Schleicher
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 19, 2014, 05:07:14 PM |
|
I managed to increase the quark and nist5 speed a little bit. Source code: https://github.com/KlausT/ccminerCompared to djm's m7v7 version: about 6% faster for quark about 4% faster for nist5 (on my 750Ti, not overclocked)
|
|
|
|
AizenSou
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 08:22:08 AM |
|
I managed to increase the quark and nist5 speed a little bit. Source code: https://github.com/KlausT/ccminerCompared to djm's m7v7 version: about 6% faster for quark about 4% faster for nist5 (on my 750Ti, not overclocked) Nice. So many people have interest in cuda mining now. Thanks for this release, although I will stay with djm's version for a while.
|
|
|
|
PVmining
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 08:48:52 AM |
|
Big fat thanks! Love to see you share your optimization with us. Nice. So many people have interest in cuda mining now. Thanks for this release, although I will stay with djm's version for a while.
I like to mention tpruvots source once more. https://github.com/tpruvot/ccminerclean & mostly up to date. especially for windows guys easy to compile, because he included everything in source that is needed.
|
|
|
|
sp_
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 08:58:30 AM |
|
I managed to increase the quark and nist5 speed a little bit. Source code: https://github.com/KlausT/ccminerCompared to djm's m7v7 version: about 6% faster for quark about 4% faster for nist5 (on my 750Ti, not overclocked) Good job. These improvements are different from my maxwell mod, so it meens that it will improve the speed of the maxwell mod when I merge them in.
|
|
|
|
restless
Legendary
Offline
Activity: 1151
Merit: 1001
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 10:10:28 AM |
|
I managed to increase the quark and nist5 speed a little bit. Source code: https://github.com/KlausT/ccminerCompared to djm's m7v7 version: about 6% faster for quark about 4% faster for nist5 (on my 750Ti, not overclocked) Which VS version for compiling? VS2013 says it cant compile native x86 :S
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 10:55:05 AM Last edit: October 20, 2014, 11:16:28 AM by djm34 |
|
Big fat thanks! Love to see you share your optimization with us. Nice. So many people have interest in cuda mining now. Thanks for this release, although I will stay with djm's version for a while.
I like to mention tpruvots source once more. https://github.com/tpruvot/ccminerclean & mostly up to date. especially for windows guys easy to compile, because he included everything in source that is needed. what did I (or others) missed in the source which make my version difficult to compile ? ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) Anyway some of the change will certainly make their way to my version when I have time (which I don't at the moment... the priority being on new development... and so far I haven't had much time to look back) I looked a bit into schleicher release and there is some interesting stuff. Anyhow I am almost done with sk-1024 miner. current speed 40MH/s on gtx980 26MH/s on gtx780ti 14MH/s on gtx750ti
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
PVmining
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 11:14:17 AM |
|
what did I (or others) missed in the source which make my version difficult to compile ? ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) I compiled your source on my linux boxes many times without hazzel. but on linux I had the usual "suspects" (dependencies) already installed for the common uses. trying to play around on windows (with vs2013) tprvouts source was the only one just compiling on a fresh system and just needs cuda installed. with your and other releases are some stuff missing (phtread,. curl,...)... and some of these stuff is not as easy to get by as on linux. Anyway some of the change will certainly make their way to my version when I have time (which I don't at the moment... the priority being on new development... ). Anyhow I am almost done with sk-1024 miner. current speed 40MH/s on gtx980 26MH/s on gtx780ti 14MH/s on gtx750ti
great guy! do you open-source your keccak speedup too?
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 11:32:46 AM |
|
what did I (or others) missed in the source which make my version difficult to compile ? ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) I compiled your source on my linux boxes many times without hazzel. but on linux I had the usual "suspects" (dependencies) already installed for the common uses. trying to play around on windows (with vs2013) tprvouts source was the only one just compiling on a fresh system and just needs cuda installed. with your and other releases are some stuff missing (phtread,. curl,...)... and some of these stuff is not as easy to get by as on linux. Anyway some of the change will certainly make their way to my version when I have time (which I don't at the moment... the priority being on new development... ). Anyhow I am almost done with sk-1024 miner. current speed 40MH/s on gtx980 26MH/s on gtx780ti 14MH/s on gtx750ti
great guy! do you open-source your keccak speedup too? at some point yes (rather next week)... but not today as sk1024 will be released as a standalone miner (well nobody want pthreads and boost at the time...). Regarding how the dependencies are treated, I don't think everything should be including into ccminer. It isn't a good idea nor considered as good practice in terms of package management and make things difficult to upgrade hence at that level I prefer to stick with what has been done so far... (even though it is a pain in the ass...)
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 11:38:21 AM |
|
what did I (or others) missed in the source which make my version difficult to compile ? ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) I compiled your source on my linux boxes many times without hazzel. but on linux I had the usual "suspects" (dependencies) already installed for the common uses. trying to play around on windows (with vs2013) tprvouts source was the only one just compiling on a fresh system and just needs cuda installed. with your and other releases are some stuff missing (phtread,. curl,...)... and some of these stuff is not as easy to get by as on linux. Anyway some of the change will certainly make their way to my version when I have time (which I don't at the moment... the priority being on new development... ). Anyhow I am almost done with sk-1024 miner. current speed 40MH/s on gtx980 26MH/s on gtx780ti 14MH/s on gtx750ti
great guy! do you open-source your keccak speedup too? It IS open source - the original uint2 keccak wasn't done by djm34. yep actually it is in cudaminer... but I assume nobody want to piece things together (well the 2 meme-miners (sorry ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) but the battle of the ego is funny) might pick them faster than I include them in my release. actually it works also with skein...
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
Schleicher
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 01:17:06 PM |
|
I managed to increase the quark and nist5 speed a little bit. Source code: https://github.com/KlausT/ccminerCompared to djm's m7v7 version: about 6% faster for quark about 4% faster for nist5 (on my 750Ti, not overclocked) Which VS version for compiling? VS2013 says it cant compile native x86 :S I'm using VS Express 2013 for Desktop For the dependencies I mostly use Christian's files because I'm too lazy to compile them myself: https://mega.co.nz/#!cQQziBrb!HB1fEEcVM7HQuE3HI1ZcV00GCOA6kV4rwmAWkBaMI84(see first post in this thread) I only had to compile MPIR myself.
|
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 01:29:45 PM Last edit: October 20, 2014, 08:48:05 PM by cbuchner1 |
|
Great stuff going on here in this thread, but ChrisH and I will be sticking to developing prime number maths in CUDA for a while. Implementing a fast Montgomery multiplication / exponentiation is challenging. Almost there... so close ;-)
Christian
|
|
|
|
tbearhere
Legendary
Offline
Activity: 3178
Merit: 1003
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 10:03:30 PM |
|
Big fat thanks! Love to see you share your optimization with us. Nice. So many people have interest in cuda mining now. Thanks for this release, although I will stay with djm's version for a while.
I like to mention tpruvots source once more. https://github.com/tpruvot/ccminerclean & mostly up to date. especially for windows guys easy to compile, because he included everything in source that is needed. Including the errors. ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) I get errors when trying to compile. ![Undecided](https://bitcointalk.org/Smileys/default/undecided.gif)
|
|
|
|
tbearhere
Legendary
Offline
Activity: 3178
Merit: 1003
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 20, 2014, 11:24:22 PM |
|
Can Cudaminer do sha256 ?
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 21, 2014, 12:02:47 AM Last edit: October 21, 2014, 12:14:03 AM by djm34 |
|
Can Cudaminer do sha256 ?
