New 95MB/s SanDisk SD card. Results.


Linux-SWAT

Forum Addict!
Joined
Feb 13, 2010
Messages
9,175
Bought two 16G Extreme Pro 95MB/s SDSDXPA-016G-X46 coming from Taiwan.

First, it beat all my previous cards bandwidth, and even blasted the theoretical limit of 16MB/s.

And second, the results are different between my CC and my 1G. Don't know if it's a known fact.

Tested a lot, rebooted, both units clocked at 800MHz.

CC

Read Write

18.9  17.4

1G

Read Write

20      19.1
 
Last edited by a moderator:
24M/s is the theoretical limit. 16M/s just seemed to be the max that we were getting.
 
Who needs a card with that much read/write speed? I think it is somewhat a waste of money than to just by the slightly cheaper ones that work flawless anyways, but that just may be me.
someone with a dslr in addition of his pandora?
 
A DSLR is a digital single lens reflex camera if i recall correctly.


A faster card would also be useful for people who run the OS on the SD.
 
guess it depends on the camera itself...

being able to take longer/faster burst might be important... when you already spend thousands $ on the camera and it's lens...
 
Last edited by a moderator:
A faster Card is better for Qemu Windows Images,because the read/write Speed is important.

How find you out the real Speed of your SD in your Pandora?

Is there a Benchmark for such Things?

I thougt the Card Reader in the Pandora itself is Speed Limited.

Have you tried the same Cards in PC for testing too?
 
I have a USB 3.0 card reader, and as i do a lot of copy/saves on my PC, i need fast SD cards.

On the OP, my "old" 45MB/s extreme pro was slower than the 30MB/s extreme.
 
And second, the results are different between my CC and my 1G. Don't know if it's a known fact.
The SoC revision used in CC Pandoras had a bug which could lead to data corruption when copying between the two SD slots. As far as I know, the workaround was to throttle transfer speed in the driver. This isn't required with GC Pandoras anymore.
 
Can you also test the random(4k)-write-speed of all your cards? That's the only thing that'll show significant differences on the Pandora, and is quite important if used for the OS.
 
I wrote a small script a few years ago to test latencies, found here. It's not perfect, but will do the trick. 

For a more real-world benchmark, you can also check /sys/class/block/*/stat after using the card for a while. Divide the 8th number by the 7th to get the average wait-time per write, in ms/sector.
 
Last edited by a moderator:
Back
Top