$ fdisk -l /dev/sdc

Disk /dev/sdc: 32.5 GB, 32462864384 bytes
64 Köpfe, 32 Sektoren/Spur, 30959 Zylinder, zusammen 63404032 Sektoren
Einheiten = Sektoren von 1 × 512 = 512 Bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x000c0bcb


$ ./flashbench -a /dev/sdc --blocksize=1024
align 8589934592        pre 1.04ms      on 1.59ms       post 1.32ms     diff 402µs
align 4294967296        pre 1.04ms      on 1.67ms       post 1.33ms     diff 485µs
align 2147483648        pre 1.03ms      on 1.64ms       post 1.31ms     diff 468µs
align 1073741824        pre 1.02ms      on 1.48ms       post 1.33ms     diff 309µs
align 536870912		pre 1.02ms      on 1.5ms        post 1.35ms     diff 320µs
align 268435456 	pre 1.02ms      on 1.49ms       post 1.28ms     diff 338µs
align 134217728 	pre 1.02ms      on 1.49ms       post 1.34ms     diff 314µs
align 67108864  	pre 1.04ms      on 1.48ms       post 1.27ms     diff 332µs
align 33554432  	pre 1.01ms      on 1.5ms        post 1.32ms     diff 330µs
align 16777216  	pre 1.04ms      on 1.5ms        post 1.28ms     diff 347µs
align 8388608   	pre 995µs       on 1.51ms       post 1.31ms     diff 354µs
align 4194304   	pre 1.02ms      on 1.35ms       post 1.3ms      diff 189µs
align 2097152  		pre 1.05ms      on 1.39ms       post 1.31ms     diff 209µs
align 1048576   	pre 1.04ms      on 1.37ms       post 1.29ms     diff 203µs
align 524288    	pre 1.02ms      on 1.36ms       post 1.32ms     diff 192µs
align 262144    	pre 1.01ms      on 1.33ms       post 1.31ms     diff 176µs
align 131072    	pre 1.03ms      on 1.35ms       post 1.31ms     diff 182µs
align 65536     	pre 1.01ms      on 1.33ms       post 1.26ms     diff 195µs
align 32768     	pre 1.04ms      on 1.37ms       post 1.29ms     diff 203µs
align 16384     	pre 1.04ms      on 1.36ms       post 1.28ms     diff 204µs
align 8192      	pre 1.02ms      on 1.3ms        post 1.27ms     diff 154µs
align 4096      	pre 1.17ms      on 1.16ms       post 1.11ms     diff 22.6µs
align 2048      	pre 1.25ms      on 1.24ms       post 1.2ms      diff 20.2µs

For me it looks like 8MB erase block size and a 8KB page size. With that in mind, I made the following tests:

$ ./flashbench /dev/sdc --open-au --erasesize=$[8*1024*1024] --blocksize=$[8*1024] --open-au-nr=3 --random
8MiB    17M/s
4MiB    17.2M/s
2MiB    16.5M/s
1MiB    15.9M/s
512KiB  15M/s
256KiB  13.1M/s
128KiB  10.7M/s
64KiB   10.9M/s
32KiB   7.84M/s
16KiB   4.87M/s
8KiB    1.79M/s

darksun:/home/holger/flashbench # ./flashbench /dev/sdc --open-au --erasesize=$[8*1024*1024] --blocksize=$[8*1024] --open-au-nr=4 --random
8MiB	17.4M/s
4MiB	17.1M/s
2MiB16.8M/s
1MiB    16M/s
512KiB  15M/s
256KiB  13.4M/s
128KiB  10.8M/s
64KiB   10.9M/s
32KiB   7.89M/s
16KiB   4.87M/s
8KiB    1.79M/s


$ ./flashbench /dev/sdc -f --erasesize=$[8*1024*1024]
8MiB    17.2M/s  17.8M/s  17.1M/s  17.8M/s  17.5M/s  17.5M/s
4MiB    16.9M/s  17.7M/s  17M/s    17.8M/s  17.5M/s  17.5M/s
2MiB    16.8M/s  17.6M/s  17M/s    17.6M/s  17.5M/s  17.4M/s
1MiB    16.7M/s  17.6M/s  17M/s    17.5M/s  17.4M/s  17.5M/s
512KiB  16.6M/s  17.2M/s  17.4M/s  17.1M/s  16.9M/s  16.5M/s
256KiB  16.4M/s  17M/s    17.2M/s  16.9M/s  16.6M/s  16.6M/s
128KiB  16.5M/s  16.2M/s  16.6M/s  16.3M/s  16.1M/s  16.1M/s
64KiB   16.5M/s  17.1M/s  16.4M/s  17.2M/s  17M/s    17M/s
32KiB   15.4M/s  16.3M/s  15.5M/s  16.1M/s  16M/s    16.1M/s
16KiB   13M/s    13.5M/s  13.2M/s  13.5M/s  13.5M/s  13.5M/s