just a question ...
what is the best commandline parameters to use when mining with lyre2 ( and groestl - i just didnt want to post the same question in both threads ) under linux? ...
does it use similar settings to x11 and the like? ...
is there a git that we can compile directly? ...
tanx again ...
#crysx
I'm mostly a linux user, so the settings on the OP will work fine with the penguin. Same for groestl, just head over to the other thread linked in my signature. will do ... would you have any suggestions as to what coins and where would be a nice mine? ... hope to hear from you shortly ... as ill be offline for about 16 hours - so the compile / testing will be done tomorrow ... tanx ... #crysx I've been mining diamond for ages, if things haven't changed it's the only profitable groestl coin.
|
|
|
just a question ...
what is the best commandline parameters to use when mining with lyre2 ( and groestl - i just didnt want to post the same question in both threads ) under linux? ...
does it use similar settings to x11 and the like? ...
is there a git that we can compile directly? ...
tanx again ...
#crysx
I'm mostly a linux user, so the settings on the OP will work fine with the penguin. Same for groestl, just head over to the other thread linked in my signature.
|
|
|
Which of the two algorithms profit is greater than Quark or Lyra2RE?
With two 7970 do little more than 2 MH/s, but 280x can not in any way make them do more than 180-190kH/s. Where is wrong? This is with algorithm Lyra2RE.
My settings are the same for both types of video cards! Here it is in my file:
setx GPU_FORCE_64BIT_PTR 0 setx GPU_MAX_HEAP_SIZE 100 setx GPU_USE_SYNC_OBJECTS 1 setx GPU_MAX_ALLOC_PERCENT 100
sgminer.exe -k Lyra2RE -o stratum+tcp://lyra2re.eu.nicehash.com:3342 -u address -p p=1 --intensity 15 --worksize 128 -g 2 --gpu-reorder
What is your gpu memory type? Is that not hynex? Yes, the problem may be the memories. Memoryinfo says that 7970 have Hynix and 280x with Elpida. Perhaps this is the problem, because only 280x are not working. Thank you. I know not to agonize over. elpida surely will be slower, but not as much as you are reporting. there must be something else going wrong. you should get between 1050 and 1150 on tahiti cards.
|
|
|
2 DStefanov
Use optimized miner from pallas. It is free and open. Look in parallel thread. Don't forget thanks to the author ...
thanks see my signature for the link
|
|
|
vertcoin got pumped. time for pallas to ask for donation from amd-lovers ))) I'm on nvidia, so - sorry )))
there are donation hints all over the thread and in the miner itself... they just don't want to ;-)
|
|
|
hmm)) how to compile it on ubuntu?
Hix, did you find a fix to the the issue? I am having zero luck compiling "siad" on Ubuntu 14.04. https://github.com/NebulousLabs/Sia/blob/master/doc/Developers.mdSia's a bit trickier because we have many packages inside of the same project.You need to set up your $GOPATH, and then you need to clone the Sia repo in $GOPATH/src/github.com/NebulousLabs Alternatively, you can try installing it with go get github.com/NebulousLabs/Sia/... . In order to compile it, I had to set GOBIN as well (ubuntu).
|
|
|
no please, don't make me install go! :-) new wallet binaries?
I fetched a linux x64 binary compiled by some kind soul in the IRC; I tested the new miner with it. While Taek says it's a lot better on Nvidia (I haven't tested), it seems only somewhat faster on AMD. Kineta's 290X does something around 90MH/s - 100MH/s with it, although the code has been much improved - the AMD OpenCL compiler is butchering it with its lobotomized code generation. The protocol is nicer, though - no more fetching the whole block, just the header and target. Whole header is returned, though, not just the nonce. Anyways, I've ported mine, and I'm waiting for Kineta to find a block (and siad to verify said block) with the two non-display GPUs - totaling 1.855GH/s on a 290X and 285, but I'm being quite gentle with them, as Kineta is my main system, so definitely don't want to crash the driver. yes, for me it's as slow as before. I think there are way too many private variables and some end up in global ram. but if you are going to release your version, I'll not waste my time fixing it. EDIT: my cl file works as is with the new miner from git: Mining at 755.989 MH/s 0 blocks mined 280x @ 1100/950 Waiting for a block to come She mined a block. My CL doesn't, so I copied network.{c,h}, then made the necessary edits to my version to call the correct functions with the correct params - it was really only a few-line edit to my existing code to use the new networking portion. I also link with the C reference implementation of Blake2B, and use its headers for my regenhash-style HW error check. With that miner, a HW error will simply appear to be a reject... well, actually, a hash below target reject. great! but... will you release it? I just want to understand what's going on with this miner. And if my code is worth merging, keeping, selling.... whatever.
|
|
|
using bamt 1.6.2 with sgminer-5.1.1-lyra2re on bamt system apt-get install libcurl4-openssl-dev pkg-config libtool libncurses5-dev apt-get update cd /opt/miners git clone https://github.com/sgminer-dev/sgminer.gitcd sgminer cp /opt/AMDADL/*h ./ADL_SDK/ git submodule init git submodule update autoreconf -i CFLAGS="-O2 -Wall -march=native -std=gnu99" ./configure make cd .. mv sgminer sgminer-5.1.1-lyra2re cd /opt/bamt vi common.pl modify the sgminer-sph section to sgminer-lyra2re using the path /opt/miners/sgminer-5.1.1-lyra2re cd /etc/bamt vi bamt.conf modify the cgminer section by adding a line with sgminer-lyra2re.conf and miner-sgminer-lyra2re: 1 copy sgminer-sph.conf to sgminer-lyra2re.conf and fit the values for your graphic card i am currently running a sapphire 5850 xtreme with i=18, clock=775, mem=1000, volt=1.000 with about 350kh/s @ 170Watt @ 68°C good luck ! greetings pazor thanks for your guide ;-)
|
|
|
no please, don't make me install go! :-) new wallet binaries?