![Grin](https://bitcointalk.org/Smileys/default/grin.gif) wanna mine against TH/s asics ? ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) yes it can, but it's not worth it (also I don't know if there is one or two loop of sha256... which doesn't change anything actually) I think it is doing stuff around 300MH/s (say some optimization could double it or quadruple it... it would still be 1GH/s against several TH/s machine (not farm) so unless it is a ninja launch and you are the only one aware it, you don't have a chance to get a block. And if you get one, it will orphan...
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
tbearhere
Legendary
Offline
Activity: 3178
Merit: 1003
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 21, 2014, 01:04:33 AM |
|
Can Cudaminer do sha256 ?
![Grin](https://bitcointalk.org/Smileys/default/grin.gif) wanna mine against TH/s asics ? ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) yes it can, but it's not worth it (also I don't know if there is one or two loop of sha256... which doesn't change anything actually) I think it is doing stuff around 300MH/s (say some optimization could double it or quadruple it... it would still be 1GH/s against several TH/s machine (not farm) so unless it is a ninja launch and you are the only one aware it, you don't have a chance to get a block. And if you get one, it will orphan... yep ![Cheesy](https://bitcointalk.org/Smileys/default/cheesy.gif) your right djm34. one 750ti would get about 2 coins a day. on sha256 coin. ![Cheesy](https://bitcointalk.org/Smileys/default/cheesy.gif)
|
|
|
|
sp_
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 21, 2014, 04:33:34 AM |
|
These files contain the heartbleed version of openssl. This meens that the miner computer can be infected.
|
|
|
|
Schleicher
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 21, 2014, 05:19:55 AM |
|
These files contain the heartbleed version of openssl. This meens that the miner computer can be infected. No, not really. I don't use encrypted connections for mining.
|
|
|
|
sp_
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 21, 2014, 05:58:01 AM |
|
Can Cudaminer do sha256 ?
![Grin](https://bitcointalk.org/Smileys/default/grin.gif) wanna mine against TH/s asics ? ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) yes it can, but it's not worth it (also I don't know if there is one or two loop of sha256... which doesn't change anything actually) I think it is doing stuff around 300MH/s (say some optimization could double it or quadruple it... it would still be 1GH/s against several TH/s machine (not farm) so unless it is a ninja launch and you are the only one aware it, you don't have a chance to get a block. And if you get one, it will orphan... yep ![Cheesy](https://bitcointalk.org/Smileys/default/cheesy.gif) your right djm34. one 750ti would get about 2 coins a day. on sha256 coin. ![Cheesy](https://bitcointalk.org/Smileys/default/cheesy.gif) But if the ti is used to predict the outcome of the final hash, you might not need to compute all the hashes. Then the TI beat the TH's asiscs. Predict -> verify (cpu)
|
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 21, 2014, 06:57:56 AM Last edit: October 21, 2014, 09:21:37 PM by cbuchner1 |
|
These files contain the heartbleed version of openssl. This meens that the miner computer can be infected. Nope it's a data leakage bug - read access to 64kb of a memory pool page of recently freed allocations. It carries no infection risk.
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 21, 2014, 08:57:31 AM |
|
Can Cudaminer do sha256 ?
![Grin](https://bitcointalk.org/Smileys/default/grin.gif) wanna mine against TH/s asics ? ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) yes it can, but it's not worth it (also I don't know if there is one or two loop of sha256... which doesn't change anything actually) I think it is doing stuff around 300MH/s (say some optimization could double it or quadruple it... it would still be 1GH/s against several TH/s machine (not farm) so unless it is a ninja launch and you are the only one aware it, you don't have a chance to get a block. And if you get one, it will orphan... yep ![Cheesy](https://bitcointalk.org/Smileys/default/cheesy.gif) your right djm34. one 750ti would get about 2 coins a day. on sha256 coin. ![Cheesy](https://bitcointalk.org/Smileys/default/cheesy.gif) But if the ti is used to predict the outcome of the final hash, you might not need to compute all the hashes. Then the TI beat the TH's asiscs. Predict -> verify (cpu) ?
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
|