Author

Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1182. (Read 2347601 times)

sr. member
Activity: 506
Merit: 252
Edit miner.h:
From (start line 488):
extern char* device_name[8];
extern short device_map[8];
extern long device_sm[8];
To:
extern char* device_name[16];
extern int device_map[16];
extern long device_sm[16];
Made the edit and just compiled on my Ubuntu 14.04 system.  Also submitted a pull to git.

Yeah, forgot to checkin one file. CCminer is now supporting 16 cards per rig. Smiley

nice...

I can't get more than 6 working in win 8.1 (the same mobo had 7 amd cards running)
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
Edit miner.h:
From (start line 488):
extern char* device_name[8];
extern short device_map[8];
extern long device_sm[8];
To:
extern char* device_name[16];
extern int device_map[16];
extern long device_sm[16];
Made the edit and just compiled on my Ubuntu 14.04 system.  Also submitted a pull to git.

Yeah, forgot to checkin one file. CCminer is now supporting 16 cards per rig. Smiley
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
From release 31 to the 1/20/2015 commit, for quark I'm seeing a drop of about 25 kH/s per card (750Ti), but also a large drop in CPU usage (from ~16% to ~1%).

Yes, The miner starts well at first, but then the hashrate is dropping over time. I reverted the sync code. Now quark is using 3-5% cpu, but the hash is more stable.(hopefully)
member
Activity: 111
Merit: 10
From release 31 to the 1/20/2015 commit, for quark I'm seeing a drop of 25 kH/s gain of 50 kH/s per card (750Ti), and a large drop in CPU usage (from ~16% to ~1%).  Great stuff.
legendary
Activity: 1797
Merit: 1028
Edit miner.h:

From (start line 488):
extern char* device_name[8];
extern short device_map[8];
extern long device_sm[8];

To:
extern char* device_name[16];
extern int device_map[16];
extern long device_sm[16];

Made the edit and just compiled on my Ubuntu 14.04 system.  Also submitted a pull to git.

v31b TYPO-

THAT IS THE FIX.  It compiled and has just started running.  Initial impression:  Loss of 500kh/s for the rig from v31a at its fastest.  If this version runs smoothly and remains stable, I don't care.  Stability is more important.  I am right on 75 accepts and 100%  error-free.      --scryptr

EDIT:

It ran for a couple hundred accepts, and then dropped to double-digit kh/s.  A couple cards would resurge to 6mh/s, and then drop immediately.  When is sent "^c", I got a "SIGINT received" response, but the program would not close.  I have seen this pattern before in previous builds.   

I am going to save the config files and important data, and move to 14.10, as the 146.35 drivers now support Utopic Unicorn.  I may have junk in my OS that is a factor.       --scryptr
legendary
Activity: 3164
Merit: 1003
member
Activity: 111
Merit: 10
Edit miner.h:

From (start line 488):
extern char* device_name[8];
extern short device_map[8];
extern long device_sm[8];

To:
extern char* device_name[16];
extern int device_map[16];
extern long device_sm[16];

Made the edit and just compiled on my Ubuntu 14.04 system.  Also submitted a pull to git.
legendary
Activity: 1797
Merit: 1028
I tried v31, my fast rig is again stable, my slow rig crashes after 100-200 accepts.  I lowered the intensity to 18 (and lost 1mh/s performance!) and it was the same thing... crash.   I run stock clocks, period.
Both rigs are headless.  I plugged into the slow rig and booted normally, and noticed some error messages about nvidia driver 346.35.  It was released on 17 JAN 15, yesterday.  I think that I will have to purge the drivers and then re-install.  For Linux, v346.35 is a long-term release, no longer a beta, as 346.22 was.  I may burn the whole OS and move to 14.10, which is now supported by 346.35.      --scryptr

Please try a new build from github, the cpu usage is gone, and I have added more cuda cleanupcode. power should be down abit. cpu usage on x11, x13 was 10-20% in release 31 but reduced to 0-1% in the github version

v31-git BUILD --

I just got home and was about to burn 14.10 when I read this.  I tried to compile the code, but I suspect a TYPO:

