Originally Posted by epitaph

Hi, I'm not sure what your are talking about but I just follow your suggestion. I didn't have time to read the website. Isn't there not a simple program to benchmark? I want to test time dd if=/dev/null of=/dev/mmcblk0p1 count=512 bs=512 because I remember it is a good benchmark, too. PLEASE don't try this yourself. It's not tested!. What do you think is the difference? Me just to lazy to write my own shell-script?

EDIT:Here is a good tutorial how to use dd non-destructive and dma: http://wiki.linuxquestions.org/wiki/..._of_your_disks

Originally Posted by topet2k12001

Thanks for the share. Hm, well in my opinion this is the simplest I found. I'm sure someone will be able to make something better, or perhaps someone may come up with a GUI for the F3 program to make it "point-and-click" friendly.

Or, if you have a Microsoft-based OS you can try the other program I attached; I find it simple for those who are not used to the command line since it has a GUI ("point-and-click" with the mouse).

As far as difference, yeah I found differences too as mentioned in Post #1 (which, based in my own tests, seem to differ based on OS used, benchmarking program used, manner on which the storage device is plugged).

I did read in one of the threads here though, that there are benchmarking tests that do a "sequential" write and there are others that do a "random" write. I'm no programmer and no techie guy, so the way I understand it, the speed being reported differs depending on how the test runs (e.g. sequential or random)? Maybe the community can help me understand.

Anyway it won't matter much to me quite honestly, since there is always this variance in reported speeds, what I just took into consideration is whether my MicroSD card is what it is; in other words as long as it is reporting the speed that it should at least report (assuming that the benchmarking program is "legit"), then I would be comfortable knowing that I didn't purchase a fake card. Whether it reports a higher benchmarking speed than the "promised" speed is just a bonus or a plus, but not really a requirement for me.

Originally Posted by epitaph

I don't want to convince you to use dd but my SD-Card Class 6 has an average speed of 17 MB/s which is almost 2x the speed of your class?

Originally Posted by topet2k12001

Yup, I know.

Perhaps it's the brand, my friend. That could be one other factor that we forgot to look at among other factors already mentioned.

Class 6 cards are hard to find from where I live so I have fewer choices than others here. That's why I mentioned that as long as it reports to be genuine and hence doesn't make me feel that I wasted money, I'm all good. Thanks for your showing your concern, by the way.

I may try the dd; never said I wouldn't...I just need some time to read how to do it correctly. But I know that I will just get tired knowing that different benchmarking programs will give different results anyway...so only one thing that will give me peace of mind that is consistent will be the question as to whether my card gives me the "promised" 6MB/s or not. If another benchmarking program reports a faster speed for my card, it's a bonus and a plus. If not, I at least expect it to report that my card is a genuine Class 6.

Originally Posted by epitaph

I'm sorry I have a 8 GB Card Class 4 from Transcend!!!!!

Originally Posted by topet2k12001

Wow, that's great to hear! That means we both got the value for our money then, since we both got the "minimum" advertised/promised speeds mentioned by our respective products. Whether we got more than that is a bonus, but at least we both are getting the speed promised by the manufacturer.

Originally Posted by epitaph

I didn't show the write benchmark with dd:
dd if=/dev/zero of=mein.img bs=512 count=1000
1000+0 Datensätze ein
1000+0 Datensätze aus
512000 Bytes (512 kB) kopiert, 0,0565491 s, 9,1 MB/s

But dd is only sequentiel!

Originally Posted by shadowjk

dd with directio flags doesn't particulary mean "dma". It just means bypassing buffers/cache on linux, which gives linux less opportunity to optimize the transfers.

Which brings us to "bs=512". A 512 byte transfer at a time. Inside a MicroSD this can translate into a .5meg read/modify/write. Command overhead is also quite monstrous, so even if the card itself is able to optimize those 512 byte writes into full half meg writes, the bus itself would be saturated long before reaching 6Mbyte/s of payload data.