Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1086. (Read 2347601 times)

tanx sp ...

im familiar with writing very BASIC scripts ( as the entire farm runs on scripts ive written for its automation ) ... but the setup is the major issue with this ...

the settings are never persistent which is why im guessing you are adding the settings into the initial part of the script ...

will have a look into this much deeper with the test machine tomorrow as its late here now ...

its interesting that the overclocking brings in such a difference ...

under x11 - the most i can squeeze out on stock clocks with the gigabyte 750ti oc lp card is about 2850KH ...

whereas you are getting more than 3400KH ...

definitely worth investigating further ... as the whole farm is based on the x11 algo mining ...

will let you know ... and tanx again ...

#crysx

Crysx, I don't see the problems you are having with westhash. I've had a continuous connection to the quark
pool for 1575 minutes. As I mentioned to Masked_Immortal the only problem i see on westhash is the
one retry message on startup. Does your proxy support extranonce?

Ditto on the search for a good multipool. I have used simplemulti and xpool but the're focussed on x11/DASH, and neither
has a quark pool.  I don't know what's behind Mintsy's front page. Globalhash looked promissing but came
and went in less than a weekl. It's looking like it's nicehash & westhash for now.

Make a textfile with a script and call it sp.x:

nvidia-smi -pm 1
nvidia-smi -ac 1350,1350
ccminer -quark ...... pool btc adress etc.

(you might need to change the paths and give the correct admin rights to nvidia-smi)

chmod +arwx sp.x


then just write sp.x and press enter.


Then you need to find the correct clocks for your card. Replace the 1350 / 1350 with something bether. You should also messure the powerusage to get the optimal hash/watt rate.

Can you run the nvidia command line tool to check wich clockrates the card is running on?

I get 3420KHASH on my overclocked 750ti in windows when mining x11.

sudo?

exactly my point ...

we dont have 30 machines .. yet ...

but even with the farm in its present form - it would gain a huge boost ...

the only factor is that using the smi tool is fiddly and VERY complicated ...

hence the reason for the wait for the simple --gpuclock --memclock type of parameters ... which are a GodSend for ccminer ... ( and make it VERY simple in sgminer ) ...

ive used and tested the smi tool - it does work ... BUT the amount of work put into it is monstrous ...

nvidia do not make it easy - and easy is what we want ...

in any case ... there is much more work to be done on the current farm while we build the 'new' one ... so simplicity is the hey to ANY miner for ANY system - whether windows or linux - whether single machine or farm ...

next week im personally committing an entire week to the farm then full time into this ...

will let you know how we go with it all sp ...

tanx ...

#crysx

The results i posted are from Linux, the hash rates seem pretty comparable to Windows
results, no idea about the TDP.

thats what i figured would happen ...

but would be nice to see what they come up with AFTER the initial rebranding ...

#crysx

Adjusting the clock settings could give a boost of 20% without drawing more power on some cards in some algos.

First set the right mode

nvidia-smi -pm 1

then set the right clock.

here is the tool's helpfile:

