Author

Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1205. (Read 2347597 times)

legendary
Activity: 2912
Merit: 1091
--- ChainWorks Industries ---
BTW: with reliese21 I hit 7930kH/s on x11 with my GTX970...

Three of those would beat either of my 750ti rigs.   Grin  I might sell my 280x cards for performance like that!       --scryptr

Im considering to replace my 280x for nvidia , but Ill wait the TIs

im doing just that ...

in fact - ALL my amd will be removed and replaced with gigabyte 750ti oc cards ... and new motherboards to boot ...

the order has been placed today and will take a week to get here ... massive christmas present for moi ( with heaps of work to redo the farm ) ... Wink

this will also allow the farm to grow to about 3 times the size it is now also ...

cant wait ...

#crysx
legendary
Activity: 2912
Merit: 1091
--- ChainWorks Industries ---
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
I shouldn't have started from Wolf0 opencl code  Grin

Haha, yeah, it's hard to read... but it's a lot faster than it was!

I can run cgminer 3.7.8 neoscrypt on a 550ti at a rate of 25kh/s (Win64).  However, I cannot run cgminer 3.7.8 neoscrypt on a 6-card 750ti rig, Ubuntu 14.04.1. I can compile it, but it won't run properly, it just spits HW errors.

My 750ti rigs run smoothly with compiled ccminer 1.5.2-git code.  Are there any special command-line switches to use during compilation to make cgminer 3.7.8 or sgminer 5 work properly on nVidia Linux rigs?  The code compiles, I am pretty sure that I have the pre-requisites in place.

Personally, I think a 750ti could likely hash on a level of a 280x if the cuda code is right.       --scryptr

agreed - it would run very well if it was coded right ... if only we had the fine grained control that sgminer does with ccminer and cuda ...

as a side issue - i run sgminer/neoscrypt on the 6 card 750ti with out issue - but the hashrates are poor ... so i stick with ccminer/x11 ...

#crysx

Is there some trick to running cg/sg-miner on nVidia?  Like I said, I have the windows binary runnning on my Win64 box, with an older 550ti.  But I can't run it on Ubuntu without 100% HW errors.  And, I do set the environment variables for the GPUs, like for an AMD rig.

What kind of hash rate do you get per card / per rig (750ti) with the OpenCL code?       --scryptr

p.s.  My 550ti would get 90kh/s or better  for Litecoin with the original cudaminer 28 FEB 14 version.  About even with an AMD 5670 on cgminer 3.7.2.       --scryptr

no mate - no trick ... i compile under fedora 19 or 20 ( im a red hat kinda guy ) Wink and just run ... but the hashrate per card on a gigabyte 750ti oc is shit ... 20KH is the max i can sqeeze out of opencl sgminer/neoscrypt ... i gave up and stuck with ccminer/x11 ... granitecoin and most other coins i mine are x11 - so its more beneficial anyway ...

i started crypto with litecoin then feathercoin ... and now neoscrypt has taken hold - i decided to do a little testing - only to find that sgminer will run neoscrypt 'okay' with amd cards - but crap on nvidia cards ... for obvious reasons ...

so when the fellas here do their magic with neoscrypt and ccminer/nvidia/cuda - ill be back again mining neoscrypt ...

#crysx
legendary
Activity: 1512
Merit: 1000
quarkchain.io
BTW: with reliese21 I hit 7930kH/s on x11 with my GTX970...

Three of those would beat either of my 750ti rigs.   Grin  I might sell my 280x cards for performance like that!       --scryptr

Im considering to replace my 280x for nvidia , but Ill wait the TIs
legendary
Activity: 1797
Merit: 1028
BTW: with reliese21 I hit 7930kH/s on x11 with my GTX970...

Three of those would beat either of my 750ti rigs.   Grin  I might sell my 280x cards for performance like that!       --scryptr
legendary
Activity: 1512
Merit: 1000
quarkchain.io
BTW: with reliese21 I hit 7930kH/s on x11 with my GTX970...
legendary
Activity: 1797
Merit: 1028
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
I shouldn't have started from Wolf0 opencl code  Grin

Haha, yeah, it's hard to read... but it's a lot faster than it was!

I can run cgminer 3.7.8 neoscrypt on a 550ti at a rate of 25kh/s (Win64).  However, I cannot run cgminer 3.7.8 neoscrypt on a 6-card 750ti rig, Ubuntu 14.04.1. I can compile it, but it won't run properly, it just spits HW errors.

