CCminer(SP-MOD) Modded GPU kernels.

[krnlx]

I am modding as well.

1070: 100sol/s (default clocks) With -500 on the memclock I get 110
1060 6gb 75sol/s(default clocks)
gtx 970 70sol/s(default clocks)

I use

#define THRD 128
#define BLK (NR_ROWS/THRD)

in param.h

Did you check last changes? I updated last eXtremal's code

[Amph]

360 sols for 4x1070 g1 with no OC ..

i'm at 106 with 1070 g1 gaming, you should definitely get more try to play with msi afterburner

for the g1 gaming -502 mem 0 core and 55% tdp work great, low power while still a nice hashrate, good compromise for me

what´s wrong?? this is what i get...

sube

strange you should get 400+, what drivers? OS seems windows 7 i'm on win 10, it may be a factor

[_javi_]

amph:
win 7 64, 372.54 drivers

btw, i always wondered about those GPU usage and power oscillating patterns.. are those normal?

[Amph]

amph:
win 7 64, 372.54 drivers

btw, i always wondered about those GPU usage and power oscillating patterns.. are those normal?

372.90 here, i think yes, code is not optimized it may be also for pool side problem, mine do that as well

just checked again, and no, my gpu sit there at 98% no swings in usage, so it perhaps the problem is in your side only

[sp_]

Did you check last changes? I updated last eXtremal's code

Doesn't seem to work on the 750ti. (compute 5.0)

[_javi_]

Maybe i should try win10..
and it´s a poor old Sempron145 doing its best 

[krnlx]

Did you check last changes? I updated last eXtremal's code

Doesn't seem to work on the 750ti. (compute 5.0)

eXtremal's last code has a lot of hardcoded 64 threads...

[ioglnx]

Lovely to have 450+ Sol/s

Code:

[14:40:08][0x0000105c] Speed [300 sec]: 245.11 I/s, 454.043 Sols/s
[14:40:09][0x0000312c] stratum | Submitting share #153, nonce 000000000000000000000000020000000000000000000000000000c4
[14:40:09][0x00003a70] stratum | Accepted share #153
[14:40:12][0x0000312c] stratum | Submitting share #154, nonce 000000000000000000000000020000000000000000000000000000c9
[14:40:12][0x00003a70] stratum | Accepted share #154
[14:40:19][0x0000105c] Speed [300 sec]: 245.057 I/s, 453.9 Sols/s
[14:40:31][0x0000105c] Speed [300 sec]: 245.087 I/s, 454.507 Sols/s
[14:40:38][0x00001df4] stratum | Submitting share #155, nonce 0000000000000000000000000f000000000000000000000000002176
[14:40:38][0x00003a70] stratum | Accepted share #155
[14:40:42][0x0000105c] Speed [300 sec]: 244.547 I/s, 453.423 Sols/s
[14:40:46][0x0000298c] stratum | Submitting share #156, nonce 0000000000000000000000000d000000000000000000000000002352
[14:40:46][0x00003a70] stratum | Accepted share #156
[14:40:53][0x0000105c] Speed [300 sec]: 244.07 I/s, 452.197 Sols/s
[14:40:57][0x00001460] stratum | Submitting share #157, nonce 0000000000000000000000000a000000000000000000000000000107
[14:40:57][0x00003a70] stratum | Accepted share #157
[14:41:04][0x00003e14] stratum | Submitting share #158, nonce 0000000000000000000000000e0000000000000000000000000024ca
[14:41:04][0x00003a70] stratum | Accepted share #158
[14:41:05][0x0000105c] Speed [300 sec]: 245 I/s, 453.563 Sols/s
[14:41:07][0x00003918] stratum | Submitting share #159, nonce 00000000000000000000000006000000000000000000000000000124
[14:41:08][0x00003a70] stratum | Accepted share #159
[14:41:16][0x0000105c] Speed [300 sec]: 244.92 I/s, 453.047 Sols/s
[14:41:17][0x00001df4] stratum | Submitting share #160, nonce 0000000000000000000000000f000000000000000000000000002a71
[14:41:17][0x00003a70] stratum | Accepted share #160

[_javi_]

i´m getting 450 sols for 4x1070 but on the hardcoded nicehash miner 

[sp_]

Did you check last changes? I updated last eXtremal's code

Doesn't seem to work on the 750ti. (compute 5.0)

eXtremal's last code has a lot of hardcoded 64 threads...

