Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1210. (Read 2347597 times)

Hi SP.

I'm mining X13 on regular pool (not multipool).The hashrate showing by the pool is lower than what the ccminer shows. And there is "set stratum difficulty to x.xxxx" message appearing quite a lot.
I tried versions from 17-20. The same scene. Hashrate is lower (from 14mhs on ccminer to only 8-9mhs on pool). I then tried older version 8 and it works better (13mhs miner 12mhs pool) I have standard setting in .bat file. Do i need a manual set up of diff per rig or is there something else that I'm missing? Switching back to version 8 now...
I would appreciate any help.
Thank you.

how much the last version, does on scrypt algo?

i've checked the source now, there isn't any scrypt algo available, you removed it?

p.s. ah silly me, scypt is in cudaminer, i always forgot that lol, mmh...old cudaminer isn't working right with 970, not optimized i supposed, it use T91

Yes this version is bether on the pools when mining on high intensity. The old version only found one solution below target, this version is submitting 2 solutions. Code forked fronm 1.5.1 (tvprovot's work)

Wow, average hash reported by the pool improved, from 30MH/s to 35-40MH/s.

Nice.

Updated on github. Tried to fix 3.0, but changing the methods in cudahelper.h doesn't work. Must be something else. Will fix it later. Can anyone test on compute 3.5 please? how is the hash on 780ti?

Build20 running OK for now, nice be just one instance for mining, all in one

In X11:

3.2MH/s 750Ti@200+core/400+mems

9.6MH/s 980@150+core/0+ mems

980gtx temps at 55ºC@ EK WC.

Nice.

Cheers

Can you also update the version, as reported by the API? My tool starts a new benchmark if it detects a new version, but it didn't now
Makes it easy to show differences between versions and authors.

Also, GTX760 (compute 3.0) doesn't work: http://snag.gy/2Vo9h.jpg

Been test hashing on build20 on gtx650. Seems to be some frequent HW errors(compute 3.0). Will look into it.

Version 20 is here (windows executable):

Faster x11 (+100KHASH(750ti))
faster x13,x14,x15,x17
fixed hashing for old cards.

http://www.filedropper.com/ccminer_1

source

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


One exe this time. Compiled for compute30,32,35,50,52


cards supported:

GeForce GTX 770, GeForce GTX 760, GeForce GT 740, GeForce GTX 690, GeForce GTX 680, GeForce GTX 670, GeForce GTX 660 Ti, GeForce GTX 660, GeForce GTX 650 Ti BOOST, GeForce GTX 650 Ti, GeForce GTX 650, GeForce GTX 780M, GeForce GTX 770M, GeForce GTX 765M, GeForce GTX 760M, GeForce GTX 680MX, GeForce GTX 680M, GeForce GTX 675MX, GeForce GTX 670MX, GeForce GTX 660M, GeForce GT 750M, GeForce GT 650M, GeForce GT 745M, GeForce GT 645M, GeForce GT 740M, GeForce GT 730M, GeForce GT 640M, GeForce GT 640M LE, GeForce GT 735M, GeForce GT 730M, Quadro K5000, Quadro K4200, Quadro K4000, Quadro K2000, Quadro K2000D, Quadro K600, Quadro K420, Quadro K500M, Quadro K510M, Quadro K610M, Quadro K1000M, Quadro K2000M, Quadro K1100M, Quadro K2100M, Quadro K3000M, Quadro K3100M, Quadro K4000M, Quadro K5000M, Quadro K4100M, Quadro K5100M, Tesla K10
GeForce GTX TITAN Z, GeForce GTX TITAN Black, GeForce GTX TITAN, GeForce GTX 780 Ti, GeForce GTX 780, GeForce GT 640 (GDDR5), GeForce GT 630 v2, Quadro K6000, Quadro K5200, Tesla K40, Tesla K20x, Tesla K20
Tesla K80 GeForce GTX 750 Ti, GeForce GTX 750, GeForce GTX 860M, GeForce GTX 850M, GeForce 845M, GeForce 840M, GeForce 830M, Quadro K2200, Quadro K620, Quadro K620M GeForce GTX 980, GeForce GTX 970, GeForce GTX 980M, GeForce GTX 970M

Could you try to write once my name correctly ? tx

thanks.

The head on Gtihub only works for compute 3.5 cards or newer.

http://en.wikipedia.org/wiki/CUDA

If you have 3.0 you can use the tpruvot version:

https://bitcointalksearch.org/topic/ann-ccminer-23-opensource-gpl-tpruvot-770064

yes on linux.
GTX 680 SOC which is unfortunately not compute 5

My version is probobly failing on sm30 cards.

Rewrote skein to use uint2. 10-20 KHASH on x11,x13,x15.  Added to github.

https://github.com/sp-hash/ccminer/commit/f5c7cc861fb4ed62e473437813f4f1a637e672e1

I see that this has also been done in the skein256 (lyra) but the code is different. Is the new implementation faster DJM34?

I had this while testing keccak uint2 changes with a sm30 only binary on the 750 ti... No errors but no found nonces neither.. So i commited the optimisation differently

Works here. Is this on linux? wich card?

I just submitted an improved BMW rewrite.

32% faster. Using UINT2 instead of Uint64_t. Added new operators for uint2 operations. SHL2 and SHR2, and minus. (in cuda_helper.h)

Around +30-40 KHASH  on x11, x13 and x15

Now BMW is spilling 0 bytes to slow memory...

I also tried to use uint2 in blake, but it was slower.

tested with x11 and x13. i see a big hashrate increase, tough never any submitted shares.

it is anti-asic for now...