag strips that camera makers deliberately lay out to accommodate the muscle cars.

It took 35 entire minutes just to *format* the 8GB Class 10 "UltimateX" Kingston card I bought today - yet I could have written a complete 8GB image to it in under 10 minutes using sequential writes.  This is a card sold at a premium price as having superior performance - yet a SanDisk Extreme beats it in every category I can measure and is cheaper.  The practical performance is so bad that people might well think the computer has frozen during the formatting operation.

That's by far the worst example I've yet seen, but it was entirely expected - Kingston are well known to use the simplest, dumbest controllers possible in their cards to minimise cost, and it is also well known that anything besides camera write patterns will make them choke badly.  Some of my other cards are 10 times as quick as that, most are at least 4 times including the Transcend MicroSD card (in an adapter) that I just tested.

BTW, the Verbatim card turned out to be totally unspectacular - more like last year's less good Transcend models - but it works a lot better than the Kingston rubbish.  Transcend and SanDisk by contrast have obviously implemented true upgrades over the past year, so that any of their SDHC models are now apparently good enough for serious use with Linux.

The key to knowledge is not to rely on people to teach you it.

WereCatf
Posts: 89
Joined: Thu Jan 26, 2012 7:16 am
Contact: Website

Re: Class 10 SD cards on the production boards

Fri Jan 27, 2012 12:09 am

Chromatix said:


The straight-line performance, while easy to measure, is really *not* so important here.  Most SD cards are like muscle cars - cheap and good at drag racing but hopeless at going around corners.  Linux filesystems are mountain roads, not the drag strips that camera makers deliberately lay out to accommodate the muscle cars.


Would you like to clarify what you mean?

I get the feeling you don't really fully understand what you're talking about as Flash devices have exceedingly good seek times; they do not have mechanical parts so there is no actual moving of a head needed and seek times are constant across the whole device. On conventional hard disks seek times are generally atleast 7ms, often even up to 11ms, whereas Flash devices have <2ms seek times. This means they perform really, really good on "non-straight roads with lots of corners", ie. in situations where you need to seek often, like e.g. reading in lots of small files.

User avatar
Chromatix
Posts: 430
Joined: Mon Jan 02, 2012 7:00 pm
Location: Helsinki

Re: Class 10 SD cards on the production boards

Fri Jan 27, 2012 1:47 am

Reading isn't the problem - at least not directly.  If you do nothing but read an SD card, it will generally perform quite well - probably even with access times below those of a hard disk.  The speed does still vary with the quality of the card, but it's much easier to find a card that performs well for reading than for writing.

The problem is that a typical Linux distro, doing normal Linux desktop things, does one heck of a lot of small writes to the disk.  And at no time is this more true than when initially installing it - when the user has the least experience of what to expect and is at most risk of giving up and running back to the "devil they know" - Windows.  Would you wait 35 minutes just for the format phase (without even looking for bad sectors!) to complete?

It is these small writes which SD cards have so much trouble with.

Now, normally writes being slow isn't really a problem, as the user should be able to carry on using the computer while this super-advanced operating system gets on with the writing behind the scenes.  But there *is* a problem if the OS needs to *read* something while the disk is busy writing.  Cheap SD cards can only do one thing at a time - so while they are taking 2 seconds to write that single Firefox cache file, Firefox now wants to read a different cache file (say to do with the URL history for autocompletion), and the user is now waiting for it.

Slow writes are also a major problem during shutdown, because users have a habit of being impatient and might yank the power before everything has properly hit the disk.

There are several sites on the Web which talk about SD card performance, and some even go into detail about what's happening inside, explaining why it goes so slowly on some cards.
The key to knowledge is not to rely on people to teach you it.

Return to “General discussion”