My 750ti rigs run smoothly with compiled ccminer 1.5.2-git code.  Are there any special command-line switches to use during compilation to make cgminer 3.7.8 or sgminer 5 work properly on nVidia Linux rigs?  The code compiles, I am pretty sure that I have the pre-requisites in place.

Personally, I think a 750ti could likely hash on a level of a 280x if the cuda code is right.       --scryptr

agreed - it would run very well if it was coded right ... if only we had the fine grained control that sgminer does with ccminer and cuda ...

as a side issue - i run sgminer/neoscrypt on the 6 card 750ti with out issue - but the hashrates are poor ... so i stick with ccminer/x11 ...

#crysx

Is there some trick to running cg/sg-miner on nVidia?  Like I said, I have the windows binary runnning on my Win64 box, with an older 550ti.  But I can't run it on Ubuntu without 100% HW errors.  And, I do set the environment variables for the GPUs, like for an AMD rig.

What kind of hash rate do you get per card / per rig (750ti) with the OpenCL code?       --scryptr

p.s.  My 550ti would get 90kh/s or better  for Litecoin with the original cudaminer 28 FEB 14 version.  About even with an AMD 5670 on cgminer 3.7.2.       --scryptr
legendary
Activity: 2912
Merit: 1091
--- ChainWorks Industries ---
anyone else get frequent crashes?  I run on three separate win8 rigs and all occasionally get a pop-up window saying the ccminer.exe stopped working.  Or, sometimes it stops working without a pop-up window.

It's not bad but of course very noticeable when the original/slower ccminer doesn't crash at all.  And the crashes aren't too frequent but at least once in 24 hours so I check it daily.  I know I could do some other monitor things or just set it to restart every 12 hours but... I'm lazy!  And... just wanted to bring it up for any discussion.

thx

which ccminer is it - url/link please?

im setting a windows system up and want to test the 750ti oc on that ...

tanx in advance ...

#crysx

It's the latest release from the OP.  I've been using the latest link from the OP and some releases are better than others.  I think 19 may have been a little more stable than 20.  I'm on 20 now.

ok - so ive setup a windows vista system ( the only 'legit' version i had ) and the ccminer21 from the OP is crashing with a msvcr120.dll issue ...

ive tried to fix it by downloading teh file and even even installing the runtime - but to no avail ... it still crashes ...

its been a very long time since ive used windows on an OS level ... anyone with any ideas on how to fix this error? ...

i just want to test how it all runs under windows and whether its worth it on this system ...

compiling from scratch on linux is great - and i have a little more control over what can be done ... but im at wits end as to how to get it to run under windows ...

the original ccminer from cbuchner1 runs well but at 2430 average hash on x11 ...

any ideas guys?

#crysx
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
#1 please.  Less crashes would be really nice.  Is 21 any better than 20?

21 has got less HW errors in x11. and a littlebit faster.

Quark is really fast. 14.5 MHASH on the 980
legendary
Activity: 2296
Merit: 1031
Good job as always SP! Lyra2 hash is a bit better.

However, same issue as with x11: Pool does not report the full hashrate. I think this is a ccminer 1.5.x related problem. For the original ccminer-Lyra2RE, pools show the correct hashrate as it is displayed in the miner. Would you mind to include the Lyra2 improvements into djm34's ccminer-Lyra2RE release?

I have done some fixes in x11 , is it still underrepporting in release 21?

Lyra is currently only finding one solution and the throughput is big. I will add 2 solutions soon.

I am considering to change the fork to the DJM34 fork. Clean it and replace all the kernals.

1. Less crashes.
2. Fix the problem on some pools.

But will lack some features like the API etc..


#1 please.  Less crashes would be really nice.  Is 21 any better than 20?
legendary
Activity: 2296
Merit: 1031
anyone else get frequent crashes?  I run on three separate win8 rigs and all occasionally get a pop-up window saying the ccminer.exe stopped working.  Or, sometimes it stops working without a pop-up window.

It's not bad but of course very noticeable when the original/slower ccminer doesn't crash at all.  And the crashes aren't too frequent but at least once in 24 hours so I check it daily.  I know I could do some other monitor things or just set it to restart every 12 hours but... I'm lazy!  And... just wanted to bring it up for any discussion.

thx

which ccminer is it - url/link please?

im setting a windows system up and want to test the 750ti oc on that ...

tanx in advance ...