I fetched a linux x64 binary compiled by some kind soul in the IRC; I tested the new miner with it. While Taek says it's a lot better on Nvidia (I haven't tested), it seems only somewhat faster on AMD. Kineta's 290X does something around 90MH/s - 100MH/s with it, although the code has been much improved - the AMD OpenCL compiler is butchering it with its lobotomized code generation. The protocol is nicer, though - no more fetching the whole block, just the header and target. Whole header is returned, though, not just the nonce. Anyways, I've ported mine, and I'm waiting for Kineta to find a block (and siad to verify said block) with the two non-display GPUs - totaling 1.855GH/s on a 290X and 285, but I'm being quite gentle with them, as Kineta is my main system, so definitely don't want to crash the driver. yes, for me it's as slow as before. I think there are way too many private variables and some end up in global ram. but if you are going to release your version, I'll not waste my time fixing it. EDIT: my cl file works as is with the new miner from git: Mining at 755.989 MH/s 0 blocks mined 280x @ 1100/950 Waiting for a block to come
|
|
|
no please, don't make me install go! :-) new wallet binaries?
|
|
|
So there are at least 2 people claim they are mining at x10 times faster than the rest of us... Wanna sell us your config ? not config changes, miner rewrite ;-) EDIT: a quick run on 290x shows more than 1Gh/s
|
|
|
I have downclocked the gpu by 30% but that doesn't seem to affect the mining speed at all.. save your electricity! yes because it is limited by the ram speed. you may even get more hashes by increasing memclock. but the algo is very simple and should run on registers only. that's how you get 280x do this: Mining at 756.267 MH/s X blocks mined Of course at this speed it doesn't "save electricity" ;-) Did you make the block fetch async? If not, is that including the block fetch time, as the stock miner does? I have not yet, and I kept the block fetch time included in the hashrate counter meter so that it better relates to the stock one. I rebased my work on the new commits and left that part unchanged. Some of the changes I did to the C code are similar to what they modified so it's not worth keeping them. EDIT: when I removed block fetching from the hashrate calculation to the old tree, the hashrate changed by a couple percent only
|
|
|
I have downclocked the gpu by 30% but that doesn't seem to affect the mining speed at all.. save your electricity! yes because it is limited by the ram speed. you may even get more hashes by increasing memclock. but the algo is very simple and should run on registers only. that's how you get 280x do this: Mining at 756.267 MH/s X blocks mined Of course at this speed it doesn't "save electricity" ;-)
|
|
|
Guys this is so much fun! My 280x is hashing at 670 Mh/s, I think I can go past 1 GH/s :-)
Not even whirlpoolx was so rewarding!
Oh, come on, yeah it's fun - but don't be too proud of it - it's not like it was brain surgery :3 you're damn right: it was pretty easy ;-) Oh, by the way, 280X around 800MH/s, and this is still including the block fetch; I haven't modified this to be done async yet. I'm about there as well: removed ram usage, can get to 800 with a bit of core overclock. Nice to know I'm on the right track!
|
|
|
Guys this is so much fun! My 280x is hashing at 670 Mh/s, I think I can go past 1 GH/s :-)
Not even whirlpoolx was so rewarding!
Oh, come on, yeah it's fun - but don't be too proud of it - it's not like it was brain surgery :3 you're damn right: it was pretty easy ;-)
|
|
|
Guys this is so much fun! My 280x is hashing at 670 Mh/s, I think I can go past 1 GH/s :-)
Not even whirlpoolx was so rewarding!
Would really appreciate it if you could share your settings. I am hashing with only about 100-112 Mh/s with my 280x Modified miner and mostly rewritten opencl code. Still a work in progress. Power use is almost doubled. can you share your .cl file? Also the C sources are modified, both for speed and stability.
|
|
|
Guys this is so much fun! My 280x is hashing at 670 Mh/s, I think I can go past 1 GH/s :-)
Not even whirlpoolx was so rewarding!
Would really appreciate it if you could share your settings. I am hashing with only about 100-112 Mh/s with my 280x Modified miner and mostly rewritten opencl code. Still a work in progress. Power use is almost doubled.
|
|
|
Guys this is so much fun! My 280x is hashing at 670 Mh/s, I think I can go past 1 GH/s :-)
Not even whirlpoolx was so rewarding!
|
|
|
it looks like diff is rising 'cause I can't find blocks for many hours now. besides, how to get the current difficulty? "siac status" returns a list of numbers I can't understand.
bump!
|
|
|
I'm getting half that on linux miner. A bit more by fiddling with old git commits. Please cleanup that souces!
We are not GPU programmers. We threw together the GPU miner at the last minute because a third party had developed FPGA mining for siacoin to the tune of 500 CPUs. Without an emergency gpu client he would have been able to 51% the network. We're sorry that not everyone was able to use the gpu miner at launch. If you have cleanup or fixes, please pull request them. We are trying to fix things but GPU code is outside of our wheelbase. About hashrate: I think the windows precompiled miner looks faster but it's actually wrong hashrate calculation: it's using an older version of the sources which used an inaccurate clock source. About pulling my changes to git: I can do it if you wish, or I can keep a separate testing/optimisation tree.
|
|
|
|