CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1133.

chrysophylax

legendary

Activity: 2912

Merit: 1091

--- ChainWorks Industries ---

Quote from: sp_ on April 18, 2015, 11:46:50 PM

-g 4 -i 10 works for x11 but you get some "does not validate on the cpu". And I think the miner will crash if you let it run for a while. But the gain is around 25% on the pool. (Testrig with Gtx 960/970/750ti). For quark you probobly need to lower the intensity.

im running it at the moment with 2/3rds of the farm on westhash with qaurk at the moment as a test ... the rest of the 1/3rd is either amd 280x or crashed ...

i havent implemented the -g parameter as i didnt want to push the machines too much ... and im not in the office yet ... did this remotely ...

you can see for yourself how they are performing on westhash if you like ... the bitcoin address is 15umzHXF8NzXA4FywmeFbrDHgL8WcPs3wx to look the performance up ...

https://www.westhash.com/index.jsp?p=miners&a=12&addr=15umzHXF8NzXA4FywmeFbrDHgL8WcPs3wx - is the farm as it stands ... i will reboot the crashed systems when i get back in the next few hours ...

let me know what you think regarding this performance in the 'real' world as opposed to test bench ...

tanx ...

#crysx

sp_

legendary

Activity: 2926

Merit: 1087

Team Black developer

-g 4 -i 10 works for x11 but you get some "does not validate on the cpu". And I think the miner will crash if you let it run for a while. But the gain is around 25% on the pool. (Testrig with Gtx 960/970/750ti). For quark you probobly need to lower the intensity.

chrysophylax

legendary

Activity: 2912

Merit: 1091

--- ChainWorks Industries ---

wow guys ...

i just woke and reading up on whats been happening ...

checked the test machine on the farm and its crashed using v45 after i left it last night ... its a 5x gigabyte 750ti oc lp with stock clocks under fedora 19 x64 ...

the rest of the farm is running fine - all running v44 ...

i cant wait to get to the office and test the latest release and see if its fixed ...

what would be the best parameters to use ... ill test them anyway - but just thought id ask ... as i see -g 4 -i 10 crashes scryptrs machine and -g2 -i 10 runs but with errors ...

the test machine runs with NO parameters given except for the algo / username / password parameters ...

tanx ...

#crysx

AliMan

hero member

Activity: 1974

Merit: 502

Vave.com - Crypto Casino

Quote from: scryptr on April 18, 2015, 06:43:22 PM

Quote from: AliMan on April 18, 2015, 06:00:43 PM

Quote from: scryptr on April 18, 2015, 05:46:39 PM

Quote from: AliMan on April 18, 2015, 05:22:56 PM

Yea Quark dropped a significant bit.

QUARK and v45--

Quark dropped from 36.6Mh/s to 36.3Mh/s on my 6x750ti FTW rig, but only when running that algo at default intensity and no gputhreads "-g" parameter. If the gputhreads implementation is fixed properly, the rig will likely mine at a rate over 40Mh/s.

Currently, an AMD 290X struggles to make 5Mh/s per card for Quark with the best sgminer bins available. I get over 6Mh/s per 750ti. Pay the piper? --scryptr

I'm getting 17-18 Mh/s on a 980 and 14-15 Mh/s on a 970.

With the latest updated binary, I'm getting 5-7 Mh/s on the 980, not the precompiled release45.

WINDOWS or LINUX?

What is your OS? My 750ti rig is Linux, I compiled the latest update and am getting 36.3 Mh/s. --scryptr

Windows 8.1

scryptr

legendary

Activity: 1797

Merit: 1028

NEW SP_ FIX --

I just downloaded the git update (30 min old) and compiled. My Linux rig is back to 36.6Mh/s. --scryptr

p.s. I was able to run at 36.6Mh/s with no problems or errors. I then changed my launch config to include "-g 2" and re-launched. The rig runs at 36Mh/s with a few "does not validate on CPU" errors. What may be useful would be to include code that labled the thread as well as the GPU, i.e. "GPU #0 t #1: Geforce 750 Ti, 6100kH/s", or something similar. The miner could see if all threads are running. Currently, it appears that the combined total hash rate for all threads on a gpu is being printed to the screen. --scryptr

scryptr

legendary

Activity: 1797

Merit: 1028

Quote from: AliMan on April 18, 2015, 06:00:43 PM

Quote from: scryptr on April 18, 2015, 05:46:39 PM

Quote from: AliMan on April 18, 2015, 05:22:56 PM

Yea Quark dropped a significant bit.