#crysx

It's the latest release from the OP.  I've been using the latest link from the OP and some releases are better than others.  I think 19 may have been a little more stable than 20.  I'm on 20 now.
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
Good job as always SP! Lyra2 hash is a bit better.

However, same issue as with x11: Pool does not report the full hashrate. I think this is a ccminer 1.5.x related problem. For the original ccminer-Lyra2RE, pools show the correct hashrate as it is displayed in the miner. Would you mind to include the Lyra2 improvements into djm34's ccminer-Lyra2RE release?

I have done some fixes in x11 , is it still underrepporting in release 21?

Lyra is currently only finding one solution and the throughput is big. I will add 2 solutions soon.

I am considering to change the fork to the DJM34 fork. Clean it and replace all the kernals.

1. Less crashes.
2. Fix the problem on some pools.

But will lack some features like the API etc..
legendary
Activity: 1400
Merit: 1050
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
I shouldn't have started from Wolf0 opencl code  Grin
Don't you like 16-dimensional vectors?  Grin
would be good if they were existing natively in cuda... (sure I could probably create a similar structure...)

lol, sorry. They're so handy, though!
especially for the permutation... really miss that thing
legendary
Activity: 1400
Merit: 1050
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
I shouldn't have started from Wolf0 opencl code  Grin
Don't you like 16-dimensional vectors?  Grin
would be good if they were existing natively in cuda... (sure I could probably create a similar structure...)
legendary
Activity: 2912
Merit: 1091
--- ChainWorks Industries ---
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
I shouldn't have started from Wolf0 opencl code  Grin

Haha, yeah, it's hard to read... but it's a lot faster than it was!

I can run cgminer 3.7.8 neoscrypt on a 550ti at a rate of 25kh/s (Win64).  However, I cannot run cgminer 3.7.8 neoscrypt on a 6-card 750ti rig, Ubuntu 14.04.1. I can compile it, but it won't run properly, it just spits HW errors.

My 750ti rigs run smoothly with compiled ccminer 1.5.2-git code.  Are there any special command-line switches to use during compilation to make cgminer 3.7.8 or sgminer 5 work properly on nVidia Linux rigs?  The code compiles, I am pretty sure that I have the pre-requisites in place.

Personally, I think a 750ti could likely hash on a level of a 280x if the cuda code is right.       --scryptr

agreed - it would run very well if it was coded right ... if only we had the fine grained control that sgminer does with ccminer and cuda ...

as a side issue - i run sgminer/neoscrypt on the 6 card 750ti with out issue - but the hashrates are poor ... so i stick with ccminer/x11 ...

#crysx
hero member
Activity: 675
Merit: 514
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
I shouldn't have started from Wolf0 opencl code  Grin
Don't you like 16-dimensional vectors?  Grin
legendary
Activity: 1797
Merit: 1028
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
I shouldn't have started from Wolf0 opencl code  Grin

Haha, yeah, it's hard to read... but it's a lot faster than it was!

I can run cgminer 3.7.8 neoscrypt on a 550ti at a rate of 25kh/s (Win64).  However, I cannot run cgminer 3.7.8 neoscrypt on a 6-card 750ti rig, Ubuntu 14.04.1. I can compile it, but it won't run properly, it just spits HW errors.

My 750ti rigs run smoothly with compiled ccminer 1.5.2-git code.  Are there any special command-line switches to use during compilation to make cgminer 3.7.8 or sgminer 5 work properly on nVidia Linux rigs?  The code compiles, I am pretty sure that I have the pre-requisites in place.

Personally, I think a 750ti could likely hash on a level of a 280x if the cuda code is right.       --scryptr
legendary
Activity: 1400
Merit: 1050
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
I shouldn't have started from Wolf0 opencl code  Grin
legendary
Activity: 3164
Merit: 1003
I cant find , do we have neoscrypt available for ccminer ?
if it was already there, you would have seen it  Grin
(working on it though)
Keep us informed.   Grin
legendary
Activity: 3164
Merit: 1003
New exe file to test. Compiled with compute 30,35,50,52

-faster lyra 12% (Most of the improvement done by Klaus_t)
-faster quark 10%
-fixed HW errors in x11
-fixed compute 3.0 cards.(untested.)

x11, x13, x15 a little bit faster. (Optimized Skein)

The 21th public beta(windows executable) is available here: (21-12-2014)

http://www.filedropper.com/ccminer21

The sourcecode is available here:

https://github.com/sp-hash/ccminer

Works great sp  Wink
Jump to: