RussK1
Splendid
Yes Russ, I started configuring the 680 for testing last night while I wait for my parts to come this week. I def could use some help configuring and tweaking. Perhaps you can assist me in dialing in some settings. Well I'm off to work, can't wait to play with it later.
The following will debug best launch configuration (be patient). Once you know what it is you can manually write it into the config..
cudaminer.exe -D -H 1 -d 0 -i 0 -l auto -C 0 -m 1 -o stratum+tcp://whatever pool ur using.com:3334 -O user😛ass
This may work:
cudaminer.exe -H 1 -d 0 -i 0 -l K8x32 -C 0 -m 1 -o stratum+tcp://whatever pool ur using.com😛ort -O user😛ass
Benchmark without the latency of the network or pool.
cudaminer.exe -H 1 -d 0 -i 0 -l K8x32 (or auto) -C 0 -m 1 --benchmark
Quoted from the CGminer README:
Additional cudaminer specific command line options are:
--no-autotune disables the built-in autotuning feature for
maximizing CUDA kernel efficiency and uses some
heuristical guesswork, which might not be optimal.
--devices [-d] gives a comma separated list of CUDA device IDs
to operate on. Device IDs start counting from 0!
Alternatively give string names of your card like
gtx780ti or gt640#2 (matching 2nd gt640 in the PC).
--launch-config [-l] specify the kernel launch configuration per device.
This replaces autotune or heuristic selection. You can
pass the strings "auto" or just a kernel prefix like
F or K or T to autotune for a specific card generation
or a kernel prefix plus a lauch configuration like F28x8
if you know what kernel runs best (from a previous
autotune).
--interactive [-i] list of flags (0 or 1) to enable interactive
desktop performance on individual cards. Use this
to remove lag at the cost of some hashing performance.
Do not use large launch configs for devices that shall
run in interactive mode - it's best to use autotune!
--batchsize [-b] comma separated list of max. scrypt iterations that
are run in one kernel invocation. Default is 1024. Best to
use powers of 2 here. Increase for better performance in
scrypt-jane with high N-factors. Lower for more interactivity
of your video display especially when using the interactive
mode.
--texture-cache [-C] list of flags (0 or 1 or 2) to enable use of the
texture cache for reading from the scrypt scratchpad.
1 uses a 1D cache, whereas 2 uses a 2D texture layout.
Cached operation has proven to be slightly faster than
noncached operation on most GPUs.
--single-memory [-m] list of flags (0 or 1) to make the devices
allocate their scrypt scratchpad in a single,
consecutive memory block. On Windows Vista, 7/8
this may lead to a smaller memory size being used.
When using the texture cache this option is implied.
--hash-parallel [-H] scrypt also has a small SHA256 or Keccak component:
0 hashes this single threaded on the CPU.
1 to enable multithreaded hashing on the CPU.
2 offloads everything to the GPU (default)