hello, want to confirm, 4 independend tasks with two r9 390 works great and gives me 80-85 sol./sec. (build about 4h ago)
1 Like
Why do we desperately need hardcoded multi-GPU support? Is there something wrong with running either 2 paraell processes - 1 on each GPU - or actually more than 1 process per GPU? I have found that running multiple processes on 1 GPU gives a little bit better result (+15 percent).
1 Like
which branch will work with n1070, pls?
I remember running into this a long time ago; GCC invokes ld through LIBRARY_PATH during compilation (note not LD_LIBRARY_PATH). So something probably needs to be tweaked in the makefile. Either figure out the proper export for LIBRARY_PATH=/usr/lib or go the quick and easy way by softlinking the libOpenCL.so.1 to libOpenCL.so:
ln -s /usr/lib/libOpenCL.so.1 /usr/lib/libOpenCL.so
Should work.
1 Like
Argh, just noticed problem was already answered earlier in the thread. Same fix as what I said but go here instead for definitive answer: cannot find -lOpenCL · Issue #18 · nginnever/zogminer · GitHub
2 Likes
I just pushed the revision to master that allowsa for n threads to process the solver on ONE gpu. Use genproclimit = 2 to get two silentarmy kernels running parallel. Be warned you may get some nasty errors if you put anything higher than 2 in your config. We are working on opencl memory detection to not allow this for protection in the future… thanks hellcatz!
Age Manning is finishing up multi-gpu support while I get ready for the first night on the town in awhile to celebrate.
Age hit 48H/s today on a 290.
Happy hashing!
2 Likes
Great Job Bro! I found zogminer is totally worth the waits
~41.3 from one 470 8gb with genproclimit=2
but sometime:
Invalid solution: invalid collision length between StepRows
and
Segmentation fault (core dumped)
Please update fetch-param.sh. its fetching beta-proving keys instead of sprout. Thanks
update:
My bad. Did not look at the other repo. Thanks
Similar results for me it’s running steadily for a while now but a lot of Invalid solution: invalid collision length between StepRows and I’m getting ~35H/s with genproclimit=2 on single RX480 8GB
Hey guys…
I got multi gpu solo mining working. Still testing catalyst drivers at the moment.
To use it, clone this repo: GitHub - nginnever/zcash: zcash
checkout the multi-solo branch and build as usual.
You will need to get the mainnet chain first, .ie run zcashd -gen=0
Once updated, you can run the following
zcashd -G -allgpu (this will load all your cards and solo mine)
zcashd -G -deviceid= (this will run a specific card)
zcashd -G -allgpu -genproclimit=2 (this will run two threads per card. If you’re card has 4gb or more this can work)
I’ve still got to clean some of the output, but you can go ahead at test.
Cheers guys!
3 Likes
Uh, what is with all of the beta2 stuff still in there? I think these are the wrong params.
$ ./zcutil/fetch-params.sh
Zcash - fetch-params.sh
Retrieving: https://z.cash/downloads/beta2-proving.key
–2016-10-30 00:07:22-- https://z.cash/downloads/beta2-proving.key
Resolving z.cash (z.cash)… 104.236.171.172
Connecting to z.cash (z.cash)|104.236.171.172|:443… connected.
HTTP request sent, awaiting response… 301 Moved Permanently
Location: https://s3.amazonaws.com/zcashalpha/beta2-proving.key [following]
–2016-10-30 00:07:22-- https://s3.amazonaws.com/zcashalpha/beta2-proving.key
Resolving s3.amazonaws.com (s3.amazonaws.com)… 52.216.0.51
Connecting to s3.amazonaws.com (s3.amazonaws.com)|52.216.0.51|:443… connected.
HTTP request sent, awaiting response… 206 Partial Content
Length: 910173851 (868M), 606767124 (579M) remaining [application/pgp-keys]
Saving to: ‘/home/-name-/.zcash-params/beta2-proving.key.dl’
Is there no hope for older cards R9 280x for solo?
Unless you have a few hundred of them I highly doubt solo is the way to go.
Im disposing these cards but I still have some rigs full of it and its idle.
Edit:
I think “solardiz” is a judge on zcashminer challenge and he claimed to run zogminer on HD7790(zcash-miner-dev@zcashcommunity.slack.com) so it is still possible for 280’s. Such a waste if this wont be put to good use.
is that 16.04 or 14.04. amd-pro or flgrx driver?
1 Like
Thank you! Helped 
I am still trying to run this miner on nvidia cards but no luck. With GTX750Ti I get the following error:
Kernel run took 10 ms. (0 H/s)
[OPENCL]:CL ERROR:CL_MEM_OBJECT_ALLOCATION_FAILURE
Any suggestion?
thanks in advance
If you’re using 2 threads, i.e -genproclimit=2 set it to 1.
Your card may not have enough memory ( <4GB).
Otherwise if you’re using catalyst drivers, try convert to amdgpu-pro drivers.