128GB Lexar SDXC card benchmarks


Grench

Forum Addict!
Joined
Oct 3, 2008
Messages
6,629
Just adding a few instances here. If someone is interested they can use these to update the pandorawiki SD card list. I don't have the time or interest to figure out how to update it. This specific card did not have an entry.


I had requested this to be done several times in other threads about the Lexar 128GB card. Rather than necroposting those threads, I started a new one.


Since I was re-arranging partitions and card usage and had the opportunity to - I ran these.


Read benchmarks were done using: dd if=/dev/mmcblk0 of=/dev/null bs=4MiB count=100


Write benchmarks were done using: dd if=/dev/zero of=/dev/mmcblk0 bs=4MiB count=100


Warning - that write benchmark IS DESTRUCTIVE. Don't do it unless you are willing/able to re-partition your card and start over.


Benchmarks were run 3 times each.


Card: Lexar Professional 128GB SDXC Class 10 133x


Obtained from: http://www.bhphotovideo.com/c/product/756124-REG/Lexar_LSD128CRBNA133_128GB_Professional_133x_SDXC.html


Left slot Read benchmark: 17.2MB/s 17.2MB/s 17.2MB/s


Left slot Write benchmark: 11.1MB/s 11.8MB/s 11.8MB/s


Right slot Read benchmark: 17.2MB/s 17.2MB/s 17.2MB/s


Right slot Write benchmark: 11.8MB/s 11.8MB/s 11.9MB/s


The following work fine (three 3MB partitions made with cfdisk used for testing):


mkfs.ext2 -L one /dev/mmcblk0p1


mkfs.ext3 -L two /dev/mmcblk0p2


gparted did not like my partitions two above - so I asked it to format them as ext3 and created a third fat32 partition respectively. gparted generated an error as if it didn't work - but the next screen showed the partitions as I would expect them to look. I'm putting this part up as a weirdness in gparted for now.


Checking them in cfdisk/mmcblk0 resulted in something other than expected for the ext3 parition - so I umounted /dev/mmcblk0p2 and re-formatted it again with the mkfs.ext3 as above.


Through that mess my cfdisk /dev/mmkblk0 now looks like:


mmcblk0p1 Primary Linux ext2 [one] 3000.01


mmcblk0p2 Primary Linux ext3 [two] 3000.01


mmcblk0p3 Primary W95 FAT32 3000.01


Pri/Log Free Space 123035.62


So - formatting them worked in the end, though the gparted tool seems to have issues with creating ext3 partitions on the Pandora - though it works fine for creating fat32.


moving on...


# umount /dev/mmcblk0p1


# fsck /dev/mmcblk0p1


fsck from util-linux-ng 2.17


e2fsck 1.41.9 (22-Aug-2009)


The filesystem size (according to the superblock) is 1220702 blocks


The physical size of the device is 732422 blocks


Either the superblock or the partition table is likely to be corrupt!


Abort<y>? n


However - if I re-format the same partition again and instead run:


# e2fsck -p /dev/mmcblk0p1


I get:


one: clean. 11/183264 files. 12586/732422 blocks


So I tried to make it fail by running:


# e2fsck -fp /dev/mmcblk0p1


I get:


one: 11/183264 files (0.0% non-contiguous), 12586/732422 blocks


OK - so that seems fine. There seems to be a disconnect between what e2fsck does when called directly and what fsck does (despite it seemingly calling e2fsck). No idea.


Moving on...


# e2fsck -fp /dev/mmcblk0p2


I get:


two: 11/183264 files (0.0% non-contiguous), 28987/732424 blocks


I can't run fsck on 3 because it's fat32.


# dosfsck /dev/mmcblk0p3


I get:


Seek to -1294959104:Invalid argument


No idea...


All 3 partitions read/write data fine.


There seem to be some big inconsistencies in how different partitioning (cfdisk and gparted) and formatting (mkfs and gparted) and file system analysis (fsck, e2fsck, dosfsck) work on the Pandora.


This may help explain why one person will have rotten luck getting ext3 running on an SD card - and another person has no issues (gparted vs cfdisk). There is no consistent standard for measurement of what constitutes a 'failure' on the SD card compatibility list. In some cases this may be a failure in the utility being used rather than the card? Maybe.


Thank you.
 
In some cases this may be a failure in the utility being used rather than the card? Maybe.
Quite possibly. Equally, some are due to failures of cards with poor controller hardware - even some big-name ones.
 