{QUOTE/}

mv -f .deps/ccminer-crc32.Tpo .deps/ccminer-crc32.Po
g++ -DHAVE_CONFIG_H -I.  -fopenmp  -pthread -fno-strict-aliasing -I./compat/jans                                                                                                                                                             son -I/usr/local/cuda/include -DUSE_WRAPNVML -DSCRYPT_KECCAK512 -DSCRYPT_CHACHA                                                                                                                                                              -DSCRYPT_CHOOSE_COMPILETIME   -O3 -march=native -D_REENTRANT -falign-functions=1                                                                                                                                                             6 -falign-jumps=16 -falign-labels=16 -MT ccminer-util.o -MD -MP -MF .deps/ccmine                                                                                                                                                             r-util.Tpo -c -o ccminer-util.o `test -f 'util.cpp' || echo './'`util.cpp
ccminer.cpp:217:22: error: conflicting declaration ‘char* device_name [16]’
 char * device_name[16];
                      ^
In file included from ccminer.cpp:44:0:
miner.h:488:14: error: ‘device_name’ has a previous declaration as ‘char* device                                                                                                                                                             _name [8]’
 extern char* device_name[8];
              ^
ccminer.cpp:218:18: error: conflicting declaration ‘int device_map [16]’
 int device_map[16] = { 0, 1, 2, 3, 4, 5, 6, 7,8,9,10,11,12,13,14,15 };
                  ^
In file included from ccminer.cpp:44:0:
miner.h:489:14: error: ‘device_map’ has a previous declaration as ‘short int dev                                                                                                                                                             ice_map [8]’
 extern short device_map[8];
              ^
ccminer.cpp:219:19: error: conflicting declaration ‘long int device_sm [16]’
 long  device_sm[16] = { 0 };
                   ^
In file included from ccminer.cpp:44:0:
miner.h:490:14: error: ‘device_sm’ has a previous declaration as ‘long int devic                                                                                                                                                             e_sm [8]’
 extern long  device_sm[8];
              ^
make[2]: *** [ccminer-ccminer.o] Error 1
make[2]: *** Waiting for unfinished jobs....
mv -f .deps/ccminer-hefty1.Tpo .deps/ccminer-hefty1.Po
mv -f .deps/ccminer-scrypt.Tpo .deps/ccminer-scrypt.Po
mv -f .deps/ccminer-util.Tpo .deps/ccminer-util.Po
make[2]: Leaving directory `/home/borqd/spfresh'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/borqd/spfresh'
make: *** [all] Error 2
{/QUOTE}

Sorry for the fuss.  I hope that you can fix it, I really want to try the executable.       --scryptr
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
I tried v31, my fast rig is again stable, my slow rig crashes after 100-200 accepts.  I lowered the intensity to 18 (and lost 1mh/s performance!) and it was the same thing... crash.   I run stock clocks, period.
Both rigs are headless.  I plugged into the slow rig and booted normally, and noticed some error messages about nvidia driver 346.35.  It was released on 17 JAN 15, yesterday.  I think that I will have to purge the drivers and then re-install.  For Linux, v346.35 is a long-term release, no longer a beta, as 346.22 was.  I may burn the whole OS and move to 14.10, which is now supported by 346.35.      --scryptr

Please try a new build from github, the cpu usage is gone, and I have added more cuda cleanupcode. power should be down abit. cpu usage on x11, x13 was 10-20% in release 31 but reduced to 0-1% in the github version
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
Submitted a fix on the cpu usage. x11 is now using 0-1% cpu on a slow celeron. Quark is up another 50KHASH 750ti. !


legendary
Activity: 1400
Merit: 1000
@ sp_

Is qubit algo broken after release 24?

I get an error message "stratum extranonce id was not correct" and all "booo's" 
sr. member
Activity: 401
Merit: 250
I have solved a bug wich causes the cpu to use alot of cpu. Now x11 is hashing with 0-1% cpu usage. Fixed in the unpublished spreadcoin fork for now, but will submitt to github later. Less CPU-> less energy and more stable hashrates.
A noticable difference when using the computer while mining.

That will be most appreciated.  I had definitely noticed slugishness issues when my 2x750Ti would switch from Quark to X11.  Fortunately, it has been staying in Quark most of the time lately.
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
I have solved a bug wich causes the cpu to use alot of cpu. Now x11 is hashing with 0-1% cpu usage. Fixed in the unpublished spreadcoin fork for now, but will submitt to github later. Less CPU-> less energy and more stable hashrates.
A noticable difference when using the computer while mining.
Hi sp_.
We need the spreadx11 opt. miner asap, dif. is skyrocketing...

So is the price. 600% in one month. This coin is gold Smiley

I will test on my 970/980 rig on friday/saturday and release the first beta to the donators on sunday.
legendary
Activity: 2912
Merit: 1091
--- ChainWorks Industries ---
I have solved a bug wich causes the cpu to use alot of cpu. Now x11 is hashing with 0-1% cpu usage. Fixed in the unpublished spreadcoin fork for now, but will submitt to github later. Less CPU-> less energy and more stable hashrates.
A noticable difference when using the computer while mining.

awesome ...

let us know when that will be published and will compile and test right away ...

#crysx
legendary
Activity: 2716
Merit: 1116
I have solved a bug wich causes the cpu to use alot of cpu. Now x11 is hashing with 0-1% cpu usage. Fixed in the unpublished spreadcoin fork for now, but will submitt to github later. Less CPU-> less energy and more stable hashrates.
A noticable difference when using the computer while mining.

