SDCard Choice

Help and Support
21

Here you go @sam_nazarko and @fzinken .

OSMC_installer installed OSMC November edition. Didn’t update a thing, apart from the self updating add-ons. Did a reboot and waited for OSMC to settle down…

Pretty much the same picture:
osmc@osmc:~$ dmesg | grep sdhost osmc@osmc:~$ dmesg | grep mmc [ 0.000000] Kernel command line: dma.dmachans=0x7f35 bcm2708_fb.fbwidth=1920 bcm2708_fb.fbheight=1080 bcm2709.boardrev=0xa01041 bcm2709.serial=0xeb9f4189 smsc95xx.macaddr=B8:27:EB:9F:41:89 bcm2708_fb.fbswap=1 bcm2709.disk_led_gpio=47 bcm2709.disk_led_active_low=0 sdhci-bcm2708.emmc_clock_freq=250000000 vc_mem.mem_base=0x3dc00000 vc_mem.mem_size=0x3f000000 root=/dev/mmcblk0p2 rootfstype=ext4 rootwait quiet osmcdev=rbp2 [ 1.111910] mmc-bcm2835 3f300000.mmc: mmc_debug:0 mmc_debug2:0 [ 1.111925] mmc-bcm2835 3f300000.mmc: DMA channels allocated [ 1.184583] mmc0: host does not support reading read-only switch, assuming write-enable [ 1.186768] mmc0: new high speed SDHC card at address 0007 [ 1.187726] mmcblk0: mmc0:0007 SL16G 14.4 GiB [ 1.189451] mmcblk0: p1 p2 [ 1.381262] EXT4-fs (mmcblk0p2): mounted filesystem with ordered data mode. Opts: (null) [ 1.391296] fsck: E2fsck run: /bin/e2fsck -p -C 0 /dev/mmcblk0p2 [ 1.391773] fsck: /dev/mmcblk0p2: clean, 25357/932064 files, 235744/3727104 blocks [ 3.565937] EXT4-fs (mmcblk0p2): re-mounted. Opts: (null) [ 4.483596] FAT-fs (mmcblk0p1): Volume was not properly unmounted. Some data may be corrupt. Please run fsck.

Block Size 4k Write 0.24707 Rewrite 0.243164 Read 5.31641 Reread 5.35059 Random Read 4.81543 Random Write 0.235352 Block Size 512k Write 8.39355 Rewrite 5.84082 Read 17.9795 Reread 18.1318 Random Read 18.084 Random Write 2.6875
The thing is, it is really obvious. The bare installation is way slower, most of the notifications were just flashing by with the EVOs. With the SanDisk I can read them pretty clearly almost one by one. The whole system is noticeably slower.

…next stop, the dist-upgrade following a reboot…

osmc@osmc:~$ dmesg | grep sdhost [ 1.257124] mmc0: sdhost-bcm2835 loaded - DMA enabled (>1) osmc@osmc:~$ dmesg | grep mmc [ 0.000000] Kernel command line: dma.dmachans=0x7f35 bcm2708_fb.fbwidth=1920 bcm2708_fb.fbheight=1080 bcm2709.boardrev=0xa01041 bcm2709.serial=0xeb9f4189 smsc95xx.macaddr=B8:27:EB:9F:41:89 bcm2708_fb.fbswap=1 bcm2709.disk_led_gpio=47 bcm2709.disk_led_active_low=0 sdhci-bcm2708.emmc_clock_freq=250000000 vc_mem.mem_base=0x3dc00000 vc_mem.mem_size=0x3f000000 root=/dev/mmcblk0p2 rootfstype=ext4 rootwait quiet osmcdev=rbp2 [ 1.257124] mmc0: sdhost-bcm2835 loaded - DMA enabled (>1) [ 1.314441] mmc0: host does not support reading read-only switch, assuming write-enable [ 1.316380] mmc0: new high speed SDHC card at address 0007 [ 1.327357] mmcblk0: mmc0:0007 SL16G 14.4 GiB [ 1.328975] mmcblk0: p1 p2 [ 1.540038] EXT4-fs (mmcblk0p2): mounted filesystem with ordered data mode. Opts: (null) [ 1.550525] fsck: E2fsck run: /bin/e2fsck -p -C 0 /dev/mmcblk0p2 [ 1.551023] fsck: /dev/mmcblk0p2: clean, 27364/932064 files, 284908/3727104 blocks [ 3.746246] EXT4-fs (mmcblk0p2): re-mounted. Opts: (null) [ 4.616725] FAT-fs (mmcblk0p1): Volume was not properly unmounted. Some data may be corrupt. Please run fsck.
Block Size 4k Write 0.237305 Rewrite 0.233398 Read 3.8457 Reread 3.87695 Random Read 3.5957 Random Write 0.233398 Block Size 512k Write 4.02344 Rewrite 5.90137 Read 21.6191 Reread 21.6357 Random Read 21.582 Random Write 2.71582
There is one thing I remember. When my OSMC adventure started I had a couple of SD card brands and sizes on my Amazon wishlist. Ultimately I went for the 16GB SanDisk Ultras because of all the reasons we like the SanDisk Ultras. Anyway, a couple of days later, when I was ready to buy, there was this “there’s a newer version of this product” window when I selected the SanDisks. I could’t tell any difference so I went for the “latest model” as it was labeled. The rest is history… Maybe there is a change in firmware with the so called latest model, maybe my card is messed up. In terms of speed, it behaves like my ancient 4GB Class 4 card, which sucks. Anyway, if there’s something I can test for you guys or something else I can do, just drop me a line. Maybe we can save some people the hassle with the latest Ultras.