Left slot Read benchmark: 17.2MB/s 17.2MB/s 17.2MB/s


Left slot Write benchmark: 11.1MB/s 11.8MB/s 11.8MB/s


Right slot Read benchmark: 17.2MB/s 17.2MB/s 17.2MB/s


Right slot Write benchmark: 11.8MB/s 11.8MB/s 11.9MB/s
These are pretty similar to the 16GB Transcend class 4 cards I'm using: reads in around 16MB/s, writes at about 12MB/s. The Pandora's theoretical max of 24MB/s is under perfect conditions and doesn't account for overhead. I suspect these numbers are the best we'll ever get with the controller in the OMAP. Anything more than a reasonable class 4 is likely to be wasted unless you use a more powerful reader for copying things onto it from your PC, and even then your only benefit is reduced time copying large volumes onto it the one time.

This may help explain why one person will have rotten luck getting ext3 running on an SD card - and another person has no issues (gparted vs cfdisk). There is no consistent standard for measurement of what constitutes a 'failure' on the SD card compatibility list. In some cases this may be a failure in the utility being used rather than the card? Maybe.
Not entirely certain what you're meaning here, though.
 
I was reading about these last night on fakeflashnews and flashchiptech ... i had no idea there were SD/MicroSD available in larger than 64GB either.


These are v rare though and only a couple of companies are currently making them.


I'd like one but they a bit on the expensive side for me to fork out for, would be lovely having a couple of those in your Pandora though eh
 
Left slot Read benchmark: 17.2MB/s 17.2MB/s 17.2MB/s


Left slot Write benchmark: 11.1MB/s 11.8MB/s 11.8MB/s


Right slot Read benchmark: 17.2MB/s 17.2MB/s 17.2MB/s


Right slot Write benchmark: 11.8MB/s 11.8MB/s 11.9MB/s
These are pretty similar to the 16GB Transcend class 4 cards I'm using: reads in around 16MB/s, writes at about 12MB/s. The Pandora's theoretical max of 24MB/s is under perfect conditions and doesn't account for overhead. I suspect these numbers are the best we'll ever get with the controller in the OMAP. Anything more than a reasonable class 4 is likely to be wasted unless you use a more powerful reader for copying things onto it from your PC, and even then your only benefit is reduced time copying large volumes onto it the one time.
Were those Transcend class 4 cards made before the class 6 and class 10 were accepted by the SD association? If they are sustaining > 10MB/s writes, then if they were made today they would be called class 10 cards.

This may help explain why one person will have rotten luck getting ext3 running on an SD card - and another person has no issues (gparted vs cfdisk). There is no consistent standard for measurement of what constitutes a 'failure' on the SD card compatibility list. In some cases this may be a failure in the utility being used rather than the card? Maybe.
Not entirely certain what you're meaning here, though.
gparted (from PNDstore) would not properly format an ext3 for me. It would properly see the results of doing it on the command line though.
 
Were those Transcend class 4 cards made before the class 6 and class 10 were accepted by the SD association? If they are sustaining > 10MB/s writes, then if they were made today they would be called class 10 cards.
No, they're relatively new, certainly long after Class 6 was well established. I believe you are misunderstanding what "class" means. The class of a card is the minimum sustained write speed, not the average. A class 4 might average 12MB/s write speeds, but occasionally might drop to as low as 4MB/s, if even for a fraction of a second (but typically for a noticeable number of seconds) while the controller does whatever it needs to do to continue writing data. A class 10 might average the same 12MB/s but it is guaranteed not to drop below 10MB/s in a non-fragmented sequential write, provided the writing device continues to supply a stream of data of course. The practical benefit of the class is when you absolutely need sustained write minimums, such as when recording audio or video. There is no guaranteed correlation between class and read/write speeds in excess of that class, nor does the class make any assertion about the read speed.


In reality, minimums are just that: absolute minimums, and average write speeds tend to be much higher, with read speeds exceeding that still. In most cases, SDHC cards are fully capable of achieving average sequential writes 3 to 4 times their minimum, only occasionally dropping to their minimums to change banks or verify data or whatever else the controller chip occasionally needs to do. This also applies to class 10 cards which, under perfect conditions, have seen sustained average write speeds of 30 to 40MB/s (sometimes higher) in devices that can support such high speeds. Unfortunately, the OMAP in the Pandora is limited. That limit seems to be around 10-15MB/s, the rate at which a reasonable class 4 card will operate on average.


The real benefit of a class 10 card for the Pandora is when copying large files to it from your PC, assuming it's a reasonable card. Benchmarks of the card in a different reader attached to your computer would be very interesting.
 
Are those numbers from old or experimental kernel?


Rebirth or older pandora?

Sorry for the delay - this got buried in the thread pile.


I'm running HF6. I have had my Pandora for a year now.


