On Tuesday 12 July 2011, Peter Warasin wrote:
Hi Guys
I *think* this card is not that bad.
Yes, looks pretty good. For completeness, please also
* list size in KB (from /proc/partitions/ or fdisk -lu) * make sure that the cutoff number for --random is the same * for cards that don't do --random (not this one), also check if there is a FAT optimization (--find-fat --erasesize=xxx)
==> /sys/block/mmcblk0/device/manfid <== 0x000009
==> /sys/block/mmcblk0/device/oemid <== 0x4150
Ok, so the manufacturer id is '9', haven't seen that before. Oemid 'AP' does not remind me of any name I've heard before either, I wonder what it means. In most cards, the ASCII value of the OEMID is whoever made the controller (SD - sandisk, SM - samsung, AD - adata, PA - panasonic, ...)
./flashbench /dev/sda3 --open-au-nr=5 -O --blocksize=2048 4MiB 8.69M/s 2MiB 5.31M/s 1MiB 11M/s 512KiB 11M/s 256KiB 10.8M/s 128KiB 10.5M/s 64KiB 10.4M/s 32KiB 9.19M/s 16KiB 7.25M/s 8KiB 4.43M/s 4KiB 2.26M/s 2KiB 1.16M/s
I would guess that the page size for this is 8KB based on these numbers.
./flashbench /dev/sda3 --open-au-nr=6 -O --blocksize=2048 4MiB 8.78M/s 2MiB 3.95M/s 1MiB 1.71M/s 512KiB 969K/s 256KiB 487K/s
-> deals well with 5 blocks
Ok.
./flashbench /dev/sda3 --open-au-nr=5 -O --blocksize=2048 --random 4MiB 8.09M/s 2MiB 8.24M/s 1MiB 10.1M/s 512KiB 9.9M/s 256KiB 9.34M/s 128KiB 8.49M/s 64KiB 9.02M/s 32KiB 7.27M/s 16KiB 5.2M/s 8KiB 3.02M/s 4KiB 1.19M/s 2KiB 582K/s
./flashbench /dev/sda3 --open-au-nr=5 -O --blocksize=2048 --random 4MiB 10.2M/s 2MiB 9.95M/s 1MiB 10.1M/s 512KiB 9.89M/s 256KiB 9.7M/s 128KiB 9.3M/s 64KiB 9.07M/s 32KiB 7.31M/s 16KiB 5.21M/s 8KiB 3.02M/s 4KiB 1.2M/s 2KiB 584K/s
-> looks quite fair enough also on random access
Yes, very good in fact.
./flashbench /dev/sda3 --open-au-nr=5 -O --blocksize=2048 --random --offset=$[123 * 512] 4MiB 4.49M/s 2MiB 2.28M/s 1MiB 2.27M/s 512KiB 2.23M/s 256KiB 2.19M/s 128KiB 2.13M/s 64KiB 2.1M/s 32KiB 1.34M/s 16KiB 756K/s 8KiB 398K/s 4KiB 206K/s
-> unaligned quite slow
The 123*512 is extremely unaligned. In general you can also try half the erasesize as an offset here, i.e. '--offset=$[2048 * 1024]'. If the erasesize is correct, it should be slower.
Arnd