Yes. To make it work on compute 5.0 I had to reduce this to

(NR_SLOTS * 24 * (THRD / THREADS_PER_ROW)

to

(NR_SLOTS * 19 * (THRD / THREADS_PER_ROW) or lower..

With this

#define LDS_COLL_SIZE (NR_SLOTS * 15 * (THRD / THREADS_PER_ROW))

I get 26 sol/s

[Amph]

after the miner work for some time, there is gpu throttling and the hashrate begin to decline a bit

if i restart it work good again and then it do the same thing after 5-10 min

somethign wrong in the code maybe? with zawawa version there was not such issue

[ioglnx]

Should I do a build for each CM?
Can define that for automatic builds on my TFS

[sp_]

Should I do a build for each CM?
Can define that for automatic builds on my TFS

Just insert this code:

Code:

#if __CUDA_ARCH__ = 500
#define LDS_COLL_SIZE (NR_SLOTS * 15 * (THRD / THREADS_PER_ROW))
#else
#define LDS_COLL_SIZE (NR_SLOTS * 24 * (THRD / THREADS_PER_ROW))
#endif

Then the compiler will fix the correct coll_size when you build for compute_50

[sp_]

I get 33 sol/s now on the 750ti on standard clocks. So I guess 37-40 is possible with oc

[scryptr]

I get 33 sol/s now on the 750ti on standard clocks. So I guess 37-40 is possible with oc

MRB'S OCL CODE GETS 25 SOLS/S ON A 750TI--

If your build is getting 33 Sols/s on a 750ti, then you have finally CUDA optimized the code.  The code from mrb, release 5, compiles nicely on my linux GTX 750ti rig and my cards get 23-27Sols/s.  CPU mining is disabled; the rig has a Sempron 145 CPU.

However, on my Win 7 x64 box, my GTX 980ti + i7 2600 CPU mine at 90+ Sols/s with the NiceHash hard-coded "EQM" miner.  This is about 20 Sols/s faster than KlausT's Windows release 5.1.  If anyone is looking for pre-compiled  binaries, KlausT's GitHub has a good Windows release.  His work is Free Open-Source Software (FOSS), and not locked to the NiceHash pool, of course.

Optimizing FOSS code for a 750ti mining on a Sempron rig is a good way to measure CUDA optimization.  The mining Sols/s are not enhanced by CPU mining results.

I hope to be able to use some polished CUDA code from your work, sp_.  Keep working on it...       --scryptr

[sp_]

I will release #1 today for free, so you can test it..

[Amph]

ioglnx with your windows port, the consumption is way too high, i'm at 640 sol but 800 watt, not really that efficient, something not optimized well in the code i think

sp what is your consumption with your modded version?

also too much throttling

[bensam1231]

ZC is advertising 108sol on 1070. Tried it and only got 97. Still not worth it compared to Ethereum mining, but still there.

https://forum.z.cash/t/cuda-equihash-miner-with-devfee-108sol-s-on-gtx1070

*sigh*

Nicehash has a new miner out as well advertising 125sol on 1070, but it's locked to Nicehash so pretty much worthless.

https://forum.z.cash/t/eqm-gtx-1070-125-sol-s/8574

[sp_]

sp-mod #1 is best on the maxwell cards.

standard clocks..

gtx 970: 77
gtx 750ti:33
gtx 1060 3gb: 73
gtx 1070: 100

[sp_]

here is the build

Zcash sp-mod #1

https://github.com/sp-hash/ccminer/releases/tag/zcash-spmod1