The 128GB Lexar card has been running like a champ. After all of my fiddling and fidgeting around with it, I've settled on using a single giant ext2 partition - for the moment. The additional benchmarking that I did in another thread (inaptly named now, but the topic has shifted through time): http://boards.openpandora.org/index.php?/topic/7576-turn-on-my-swap-partition/page__st__20#entry139494


The Lexar Professional 128GB SDXC card shows faster sustained writes and reads than the Patriot LX. It's not just a huge card - it's fast in the Pandora too. It's been stable, notwithstanding an event that I caused that took down both cards - I can't blame the cards for what I did to the device.


I'm very happy with it.


As far as costs go - it's roughly 4X the price of quality 32GB SDHC media +$10 or so. It's a big chunk of change, but not outlandish for what you get. A gigantic SD card that just happens to be quite fast too.
 
Last edited by a moderator:
Are those numbers from old or experimental kernel?


Rebirth or older pandora?

Sorry for the delay - this got buried in the thread pile.


I'm running HF6. I have had my Pandora for a year now.


The 128GB Lexar card has been running like a champ. After all of my fiddling and fidgeting around with it, I've settled on using a single giant ext2 partition - for the moment. The additional benchmarking that I did in another thread (inaptly named now, but the topic has shifted through time): http://boards.openpa..._20#entry139494


The Lexar Professional 128GB SDXC card shows faster sustained writes and reads than the Patriot LX. It's not just a huge card - it's fast in the Pandora too. It's been stable, notwithstanding an event that I caused that took down both cards - I can't blame the cards for what I did to the device.


I'm very happy with it.


As far as costs go - it's roughly 4X the price of quality 32GB SDHC media +$10 or so. It's a big chunk of change, but not outlandish for what you get. A gigantic SD card that just happens to be quite fast too.

Could you give me a quick run down of how you formatted it? I've been fiddling with a 128GB card for 3 days andjust when I found a stable way to format it to fat32 I accidentaly wipe it while trying to fix another card. Instead of trying to recreate what I did again I figured I might as well ask. I'm pretty sure the way I did it was wrong anyways.


I had basically created a fat32 partition with gparted first. then I went into that partition with fdisk (ie using /dev/mmcblk0p1) and split that up into 4 equal sized partitions.For gods sake and all that is holy someone correct me if I'm wrong in doing this. It seemed stable and fast but I didnt give myself time to actually test that because of what I mentioned earlier.


The main problem I encounterd before I did what I mentioned above is that it occasionally locks up and either wont let me read/write or wont let programs read/write. I run fsck -a and fsck -r on its ass but the solution it comes up with usually involves deleting whatever new files or folders I or a program created. Sometimes t would mention something about a bad superblock but I dont remember exactly what. Could go through recreating the rror again if anyone cares. So yea, I think I'm pretty much ready to move on to ext2 if that seems to be working for you. I dont think it matters but my sd card is sandisk brand, not lexar. Als, it's not faulty and definitely hasthe full 128 GB of space. (125.4 after ft32 format).


#imnotsmartenoughforthis
 
Last edited by a moderator:
Hmmm. I'm not sure how much I can help you with FAT32 - Right now I'm not using it on anything.


I may have to start using FAT32 or ExFAT though - my wife's new digital camera uses ExFat by default.


A quick search turned up NearTao's instructions for formatting just this card as FAT32 though:





The big catch with FAT32 on the Pandora is that you have to opkg your way into another package - which in my experience was dangerous. I soft-bricked my SD card OS once trying to opkg my way through something.


Hey - an idea - can someone create a FAT32.pnd so that this could be easy?


Grench

Oh - found something else. Free open source code for ExFAT on Linux.


http://code.google.com/p/exfat/


So - those could theoretically allow the Pandora to read any common SD card format, right?
 
No worry about your cards, this seems to be rather a GParted / OpenPandora OS software than a hardware issue.


Already reported this ages ago en detail (on the gp32x forums), seems that development of the GParted PND is stalled.


Maybe someone (of the core team) carries on? Formatting SD cards seems like a core utility, doesn't it?


THE ISSUE SUMMARIZED:


GParted - Most operations are aborted as the partitions get automatically mounted again


FOREWORD:


Of course I run the GParted PND from the SD card on which I perform NO operations.


BUG REPORT:


For the most operations, GParted aborts with an error!


It is practically un-useable. Many attempts, very few successes.


I've looked into the error messages, and their common problem seems to be:


Between the steps of an operation, some OS process seems to re-mount the partition. And for safety reasons, GParted has to abort, as it cannot operate on mounted partitions.
 
Last edited by a moderator:
I was recently chastised for stating that Gparted didn't work properly. It sounds like I'm not alone in that.


Combined gparted fixes and making fat32 and ExFAT formatting/use run smooth is, IMHO, hugely important.
 