Hi sp_.

We need the spreadx11 opt. miner asap, dif. is skyrocketing...

sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
I have solved a bug wich causes the cpu to use alot of cpu. Now x11 is hashing with 0-1% cpu usage. Fixed in the unpublished spreadcoin fork for now, but will submitt to github later. Less CPU-> less energy and more stable hashrates.
A noticable difference when using the computer while mining.
full member
Activity: 168
Merit: 100
Hi guys, i need help i have driver crashing every hour or two, with all editions since 1.5.24 to realase 31 by sp of ccminer i mine with last drivers 347.09 on win 8.1 with 970gtx gpus, cards are not overclocked and driver is not crashing in games or any other app just with ccminer mining mjollnir algo.
legendary
Activity: 2912
Merit: 1091
--- ChainWorks Industries ---
First, I run two nearly identical Linux rigs, same MB,OS, different CPUS, slightly different cards.  Six cards per rig, the fast, reliable rig has 6x EVGA FTW 750ti cards, the slower rig has 6x EVGA SC 750ti cards.  The 750ti SC cards have no 6-pin PCI connector, and these cards run on a MB with a 2-core Sempron 145 (unlocked).

swap 3 of the cards with a 6 pin power connector to the slow rig. Some motherboards doesn't work if you have 6 cards without powered raisers. How many watt is the PSU. and is it gold rated?

All my cards have powered USB 3 risers, connected to the PSU.  None draw power from the MB directly on either rig.  The slow rig has a 600w bronze PSU, the fast one has a 750W bronze PSU.  Neither draws 500w at the wall.

I am thinking that I might lower the stock intensity.  It is a simple setting.  What are your current stock intensities for x11/x13/x14/x15/quark?       --scryptr

all my rigs are similar scryptr ...

all have powered usb3 risers and all run bronze or gold psu's ...

i still think this could be a compilation issue ...

but if you have already tried that - then it could be any number of things including a faulty card that is killing the others / system as sp mentioned earlier ...

start it off without ANY oc - all stock and see what happens to it ... if it still crashes after you have locally compiled the miner AND stocked all the clocks - then a hardware issue may be the answer ...

#crysx

crysx--

I tried v31, my fast rig is again stable, my slow rig crashes after 100-200 accepts.  I lowered the intensity to 18 (and lost 1mh/s performance!) and it was the same thing... crash.   I run stock clocks, period.

Both rigs are headless.  I plugged into the slow rig and booted normally, and noticed some error messages about nvidia driver 346.35.  It was released on 17 JAN 15, yesterday.  I think that I will have to purge the drivers and then re-install.  For Linux, v346.35 is a long-term release, no longer a beta, as 346.22 was.  I may burn the whole OS and move to 14.10, which is now supported by 346.35.      --scryptr

How long are your USB 3 cables for the risers?  Are any making sharp bends or twisting unnecessarily? 

I ordered 6 risers which were supposed to have 30cm cables but they came with about 60cm ones and I constantly had problems until I adjusted the usb connectors to fit more snug in the pci extension and smoothly routed the cables far away from each other straight out from the card and around the plastic crate I had them in.  The quality of most of the cables in kits is questionable and I don't think the longer ones can take the interference or cross-talk if they're bundled to close to each other or routed near high heat components.  I never really did figure out which one it was, but once I got them zip-tied to the crate holding them in a place they all worked, it stayed working well unless I did a major overhaul in the config.  I run windows and it was usually easy to tell what happened because a card would show up as 0 on the overclocks and the overboost wouldnt' be showing in Precision x anymore.

Never really had problems with my cables that were about 8" or however many cm that is.
Don't forget to get these usb hold down clips...they work great...no more loose cables.  Wink
http://www.miningcryptocurrency.com.au/store/p4/RiserSafe_By_the_Inventor_Patent_%28Pending%29.html

i know the guy that invented those clips ... he is one brilliant guy ... really nice too and doesnt live very far away from me ...

i have use some of those clips i got from him ... they work very well ...

#crysx
legendary
Activity: 1512
Merit: 1000
quarkchain.io
some progress on neoscrypt:
750ti: 81kh/s 110kh/s 107kh/s
780ti: 165kh/s 182kh/s 193kh/s
980:   290kh/s 299kh/s 335kh/s
(still doesn't scale right   Huh)

Let me know when the 980 hits 600kh/s.  Grin
done  Grin 630kh/s

nice , man , you rock... Smiley
legendary
Activity: 1512
Merit: 1000
quarkchain.io
@SP_
have issues with r.31 on qubit, cant accept share. with 750ti and 970 , driver 347.09...
Jump to: