Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1155. (Read 2347599 times)

If I remember right there were some strange issues with windows 8 & 8.1 and tsiv's cryptonight miner. Had to use 8x30, higher values and it crashed. On windows 7 8x60 worked and was faster too.

I believe you just need to disable the driver checking (or make the check interval larger):
https://msdn.microsoft.com/en-us/library/windows/hardware/ff569918%28v=vs.85%29.aspx

My working setup was to just set TdrLevel off (zero).
A more conservative launch value is 8x30 (though 8x60 does work best for me).

Cheers!

but it should work without launch config. the default tsiv used were quite conservative.

The hashrate is too high and the driver is crashing..

Seems to work now, I fixed a bug in benchmark mode

8x60 crashes

8x30 works.

never used it on win, on linux -l 8x60 was my always working launch config with a 750ti.

I have started on the cryptonight core now. But tsivs compiled version seems to crash on cuda 6.5 win8. Anyone have any tips ?
I tried different -l parameters but doesn't seem to work

sounds like a great deal of work ...

if version 8 can run faster - then wouldnt that be a 'leaner' miner? ...

would be nice to see if there was a leap in hashrate ...

#crysx

yeah, forgot to change it. The bottleneck in the spreadcoin hash is the cuda_sha256.cu function that iterates a 200000 element table.
I moved the table to a texture and got a small boost, but it broke the hash.
I will release a version 8 with higher speed soon.

tanx sp ... that is something that will be kept in the knowledgebase ...

i edited earlier - which i missed obviously - but the title of spreadminer is still at version 5 ...

thats not a big thing at all - but i thought you might want to know ...

...

#crysx

Different intensity  parameters for different cards.

For the 750ti -x parameters should be a multiple of 5
for the 750, 960,980  -x paramameter should be a multiple of 8
and the 970 -x parameter should be a multiple of 13

ive changed the parameter to -x 5 and it seems to perform about the same ... so ill keep it at that ...

as for the x32 builds ... ill have to test that another day - as ill setup a x32 machine ( fedora 19 or 20 x32 ) and build it from a fully 32bit platform ...

what other commandline parameters do you think would benefit the hashrate / processing? ...

btw - just a small thing - but the title still says spreadminer 5

#crysx

I run without launchparameter (default is set to -x 5 on the 750ti). The 750ti will reduce the clockspeed if the cards get too hot. Lower intensity might be bether in the long run. I also think that 32bit builds will be faster. My windows compile (x86) is performing much bether.

sp ...

it compiled after i changed the filenames as per your instruction in the Makefile.am file - though the second file required the .cu extension ...

ive now moved the compiled spreadminer to one of the production systems for testing ...

there is an improvement in performance - though not as much as i would have hoped using stock clocks per gpu - but nonetheless - an improvement in comparison to tsivs miner ...

below are the output of both the currently running spreadminer-tsiv and also the newly compiled spreadminer-sph both using the -x 20 parameter ( they are machines exactly the same specs ) ...


spreadminer-tsiv ...

[2015-03-02 18:15:35] accepted: 66765/68063 (98.09%), 7264 khash/s (yay!!!)
[2015-03-02 18:15:36] GPU #1: GeForce GTX 750 Ti, 1448 khash/s
[2015-03-02 18:15:37] GPU #2: GeForce GTX 750 Ti, 1464 khash/s
[2015-03-02 18:15:37] GPU #4: found a solution, nonce $CC22B141
[2015-03-02 18:15:37] GPU #4: GeForce GTX 750 Ti, 1445 khash/s
[2015-03-02 18:15:37] accepted: 66766/68064 (98.09%), 7259 khash/s (yay!!!)
[2015-03-02 18:15:38] GPU #3: found a solution, nonce $994E71C5
[2015-03-02 18:15:38] GPU #3: GeForce GTX 750 Ti, 1451 khash/s
[2015-03-02 18:15:38] accepted: 66767/68065 (98.09%), 7257 khash/s (yay!!!)
[2015-03-02 18:15:39] GPU #0: GeForce GTX 750 Ti, 1465 khash/s
[2015-03-02 18:15:42] GPU #1: GeForce GTX 750 Ti, 1447 khash/s
[2015-03-02 18:15:42] GPU #2: GeForce GTX 750 Ti, 1464 khash/s
[2015-03-02 18:15:43] GPU #4: GeForce GTX 750 Ti, 1453 khash/s
[2015-03-02 18:15:44] GPU #3: GeForce GTX 750 Ti, 1452 khash/s
[2015-03-02 18:15:45] GPU #0: GeForce GTX 750 Ti, 1463 khash/s
[2015-03-02 18:15:46] GPU #1: found a solution, nonce $3345AA8F
[2015-03-02 18:15:46] GPU #1: GeForce GTX 750 Ti, 1445 khash/s
[2015-03-02 18:15:46] accepted: 66768/68066 (98.09%), 7276 khash/s (yay!!!)
[2015-03-02 18:15:48] GPU #2: GeForce GTX 750 Ti, 1463 khash/s
[2015-03-02 18:15:49] GPU #4: GeForce GTX 750 Ti, 1454 khash/s
[2015-03-02 18:15:50] GPU #3: GeForce GTX 750 Ti, 1453 khash/s
[2015-03-02 18:15:50] GPU #0: found a solution, nonce $005965AD
[2015-03-02 18:15:50] GPU #0: GeForce GTX 750 Ti, 1465 khash/s
[2015-03-02 18:15:50] accepted: 66769/68067 (98.09%), 7280 khash/s (yay!!!)
[2015-03-02 18:15:52] GPU #1: GeForce GTX 750 Ti, 1447 khash/s
[2015-03-02 18:15:52] GPU #0: found a solution, nonce $0022EAEA
[2015-03-02 18:15:52] GPU #0: GeForce GTX 750 Ti, 1458 khash/s
[2015-03-02 18:15:53] accepted: 66770/68068 (98.09%), 7276 khash/s (yay!!!)
[2015-03-02 18:15:53] GPU #2: GeForce GTX 750 Ti, 1464 khash/s
[2015-03-02 18:15:54] GPU #3: found a solution, nonce $99469AD2
[2015-03-02 18:15:54] GPU #3: GeForce GTX 750 Ti, 1450 khash/s
[2015-03-02 18:15:54] accepted: 66771/68069 (98.09%), 7274 khash/s (yay!!!)


spreadminer-sph ...

     *** Spreadminer 5 - a SpreadX11 miner for nVidia GPUs ***
           based on ccminer 1.2 Copyright 2014 Christian Buchner, Christian H.
           based on pooler-cpuminer 2.3.2 (c) 2010 Jeff Garzik, 2012 pooler
           SpreadX11 additions by tsiv
[2015-03-02 18:14:09] 5 miner threads started, using 'spreadx11' algorithm.
[2015-03-02 18:14:10] GPU #0: GeForce GTX 750 Ti, 1280 khash/s
[2015-03-02 18:14:11] GPU #2: GeForce GTX 750 Ti, 799.36 khash/s
[2015-03-02 18:14:13] GPU #2: found a solution, nonce $6607585D
[2015-03-02 18:14:13] GPU #2: GeForce GTX 750 Ti, 1458 khash/s
[2015-03-02 18:14:13] GPU #3: GeForce GTX 750 Ti, 582.66 khash/s
[2015-03-02 18:14:13] accepted: 1/1 (100.00%), 3321 khash/s (yay!!!)
[2015-03-02 18:14:15] GPU #0: GeForce GTX 750 Ti, 1573 khash/s
[2015-03-02 18:14:16] GPU #3: GeForce GTX 750 Ti, 1607 khash/s
[2015-03-02 18:14:16] GPU #2: found a solution, nonce $664BC097
[2015-03-02 18:14:16] GPU #2: GeForce GTX 750 Ti, 1546 khash/s
[2015-03-02 18:14:17] accepted: 2/2 (100.00%), 4726 khash/s (yay!!!)

and after the miner has settled and processed a few more shares ...

[2015-03-02 18:18:40] accepted: 85/87 (97.70%), 7635 khash/s (yay!!!)
[2015-03-02 18:18:40] GPU #0: found a solution, nonce $001056A8
[2015-03-02 18:18:40] GPU #0: GeForce GTX 750 Ti, 1359 khash/s
[2015-03-02 18:18:41] accepted: 86/88 (97.73%), 7449 khash/s (yay!!!)
[2015-03-02 18:18:41] GPU #1: GeForce GTX 750 Ti, 1510 khash/s
[2015-03-02 18:18:41] GPU #4: GeForce GTX 750 Ti, 1573 khash/s
[2015-03-02 18:18:42] GPU #3: GeForce GTX 750 Ti, 1553 khash/s
[2015-03-02 18:18:45] GPU #2: GeForce GTX 750 Ti, 1545 khash/s
[2015-03-02 18:18:46] GPU #0: GeForce GTX 750 Ti, 1545 khash/s
[2015-03-02 18:18:47] GPU #1: GeForce GTX 750 Ti, 1510 khash/s
[2015-03-02 18:18:47] GPU #3: GeForce GTX 750 Ti, 1552 khash/s
[2015-03-02 18:18:48] GPU #4: GeForce GTX 750 Ti, 1574 khash/s
[2015-03-02 18:18:49] GPU #2: found a solution, nonce $66497DF1
[2015-03-02 18:18:49] GPU #2: GeForce GTX 750 Ti, 1495 khash/s
[2015-03-02 18:18:49] accepted: 87/89 (97.75%), 7675 khash/s (yay!!!)
[2015-03-02 18:18:50] GPU #1: found a solution, nonce $332952A2
[2015-03-02 18:18:50] GPU #1: GeForce GTX 750 Ti, 1445 khash/s
[2015-03-02 18:18:50] accepted: 88/90 (97.78%), 7610 khash/s (yay!!!)
[2015-03-02 18:18:51] GPU #0: found a solution, nonce $005DB092
[2015-03-02 18:18:51] GPU #0: GeForce GTX 750 Ti, 1505 khash/s
[2015-03-02 18:18:51] accepted: 89/91 (97.80%), 7570 khash/s (yay!!!)


so as you can see there is a marked difference between the two workers ...

ill now propagate the latest spreadminer-sph across the farm and see how they perform overnight ...

tanx again sp ... this will help a great deal overall ...

#crysx

ok - changed and compiling once again - lets see what happens here ...

tanx sp ...

#crysx

keep the filename, but rename filename in the makefile. Cases are important, so you need to rename the cuda_jh512keccak512  with a small k in Keccak

sp ...

the first was already there in the directory ...

the second is named currently cuda_jh512keccak512.cu ...

should i rename it to cuda_jh512keccak512 ? ...

#crysx

You managed to compile tsivs version, then you should compile mine as well. I use cuda6.5 on windows.

There are spelling errors in the makefile.am You need to rename these files:

spreadx11/x11_luffaCubehash512_cpu_hash_64.cu  rename to  spreadx11/cuda_x11_luffa512_Cubehash.cu
spreadx11/cuda_jh512Keccak512_cpu_hash_64    rename to spreadx11/cuda_jh512keccak512

ok - errors out with the following ...

#######

mv -f sph/.deps/spreadminer-simd.Tpo sph/.deps/spreadminer-simd.Po
In file included from /usr/local/cuda/bin/../targets/x86_64-linux/include/cuda_runtime.h:59:0,
                 from :0:
/usr/local/cuda/bin/../targets/x86_64-linux/include/host_config.h:82:2: error: #error -- unsupported GNU version! gcc 4.9 and up are not supported!
 #error -- unsupported GNU version! gcc 4.9 and up are not supported!
  ^
Makefile:1232: recipe for target 'cuda_util.o' failed
make[2]: *** [cuda_util.o] Error 1
make[2]: Leaving directory '/mnt/horustor/compile/ccminer-all/sp-hash/spreadminer/Spreadminer7_source.7_FILES/sptest'
Makefile:808: recipe for target 'all-recursive' failed
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory '/mnt/horustor/compile/ccminer-all/sp-hash/spreadminer/Spreadminer7_source.7_FILES/sptest'
Makefile:393: recipe for target 'all' failed
make: *** [all] Error 2

#######

does this mean i have conflicting versions of cuda / gcc ? ...

#crysx

i copied the files directly to the spreadminer7 folder and it seems to have started the compile ...

if the ccminer-spmod is anything to go by - this should take a little bit of time to finish ...

ill keep you informed in the next 15-30mins as to how this goes ...

one question though - does this miner run via a pool ( like tsivs one does ) ? ...

tanx ...

#crysx