Cheers
Ralph

22

Hi Ralph,

I can tell you that the SanDisk cards we stock certainly don’t behave like this. @DBMandrake has some of our SanDisk cards. While he finds the Samsung ones faster, the SanDisk ones do perform very well and one thing is for certain, they’re better than your run of the mill Class 4 SD card.

Not sure about this, but we ship the very latest offering from SanDisk and haven’t had problems with it. Do you have another SanDisk card? Are the results similar? Otherwise I would be inclined to believe you have a faulty / bad card. SanDisk and Samsung cards should be near identical with performance, unless you’re doing something like torrenting to an SD card, which isn’t recommended anyway.

Sam

23

If you bought them online there’s a good chance they’re fake Fake memory cards out there – Don't be fooled- Technology News, Firstpost

24

Got the card and ran the first benchmarks … this looks promissing and as expected if you look at the benchmarks posted on the RPi forum posted above.

I’m installing at the moment … so I can run the iozone benchmarks when that is finished :smile:

iozone - Samsung EVO 32GB [MB-MP32D(A/EU)] - mmc

Block Size 4k Write 1.68652 Rewrite 2.36816 Read 6.98438 Reread 7.02832 Random Read 7.00586 Random Write 2.32715 Block Size 512k Write 16.4824 Rewrite 16.0771 Read 18.4512 Reread 18.4521 Random Read 18.4482 Random Write 17.3564

iozone - Samsung EVO 32GB [MB-MP32D(A/EU)] - sdhost

Block Size 4k Write 1.33301 Rewrite 2.16895 Read 5.62598 Reread 5.63477 Random Read 5.62891 Random Write 2.33789 Block Size 512k Write 11.9189 Rewrite 12.0498 Read 12.5283 Reread 12.4961 Random Read 12.5273 Random Write 12.1367

flash-bench - Samsung EVO 32GB [MB-MP32D(A/EU)] - mmc

Benchmark: Sequential read Limits: Total size: 512MB Duration: 60s 512.0MB processed in 27.76s (18.45MB/s), CPU: user 0.06%, sys 1.70% Benchmark: Sequential write Limits: Total size: 512MB Duration: 60s 512.0MB processed in 38.11s (13.44MB/s), CPU: user 0.11%, sys 2.43% Benchmark: Random read Limits: Total size: 512MB Duration: 60s 293.8MB processed in 60.04s (4.89MB/s), CPU: user 0.00%, sys 2.84% Benchmark: Random write Limits: Total size: 512MB Duration: 60s 236.9MB processed in 90.22s (2.63MB/s), CPU: user 0.01%, sys 0.72%

flash-bench - Samsung EVO 32GB [MB-MP32D(A/EU)] - sdhost

Benchmark: Sequential read Limits: Total size: 512MB Duration: 60s 512.0MB processed in 41.01s (12.49MB/s), CPU: user 0.06%, sys 1.14% Benchmark: Sequential write Limits: Total size: 512MB Duration: 60s 512.0MB processed in 54.71s (9.36MB/s), CPU: user 0.09%, sys 2.05% Benchmark: Random read Limits: Total size: 512MB Duration: 60s 258.4MB processed in 60.05s (4.30MB/s), CPU: user 0.04%, sys 2.49% Benchmark: Random write Limits: Total size: 512MB Duration: 60s 216.1MB processed in 92.21s (2.34MB/s), CPU: user 0.01%, sys 0.64%

Looks like sdhost needs some more loving since results have gone down it seems :hushed:

25

Hi Sam,

I bought 3 of 'em but only used one so far. The other packages are still sealed and as benchmarks are showing I’m inclined to send them back and get the Samsung PROs instead.

