Author

Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1221. (Read 2347597 times)

legendary
Activity: 2716
Merit: 1116
@jpouza I set -i 20 and managed to get the figures above.

The default is set to -i=19 because -i=20 will cause out of memory on the 1GB cards (750,750ti) On 970/980 -i should be set to 20 or higher.

-i 21 on 980 boosted almost 200KH/s!

-i 20 on 750Ti 2GB nothing changed, will try to increase memory clock.

EDIT: nothing, still can't reach 3MH/s with 750 Ti, even with 1600MHz mem clock and -i 20.
980 just amazing, 9,8 MH/s achieved with -i 21

WOW


subefotos
legendary
Activity: 1484
Merit: 1082
ccminer/cpuminer developer

http://pastebin.com/sV7C2TWR

Try that for Hamsi. It sure is not pretty but it's a whole lot faster than the one I originally left X13 with. My notes say +59% on the 750 Ti Tongue

Yep, i get much more with your code Smiley i keep it tx.

2250KHs on a 750 Ti, vs 2150 before changes, and slower with sp commit (tried differents maxregcount on linux)

edit: i used 72 registers, 64 is too low
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
@jpouza I set -i 20 and managed to get the figures above.

The default is set to -i=19 because -i=20 will cause out of memory on the 1GB cards (750,750ti) On 970/980 -i should be set to 20 or higher.
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
Commited tsivs hamsi, and removed memoryleaks in x11 and x13. Also merged in tvprovot's stratum bugfix.
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
Yes, faster indeed! Thanks, I will add it.
Tested it to be 14.34% faster than my last version. (750ti)

Actually it was even faster with 64 registers But I guess your default already was 64 regs..
My default is 80, so I had to add  launchbounds to force 64 registers, and running only with 32 threads.

Then I get it to be 20% faster.
sr. member
Activity: 292
Merit: 250
Release 10 works perfectly. No problems whatsoever.

Hashrates (750Ti O/C)
X11: 3000 kH/s
X13: 2320 kH/s
X15: 2000 kH/s

What are your clocks with the 750Ti?

I can get only 2950-2970KH/s with 1246MHz Core / 1580MHz Memory

1210MHz Core / 1600MHz Memory

@jpouza I set -i 20 and managed to get the figures above.
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
Release 10 works perfectly. No problems whatsoever.
Hashrates (750Ti O/C)
X11: 3000 kH/s
X13: 2320 kH/s
X15: 2000 kH/s
Congratulations, sp_. You have beaten the code I did for CUDA... I think a couple months ago now, on X11. I think all three.

You should help me push it up to 3,5 MHASH. Still some easy pickings.
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
what is the range of intensity parameter we can set?

try. max is 30, but the miner will probobly crash and spit out wild hashes. on the 970 20 or 21 @x11?

Default is 19
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
New build from github
http://www.filedropper.com/release10
Contains 2 bugfixes from the tpruvot (stratum disconnects)
X13 and x15 is a bit faster. (hamsi+8,3%)
fixed linuxbuild (untested)
The sourcecode is available here:
https://github.com/sp-hash/ccminer
http://pastebin.com/sV7C2TWR
Try that for Hamsi. It sure is not pretty but it's a whole lot faster than the one I originally left X13 with. My notes say +59% on the 750 Ti Tongue

Yes, faster indeed! Thanks, I will add it.
Tested it to be 14.34% faster than my last version. (750ti)
sr. member
Activity: 285
Merit: 250
what is the range of intensity parameter we can set?
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
ccminer12 is for compute 5.2 cards.  970/980. Since this is just a build from the github it doesn't contain the dependency files. You can download the files in the latest tvpovot release.
sr. member
Activity: 305
Merit: 250
Hey sp_ quick question.

I've just downloaded your Release10 (my first download of yours).
What is the difference between ccminer and ccminer12?

Also it says that its missing msvcr120.dll

I managed to copy this from tpruvot.

Is this correct?
Or do you have a better one to use?
full member
Activity: 137
Merit: 100
New build from github

http://www.filedropper.com/release10

Contains 2 bugfixes from the tpruvot (stratum disconnects)
X13 and x15 is a bit faster. (hamsi+8,3%)
fixed linuxbuild (untested)

The sourcecode is available here:

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

http://pastebin.com/sV7C2TWR

Try that for Hamsi. It sure is not pretty but it's a whole lot faster than the one I originally left X13 with. My notes say +59% on the 750 Ti Tongue
sr. member
Activity: 292
Merit: 250
Release 10 works perfectly. No problems whatsoever.

Hashrates (750Ti O/C)
X11: 3000 kH/s
X13: 2320 kH/s
X15: 2000 kH/s

What are your clocks with the 750Ti?

I can get only 2950-2970KH/s with 1246MHz Core / 1580MHz Memory

1210MHz Core / 1600MHz Memory
legendary
Activity: 1484
Merit: 1082
ccminer/cpuminer developer
i found another side effect of the extranonce subscribe (its the possible 0khs accept on start)...

Stratum Request id was not tested on share answers (it was tested nowhere in fact before my recent changes)

Just commited

Code:
[2014-11-24 19:34:42] < {"id":3,"result":true,"error":null}
[2014-11-24 19:34:42] accepted: 1/1 (100.00%), 0.00 khash/s yay!!!
legendary
Activity: 2716
Merit: 1116
Release 10 works perfectly. No problems whatsoever.

Hashrates (750Ti O/C)
X11: 3000 kH/s
X13: 2320 kH/s
X15: 2000 kH/s

What are your clocks with the 750Ti?

I can get only 2950-2970KH/s with 1246MHz Core / 1580MHz Memory
sr. member
Activity: 292
Merit: 250
Release 10 works perfectly. No problems whatsoever.

Hashrates (750Ti O/C)
X11: 3000 kH/s
X13: 2320 kH/s
X15: 2000 kH/s
sr. member
Activity: 305
Merit: 250
Thanks sp_, I was just LITERALLY asking for a new build!

You rock!
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
New build from github

http://www.filedropper.com/release10

Contains 2 bugfixes from the tpruvot (stratum disconnects)
X13 and x15 is a bit faster. (hamsi+8,3%)
fixed linuxbuild (untested)

The sourcecode is available here:

https://github.com/sp-hash/ccminer
sp_
legendary
Activity: 2926
Merit: 1087
Team Black developer
thanks. I checked in the fix.
I will checkin some more kernal improvements (x13).
Jump to: