Cat 4 and Cat 10 micro SD cards have about same performance in two tests on the Ben!
ddf at sonic.net
Wed Oct 10 19:35:23 EDT 2012
I have just run some simple tests to see if there is much difference
between using a Cat 4 and a Cat 10 micro SD card in the Ben. I have
two Bens now and two Cat 4 32 GB cards and two Cat 10 32 GB cards.
The Cat 4 cards are made by Transcend: one is made in Taiwan and the
other in China. One of the Cat 10 cards is made by Samsung and the
other by Sandisk.
The first test was to transfer an ASCII file with a size of about 14
MB from my desktop to each Nanonote, doing it for two different cards.
The scp command was issued on the desktop computer and the Nanonotes
were connected via USB to that computer as well, to reduce network
overhead. Public/private keys were used so that no password was
entered. The average transfer speed for the Cat 4 cards was: 580 KB/s
(K=1024) and for the Cat 10 cards was: 620 KB/s. That's an increase
of transfer speed of about 7 per cent.
This change is much less than the alleged change in the maximum
transfer rate for each of the two categories of cards. Not much
advantage to using a Cat 10 card:-) This probably follows from the many
layers of software involved in such a transfer over USB. Lots of
factors contribute to transfer speed. For example, a larger file
would have had a faster rate in that the hand-shaking at the start of
the transfer adds a second or two of time.
The second test was compiling and linking a C program with 73 small
C-program files at opt level 2. This makes heavy use of swap space so
perhaps the Cat 10 cards would show some advantage. The time for the
Cat 4 cards, averaged over both cards and both Nanonotes was 713
seconds. The time for the Cat 10 cards was 666 seconds. This is a
decrease in time of about 7 per cent. However, it turned out that one
of the Cat 10 cards had an average time 4 seconds longer than one of
the Cat 4 cards! So again, real-world performance may not have much
relationship to the category assignment. Many other factors enter
into real-world performance.
Comparing the fastest Cat 10 card with the slowest Cat 4 card on the
compiling test showed a decrease in time of about 13 per cent. Not
much given the difference in cost for these two categories.
My overall conclusion is that for these two test, the difference
between Cat and Cat 10 is quite small. Often ratings of hardware are
based on conditions that maximize its performance. Real-world
applications often do not show much difference.
It would be interesting to have a small sample of cards of the same
category and same maker and then test them. I suspect that there
would be variations between cards nearly as large as I found here.
However, I do not want to spend more time on that. Getting precision
in benchmarking is difficult given the complexity of modern OS's and
So save your money and buy a Cat 4 card for the Nanonote:-)
More information about the discussion