QUARK and v45--

Quark dropped from 36.6Mh/s to 36.3Mh/s on my 6x750ti FTW rig, but only when running that algo at default intensity and no gputhreads "-g" parameter. If the gputhreads implementation is fixed properly, the rig will likely mine at a rate over 40Mh/s.

Currently, an AMD 290X struggles to make 5Mh/s per card for Quark with the best sgminer bins available. I get over 6Mh/s per 750ti. Pay the piper? --scryptr

I'm getting 17-18 Mh/s on a 980 and 14-15 Mh/s on a 970.

With the latest updated binary, I'm getting 5-7 Mh/s on the 980, not the precompiled release45.

WINDOWS or LINUX?

What is your OS? My 750ti rig is Linux, I compiled the latest update and am getting 36.3 Mh/s. --scryptr

AliMan

hero member

Activity: 1974

Merit: 502

Vave.com - Crypto Casino

Quote from: scryptr on April 18, 2015, 05:46:39 PM

Quote from: AliMan on April 18, 2015, 05:22:56 PM

Yea Quark dropped a significant bit.

QUARK and v45--

Quark dropped from 36.6Mh/s to 36.3Mh/s on my 6x750ti FTW rig, but only when running that algo at default intensity and no gputhreads "-g" parameter. If the gputhreads implementation is fixed properly, the rig will likely mine at a rate over 40Mh/s.

Currently, an AMD 290X struggles to make 5Mh/s per card for Quark with the best sgminer bins available. I get over 6Mh/s per 750ti. Pay the piper? --scryptr

I'm getting 17-18 Mh/s on a 980 and 14-15 Mh/s on a 970.

With the latest updated binary, I'm getting 5-7 Mh/s on the 980, not the precompiled release45.

scryptr

legendary

Activity: 1797

Merit: 1028

Quote from: AliMan on April 18, 2015, 05:22:56 PM

Yea Quark dropped a significant bit.

QUARK and v45--

Quark dropped from 36.6Mh/s to 36.3Mh/s on my 6x750ti FTW rig, but only when running that algo at default intensity and no gputhreads "-g" parameter. If the gputhreads implementation is fixed properly, the rig will likely mine at a rate over 40Mh/s.

Currently, an AMD 290X struggles to make 5Mh/s per card for Quark with the best sgminer bins available. I get over 6Mh/s per 750ti. Pay the piper? --scryptr

p.s. It's like getting an invisible 290X that runs on invisible electricity that doesn't cost anything, and the shipping was free, too! Not to mention that you wouldn't have any extra setup hassles... --scryptr

AliMan

hero member

Activity: 1974

Merit: 502

Vave.com - Crypto Casino

Yea Quark dropped a significant bit.

scryptr

legendary

Activity: 1797

Merit: 1028

Quote from: sp_ on April 18, 2015, 04:28:58 PM

X11,x13,x15 +25% on the pool with -g 4 -i 10, but the crash needs to be removed. Donate some beers and i will fix it:) Quark is up less than 10%because it is already running the kernals in paralell.

SP_

I sent my 2 bits worth a few seconds ago. I hope that others also contribute. The "gputhreads" should be a valuable option for the miner. --scryptr

p.s. 21Mh/s plus! --

I am getting 21Mh/s on my 6x750ti FTW rig mining X11 with intensity at default and "-g 2". This is up from a solid 18.9 Mh/s with v44. Unfortunately, "-i 10 -g 4" crashed on launch. --scryptr

sp_

legendary

Activity: 2926

Merit: 1087

Team Black developer

X11,x13,x15 +25% on the pool with -g 4 -i 10, but the crash needs to be removed. Donate some beers and i will fix it:) Quark is up less than 10%because it is already running the kernals in paralell.

scryptr

legendary

Activity: 1797

Merit: 1028

Quote from: sp_ on April 18, 2015, 11:13:25 AM

Quote from: scryptr on April 18, 2015, 10:42:15 AM

SP_ v45-fixed --

My 6x750ti Linux rig running Quark at "i 21" and "g 2" is hashing at 39,000+ kh/s, albeit with errors and rate fluctuation:

--scryptr

What about without the -g parameter?

Looks like the -g need some more work. It might not work on all algos..

THE "-g 2" FLAG --

It shows a lot of potential. I was getting 3Mh/s more than my rig was getting before using the flag, until it crashed after about 90 minutes. I tried "-i 10 -g 4" and it crashed. I wonder if the 4gb 750ti could handle it better? --scryptr

p.s. I think that the 4GB 750ti (STRIX-GTX750TI-DC2OC-4GD5) is not currently in production by ASUS. --scryptr

sp_

legendary

Activity: 2926

Merit: 1087

Team Black developer

-i 10 -g 4 seems to work with the latest on git.

x11 is 25% faster on my 3 card testrig. gtx970,960,750ti

AliMan

hero member

Activity: 1974

Merit: 502

Vave.com - Crypto Casino

Any idea about the expected payouts on yaamp for quark?

chrysophylax

legendary

Activity: 2912

Merit: 1091

--- ChainWorks Industries ---

Quote from: sp_ on April 18, 2015, 10:37:43 AM

Quote from: chrysophylax on April 18, 2015, 10:11:29 AM

Quote from: sp_ on April 18, 2015, 09:32:49 AM

looks like there is a bug in build 45 on some rigs. only 2 of 6 cards are hashing. working on it

the current compile of 45 actually broke x11 for the systems here sp ...
quark runs nicely though

...
out of memory and core dump errors ... and 44 is fine with those algos though ...
#crysx

Get the latest version and try x11 with -g 4!

the 750ti is peaking at 3200 at standard clocks!

edit:

in benchmark mode, but on the pool I get some does not validate on the cpu...

the -g 4 parameter ( actually -g 2 also ) just doesnt work - it drops just after it starts with no error message that i can see ...

take the -g 4 parameter out - and its all ok with this newer build ... but only on some of the machines in the farm ...

with others - v44 needed to be reinstated due to v45 giving errors with the aforementioned - core dump and memory out errors ...

will get to this tomorrow - as its late here now and sleep awaits

...

tanx again sp ...

#crysx

sp_

legendary

Activity: 2926

Merit: 1087

Team Black developer

Quote from: scryptr on April 18, 2015, 10:42:15 AM

SP_ v45-fixed --

My 6x750ti Linux rig running Quark at "i 21" and "g 2" is hashing at 39,000+ kh/s, albeit with errors and rate fluctuation:

--scryptr

What about without the -g parameter?

Looks like the -g need some more work. It might not work on all algos..

Bombadil

hero member

Activity: 644

Merit: 500

Quote from: sp_ on April 18, 2015, 08:33:30 AM

I have added a threads per gpu parameter just like in sgminer. ex: -g 2

Some algos get a boost.

Most impovement on compute 5.2 devices. (960,970,980)

x11 with -g 2 on the 750 ti, the performance is reduced by 0.5MHASH.
The parameter is not always producing higher hashrates.

The stats are reported by each thread, so some more work is needed. (run with -q, so the screen is not flooded with hashrates)

Note that -g 2 will reduce the intensity by one. this can be overridden with the -i parameter.

I will build release 45 now.

-d 0,0,1,1 ^^'

scryptr

legendary

Activity: 1797

Merit: 1028

SP_ v45-fixed --

My 6x750ti Linux rig running Quark at "i 21" and "g 2" is hashing at 39,000+ kh/s, albeit with errors and rate fluctuation:

--scryptr

sp_

legendary

Activity: 2926

Merit: 1087

Team Black developer

Quote from: chrysophylax on April 18, 2015, 10:11:29 AM

Quote from: sp_ on April 18, 2015, 09:32:49 AM

looks like there is a bug in build 45 on some rigs. only 2 of 6 cards are hashing. working on it

the current compile of 45 actually broke x11 for the systems here sp ...
quark runs nicely though

...
out of memory and core dump errors ... and 44 is fine with those algos though ...
#crysx

Get the latest version and try x11 with -g 4!

the 750ti is peaking at 3200 at standard clocks!

edit:

in benchmark mode, but on the pool I get some does not validate on the cpu...

sp_

legendary

Activity: 2926

Merit: 1087

Team Black developer

fixed the bug, and corrected the hashrate statistics. Please try the new version.

Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1133. (Read 2347601 times)