grendel25
Legendary
Offline
Activity: 2296
Merit: 1031
|
|
January 08, 2015, 06:24:11 AM |
|
Submitted the optimized blake kernals to gihub.(25% faster)
I had to split the kernals and implement 2 seperate implementations. one for the x series. x11,x13,x14,x15,x17 and one for the others, quark, nist5, jackpotcoin etc. In the x'series implementation I moved some precalculation from const mem into the instructioncache. I also improved the uint2 implementation.
Building release 26
results on 750ti: (up from release 25)
x11 +30-50 KHASH quark +50-100 KHASH
970 and 980 untested. Please test if you have time
I think after release 23 or 24 I reverted to something like 19 or 21 (it's the one dated: 12/21/2014). I like this one because I consistently get over 3 Mh/s in X11 and it doesn't have much noticeable affects on regular usage of the 3 PCs it's running on. So the reason I'm saying this... How is the latest as far as computer usability? I know there's probably more support for rig based miners as opposed to average joe blows like me but just figured I'd ask for some feedback in general from anyone who might have similar interests like me.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 08, 2015, 08:16:14 AM |
|
Klaus_t has been busy@github, I think I will re-merge his fork with mine.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 08, 2015, 08:20:18 AM |
|
I think after release 23 or 24 I reverted to something like 19 or 21 (it's the one dated: 12/21/2014). I like this one because I consistently get over 3 Mh/s in X11 and it doesn't have much noticeable affects on regular usage of the 3 PCs it's running on. So the reason I'm saying this... How is the latest as far as computer usability? I know there's probably more support for rig based miners as opposed to average joe blows like me but just figured I'd ask for some feedback in general from anyone who might have similar interests like me.
The latest versions uses more GPU and CPU in x11. It's just to get a few extra khash. Not really needed. The 21 build is ok, but it will report a lower hashrate on some pools. (coinmine.pl, suchpol.pw etc)
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 08, 2015, 08:23:56 AM |
|
Looks like you pushed it too hard yes. Use a tool like GPUZ to messure the temperature. Open the case if the card is too hot. wich card model, and wich algorithm are you mining?
its a Asus Strix OC 4gb GTX 980 and mining quark Pretty good hash for quark! [2015-01-07 14:42:37] accepted: 200/202 (99.01%), 16938 khash/s yay!!! This is around the same speed as 6 750TI's mining x11. (but with one 980 card)
|
|
|
|
Handofmidas
Member
Offline
Activity: 61
Merit: 10
|
|
January 08, 2015, 09:31:10 AM |
|
Release26 seems to be a tiny bit slower than 25 for me (GTX970, X11).
|
|
|
|
tbearhere
Legendary
Offline
Activity: 3178
Merit: 1003
|
|
January 08, 2015, 11:09:30 AM |
|
Looks like you pushed it too hard yes. Use a tool like GPUZ to messure the temperature. Open the case if the card is too hot. wich card model, and wich algorithm are you mining?
its a Asus Strix OC 4gb GTX 980 and mining quark Pretty good hash for quark! [2015-01-07 14:42:37] accepted: 200/202 (99.01%), 16938 khash/s yay!!! This is around the same speed as 6 750TI's mining x11. (but with one 980 card) Make that 3 750ti sp 6mh/s 750ti. 3 750ti 6mh x 3 = 18mh 18,000kh/s
|
|
|
|
kaltar
|
|
January 08, 2015, 03:15:45 PM |
|
Looks like you pushed it too hard yes. Use a tool like GPUZ to messure the temperature. Open the case if the card is too hot.
You where right, i reduced the core by 5% and now it`s been running for 18hrs non stop, thank for the hint and quick reply
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 08, 2015, 07:37:52 PM Last edit: January 08, 2015, 07:55:16 PM by sp_ |
|
Release26 seems to be a tiny bit slower than 25 for me (GTX970, X11).
