Which microSD card does speed matter?

Mike (mwester) mwester at dls.net
Wed Nov 26 14:20:42 CET 2008


Joop Boonen wrote:
> Yesterday I bought a SanDisk Mobile Ultra  microSD 4GB class 6 it seems as
> it was cheaper with an usb adaptor than a slower class2 version.
> 
> I have a feeling that the boot processes if much faster than the boot
> process with the supplied class 2 card.

There are multiple timing factors to consider, and the generic class
designation is really just "marketing-speak".  The transfer rate that we
can get on the GTA02 is limited, and will be far less than the 25MHz max
clock frequency (the same is true on the GTA01, although for different
reasons).  The kernel will tell you what speed it selected for the card
at boot time.  But that's just the speed at which the data is clocked in
once its ready.

The other factor is how long it takes from issuing the command to the
card to fetch a block of data until that data is ready to be transferred
in (latency).  Nothing currently collects and displays this bit of data
for the user, but it would be a very important factor in overall
performance, and it seems logical that higher-quality, higher-speed
cards should have better specs here.

It would be neat if someone could work up a patch to collect and display
a min/max and running average of the latency time as observed by the
kernel, but in reality I think that a quick test with bonnie++ on the SD
filesystem will provide a much truer picture of how the various cards
perform on the various devices.  Maybe someone should add such a page to
the wiki for users to submit their test results with various brand cards?

Mike (mwester)



More information about the hardware mailing list