  -pm,  --persistence-mode=   Set persistence mode: 0/DISABLED, 1/ENABLED
    -e,   --ecc-config=         Toggle ECC support: 0/DISABLED, 1/ENABLED
    -p,   --reset-ecc-errors=   Reset ECC error counts: 0/VOLATILE, 1/AGGREGATE
    -c,   --compute-mode=       Set MODE for compute applications:
                                0/DEFAULT, 1/EXCLUSIVE_THREAD,
                                2/PROHIBITED, 3/EXCLUSIVE_PROCESS
          --gom=                Set GPU Operation Mode:
                                    0/ALL_ON, 1/COMPUTE, 2/LOW_DP
    -r    --gpu-reset           Trigger reset of the GPU.
                                Can be used to reset the GPU HW state in situations
                                that would otherwise require a machine reboot.
                                Typically useful if a double bit ECC error has
                                occurred.
                                Reset operations are not guarenteed to work in
                                all cases and should be used with caution.
                                --id= switch is mandatory for this switch
    -ac   --applications-clocks= Specifies  clocks as a
                                    pair (e.g. 2000,800) that defines GPU's
                                    speed in MHz while running applications on a GPU.
    -rac  --reset-applications-clocks
                                Resets the applications clocks to the default values.
    -acp  --applications-clocks-permission=
                                Toggles permission requirements for -ac and -rac commands:
                                0/UNRESTRICTED, 1/RESTRICTED
    -pl   --power-limit=        Specifies maximum power management limit in watts.
    -am   --accounting-mode=    Enable or disable Accounting Mode: 0/DISABLED, 1/ENABLED
    -caa  --clear-accounted-apps
                                Clears all the accounted PIDs in the buffer.
          --auto-boost-default= Set the default auto boost policy to 0/DISABLED
                                or 1/ENABLED, enforcing the change only after the
                                last boost client has exited.
          --auto-boost-permission=
                                Allow non-admin/root control over auto boost mode:
                                0/UNRESTRICTED, 1/RESTRICTED

R9 300 series will not be a rebrand only: there is new memory technology and other goodies; but they'll probably, at least intially, still be based on 28nm tehnology.

Find an updated version of the nvidia-smi tool. Does it work to change the clocks with this tool for your graphic cards?

If not, I cannot help you on linux. It's a API/Driver problem.

on windows the tool is located here:

C:\Program Files\NVIDIA Corporation\NVSMI

we appreciate it sp ... especially all the work you have done so far ...

but i have to admit that such an advancement like this in ccminer would be a much missed feature for linux ...

what is it specifically that is the issue with this ( for it to be 'just' windows currently )? ...

is it the drivers that nvidia bring out? ... or the way that ccminer is built? ... or a combination of both? ... or something else entirely? ...

#crysx

I will continue to support linux. But the memclock/gpuclock is windows only for now.

if amd are going the way they did with 7970 etc which were rebranded ( and that was it for the first round - just a rebrand ) as 280x - then for the next 6 months - we will not really see anything worth the while ...

unless of course they decide that an actual rebrand AND redesign is what they will bring out ...

but its all conjecture for the moment ... would be nice for a redesign though ...

...

#crysx

our farm was completely amd ... now its a hybrid amd / nvidia - though mostly nvidia now ...

the headaches involved with catalyst drivers and the lack of linux support was beyond us - so was one of the major reasons we changed over to nvidia - apart from the power consumption / hash savings ...

either way - it was worth the effort ... it still is ...

but to change to cumbersome top heavy os like windows JUST to mine? ... not a chance ...

will find another way and support that way ...

until microsoft decide to release a FREE version of windows ( which looks like version 10 anyway ) - even then we will be hard pressed to reconfigure and redesign the ENTIRE farm for that os ...

will consider where we go next once we see how the 'scene' goes on this end ...

#crysx

Take it as a rumor for now but the only cards probably worth it will be the top of the line 3xx.

http://videocardz.com/55499/amds-hawaii-gpu-to-return-with-radeon-300-series

i wont be switching ( about to be ) 30 + machines to windows sp ...

it becomes way too much maintenance for each system - and way too expensive for the licensing to microsoft ...

id rather spend all the money upgrading the farm AND donating to devs that will support the effort ...

you have a fantastic ccminer fork here - one that stands out from many that are out there ...

not supporting linux in this way dulls the shine you have with this fork ...

and the community support that you have with those that mine with linux - will be disappointing to see that you wont support this side of the community ...

im actually a little taken back with that mate - to be totally honest with you ...

#crysx

reason why I'm not buying nvidia: amd drivers for linux are bad but the situation with nvidia is even worse.
another reason is I wanna see the r9 3XX in action before buying for the winter.

Lazy..

The windows builds produce more hash, and use less power. You can get 100% more  profit if you switch to windows.