I've seen serveral numbers, in this thread, for the --icarus-timing parameter, but most of them seem wrong when used with twin_test.bit.
Cgminer sees each hashing FPGA as a full icarus board; a real icarus board splits the nonce range between the two FPGAs it has onboard, so, since in our case only one FPGA in each pair is active, our FPGAs are doing just half of the nonce range, the other half is thrown away (or sent to the disabled FPGA, I don't know).
Given that twin_test.bit hashes at 190MH/s every hash takes 1s/190MH or 5.2632 ns. so this is the first number.
How long does it take to scan the "full" nonce range? 4GH / 2 (because half nonce range is not hashed) / 190MH/s or 11.3 seconds.
So, the second number should be 112, I've started mine with
Code:
--icarus-timining 5.2632=111
I've used 111 to be on the safe side.
Sice I've started it this way the yellow led is always off, so FPGA utilization is full, but now the board speed is around 210-230 MH/s on cgminer with my pool reporting an hashing speed of 185 MH/s.
spiccioli