I tested on the 980 and the 970. The launch configuration was way off. the new blake code needed different launchparameters. I modified a bit, and booom 150KHASH more (x11) will test some more and submit to github. The standard clocked 980 is now hashing at 9MHASH.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 08, 2015, 09:13:07 PM Last edit: January 08, 2015, 09:33:45 PM by sp_ |
|
fixed x11 on 970 and 980 bether performance on 750ti but will drop if the cards get too hot. new beta: The 27th public beta(windows executable) is available here: (08-jan-2015) http://www.filedropper.com/release27The sourcecode is available here: https://github.com/sp-hash/ccminerPost your stats here. Card name/gpu clock/memclock
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 08, 2015, 09:59:00 PM |
|
spreadcoin is using sha256 (bitcoin) hashing in addition to other algorithms. Don't throw your old btc Asic miners(usb?)... Perhaps they can be used in a hybrid miner? NVIDIA + ASICS +SP-HASH=$$$$ I heard they are selling useless miners at ebay
|
|
|
|
Bombadil
|
|
January 08, 2015, 10:06:05 PM |
|
spreadcoin is using sha256 (bitcoin) hashing in addition to other algorithms. Don't throw your old btc Asic miners(usb?)... Perhaps they can be used in a hybrid miner? NVIDIA + ASICS +SP-HASH=$$$$ I heard they are selling useless miners at ebay How much % of the total algo is that SHA part? If it's 50% of the time, it might be worth it (cuda </3 sha256) Even with just a simple 2GH USB miner. Need one? I have 10x antminer U2 ^^ Also, get on IRC, sp_, and idle there
|
|
|
|
neonman
Newbie
Offline
Activity: 3
Merit: 0
|
|
January 08, 2015, 10:25:45 PM |
|
does not compile on linux with gcc 4.8 from latest git pull: cuda_myriadgroestl.o:(.bss+0x40): multiple definition of `d_resultNonce' Algo256/cuda_fugue256.o:(.bss+0x40): first defined here collect2: error: ld returned 1 exit status
also error on git clone from KlausT fork
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 08, 2015, 10:30:50 PM Last edit: January 09, 2015, 12:55:00 AM by sp_ |
|
does not compile on linux with gcc 4.8 from latest git pull: cuda_myriadgroestl.o:(.bss+0x40): multiple definition of `d_resultNonce' Algo256/cuda_fugue256.o:(.bss+0x40): first defined here collect2: error: ld returned 1 exit status
also error on git clone from KlausT fork working on it. Merging the fork. But it wil take a while. The hashrate has dropped. But the PTX assembly is bether. (less instructions) release 27 is after this commit: 4de3436ba38702feda3c625c43cdc07530e82bb8 but before the merge
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 08, 2015, 11:02:59 PM Last edit: January 09, 2015, 11:35:54 AM by sp_ |
|
I fixed the pool bug some days ago in release 25. CCminer didn't submit all the solutions to the pool. Only for some some pools though. The bug introduced in the tvpruvot release 1.5.0. and fixed in 1.5.24(sp-hash)
release<9 and release >23 is works on Coinmine.pl,suchpool.pw, Coinotron.com etc.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 09, 2015, 12:09:02 AM |
|
I think I need to rollback the klausT merge, and merge 1 kernal at a time. The hashrate has dropped, quark is broken @ head..
|
|
|
|
Epsylon3
Legendary
Offline
Activity: 1484
Merit: 1082
ccminer/cpuminer developer
|
|
January 09, 2015, 03:50:09 PM |
|
I fixed the pool bug some days ago in release 25. CCminer didn't submit all the solutions to the pool. Only for some some pools though. The bug introduced in the tvpruvot release 1.5.0. and fixed in 1.5.24(sp-hash)
release<9 and release >23 is works on Coinmine.pl,suchpool.pw, Coinotron.com etc.
its not a bug but a work around for buggy pools
|
|
|
|
scryptr
Legendary
Offline
Activity: 1797
Merit: 1028
|
|
January 09, 2015, 06:06:03 PM |
|
SP_ v26- This build is running smoothly on both of my rigs for more than a day. Mining quark, my slower rig gets 34,710kh/s, my faster rig gets 35,930kh/s. Both are nearing 3,000 accepts at less than 1% rejects. I'd try v27 but the author is negative about it in his posts. I think that the git code may be less-than-stable from his remarks. Excuse me, does quark consume more or less energy than x11? Another note, I mined x14 for about an hour yesterday on both rigs. Individual cards clocked at 2350kh/s-2450kh/s, but for a brief while x14 was worth it. --scryptr
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
|
January 09, 2015, 06:36:19 PM |
|
I have fixed it. Working on more hash . Will submit more changes to the blake algo today.
|
|
|
|
flipclip
Member
Offline
Activity: 111
Merit: 10
|
|
January 09, 2015, 06:39:03 PM |
|
SP_ v26- This build is running smoothly on both of my rigs for more than a day. Mining quark, my slower rig gets 34,710kh/s, my faster rig gets 35,930kh/s. Both are nearing 3,000 accepts at less than 1% rejects. I'd try v27 but the author is negative about it in his posts. I think that the git code may be less-than-stable from his remarks. Excuse me, does quark consume more or less energy than x11? Another note, I mined x14 for about an hour yesterday on both rigs. Individual cards clocked at 2350kh/s-2450kh/s, but for a brief while x14 was worth it. --scryptr I believe quark consumes roughly (less then a watt difference) the same as x11. I haven't noticed any power fluctuations from my kill-a-watt from recent version updates. x14 can be profitable but it is pretty short lived (as you noticed).
|
|
|
|
flipclip
Member
Offline
Activity: 111
Merit: 10
|
|
January 09, 2015, 06:51:19 PM |
|
does not compile on linux with gcc 4.8 from latest git pull: cuda_myriadgroestl.o:(.bss+0x40): multiple definition of `d_resultNonce' Algo256/cuda_fugue256.o:(.bss+0x40): first defined here collect2: error: ld returned 1 exit status
also error on git clone from KlausT fork I submitted a patch (which allows it to compile. you'll see the pull request in sp_'s git), but it may be unnecessary with the work sp_ is undertaking now.
|
|
|
|
|