I’d like to hear your opinion on returning the one we we’re talking about as faulty. My guess, it’s a long shot. I reformatted the card as ExFAT and copied a 500MB file back and forth on a Windows box. 12MB/s read speed and 8.5MB/s write speed. Pretty good actually…

At the end of the day, what the heck, it’s 8 bucks. I slap on an SD adapter and put it my DSLR.

Ralph

26

I would suggest a program like Crystal Disk Mark 3.0.3 for a better benchmark…

27

While certainly worth a try, I prefer the “real world” stats on the platform I intend to run my stuff, which in this case is the PI2 on the various software platforms (primarily OSMC for media centre). Therefore I think I stick with iozone.

For diagnosing if a card is faulty or not or general benchmarking, it would certainly help to switch software and/or platforms.

Nonetheless, thanks for testing and sharing man. I’ll stick with the Samsung PROs for the time being.

Now, I think I’m going to actually use OSMC for the rest of this year instead of benchmarking it, for it is a nice piece of software! :grin:

Cheers to y’all
Ralph

28

I meant instead of a windows file copy :wink:
Lol offcourse I prefer iozone as well since you compare on the same platform :relieved:

29

That’s what counts. But a CrystalDiskMark benchmark would be interesting. It would be good to see if you get the same performance on the PC as you do with the Pi. This not being the case could hint at a regression with SDHost.

Sam

30

Is that on a Pi 1 or a Pi 2 ?

Your evo results are quite a bit slower than mine, but mine were taken on a Pi 2.

31

RPi2

32

First and foremost, Happy New Year to everyone!

Back to the topic. Here’s a good one:

I did a CrystalDiskMark in two different SDCard adapter and here’s my result. Now I did not fiddle with the settings and if I didn’t overlook something, I’d conclude that the iozone results are somewhat consistent with these. Or CrystalDiskMark doesn’t show results less than .5 . I did the test multiple times, because at one point I wasn’t sure if it was benchmarking altogether.

Seems to me the card is the culprit, right?

Ralph

33

I would suggest version 3.0.3 that I linked, it has 512 and 4k, should display it better though, but 0.000 write … autch :anguished:

34

…redid the tests with version 3.0.3… This one seems to be a tad more chatty. Gave me a write error when it reached the 4k test. Reformatted the card to FAT32 (full format, not the quick format), and reran the test. Same error when it came by the 4k stuff. Since I don’t want to bore you any longer with this stuff, is it safe to assume that the card is faulty/sub par at this point?

If there are some more formatting options/tricks/best practices that might be worth trying, feel free to share. :smiley: Happy to test them…

Ralph

35

Almost save to presume it’s close to a near dead …
You could try sd card formatter or something from SanDisk self … that would suggest a last try … next stop would be the bin.

36

SD cards are so cheap now I would just bin any one that is giving errors when trying to test it…

Of all the SD cards I’ve got none have failed on me yet, but they are all Samsung, sandisk and noobs cards obtained from reputable sources. (Eg not counterfeits) There is a surprising percentage of counterfeit SD cards on places like eBay where any random person can sell…

37

Except for the 8GB SanDisk which came from the PiHut as part of my set, the rest I got from Amazon so I’m naive enough to think they’re the real deal. No worries here, I can still return them. In fact I already triggered that.

Probably a faulty/sub par card and that’s it. It happens. I’ll call it a day for my part on this one.

38

I bought one of my Samsung evo cards from Amazon - I think Amazon are a lot safer than eBay when it comes to avoiding counterfeits but it’s not out of the question.

It may just be faulty.

39

I’d trust sdcards direct from Amazon.
However Amazon Marketplace isn’t much different from eBay, so I’d be wary.

40

Indeed, and if you have issues, Amazon will allow you to return those cards without any problems.