I was recently chastised for stating that Gparted didn't work properly.
You were chastised for making provably false statements and carrying on an argument that no one was debating against.

After I made this post:


http://boards.openpandora.org/index.php?/topic/7963-firmware-suggestions-and-contributions/page__st__60#entry148098


I received 9 posts in reply telling me how wrong I was.


The last one is actually from you suggesting that fat32 was all solved because mkfs.vfat is available. You are partially right - but I was specifically referencing gparted in my comment. I had assumed that it was failing because the fat32 formatting tools were still absent - my mistake. It still doesn't work right with gparted - which is everyone's default direction to send all of the newbie questions to.


Yah, OK. No one debating against. OK. If you say so then it must be right.


The root of the argument - I think you and I agree on.


/************ back on topic **************/


The Pandora needs as much flexibility around SD card media as possible. That includes full functional support (formatting, reading, writing, repairing) FAT32, ExFAT, EXT2 and maybe EXT3 and EXT4 though those can get goofy on removable media.


The easiest way for a new-to-Linux user to partition and format SD cards is Gparted - which is currently only quasi-working.


Describing the Gparted issue... If you start with a card that the Pandora can not mount (no partition or EXT3 formatted partition) then you can use Gparted to re-partition and format EXT2 (and apparently FAT32 now though I haven't tried it myself). Works - that time. Reboot or re-insert the card and Gparted can't seem to keep it unmounted long enough to actually perform any actions on it.


Is it Gparted that is remounting the SD slots or is it something to do with the way the Pandora's OS is dealing with SD cards? Are we auto-remounting them every 3 seconds or something?


I recall Gparted working better on the earlier versions of the OS, but that could just be 1st time on a new card related. I'm not sure if or what changed.
 
Last edited by a moderator:
The root of the argument - I think you and I agree on.
Absolutely. What I don't agree with and what I absolutely don't appreciate is when you try to make yourself a martyr to a cause that doesn't need one. If there's a problem with gparted, say there's a problem with gparted, and if someone points out that the problem you are experiencing really can't be what you think it is say "oops, good point. Here is what is happening and maybe it is something else then..." not "I keep getting chastised for pointing out problems". There's no fault in being wrong, I'm wrong frequently, but when you try to cover it up like that it doesn't help anyone.


Thank you.
 
The root of the argument - I think you and I agree on.
Absolutely. What I don't agree with and what I absolutely don't appreciate is when you try to make yourself a martyr to a cause that doesn't need one. If there's a problem with gparted, say there's a problem with gparted, and if someone points out that the problem you are experiencing really can't be what you think it is say "oops, good point. Here is what is happening and maybe it is something else then..." not "I keep getting chastised for pointing out problems". There's no fault in being wrong, I'm wrong frequently, but when you try to cover it up like that it doesn't help anyone.


Thank you.

Meanwhile, I hope you don't mind that I take offense at your slamming my moral and ethical integrity (cover up? WTF?) in casual posts where I have backed up my claims to levels far beyond any required of any other user on here.


Frankly, I could state that, "The time is 12:25 PM," and have 4+ people require proof that I was, indeed, in a time zone where at that moment it was 12:25 +/- 60 seconds.


Not a martyr, just a guy with a big target on his head for having called BS on several others here. I'm fine with that.


However, I state things as I see them. I provide information to back up my claims. I even answer questions as they arise to the best of my ability based on the information at hand. There is a huge tendency on these boards to answer serious questions with personal quips that don't answer anything.


You're welcome to provide additional information around your experience and alternate solutions and information that may change the outcome - yes, my opinion can change with new information.


If you're going to make 100% clearly false statements about my personal integrity, though, I can and will take offense at it.


"You were chastised for making provably false statements and carrying on an argument that no one was debating against." False.


"you try to cover it up like that it doesn't help anyone." False. I have -never- covered up my mistakes. I do not re-edit posts. I post resolutions to issues that I found - as well as the false tracks I followed on the way to my solution. If anything, I'm guilty of providing too much information of the problem solving involved.


" if someone points out that the problem you are experiencing really can't be what you think it is say "oops, good point. Here is what is happening and maybe it is something else then...""


Because clearly if someone else states, "It can't do that," while offering no explanation or proof - they must be right and I must be wrong even though I am watching it "do that". OK. Sure. Do you have any clue how asinine that logic is? If "someone" states it "can't be what you think it is"? Wow - we're going to need a list of these people who are always right so we know who's statements are infallible and should never be questioned.


In that previous thread I supplied proof that what the naysayers were stating wasn't actually true - that we could get the source code for ExFAT and that we could use it for free under a GPL license. None of them bothered to even look at the link I provided to the source and license before going off on it.


Do you have any idea how insulting you're being?
 
Back
Top