CCminer(SP-MOD) Modded GPU kernels.

[sp_]

Had the same issue, reverted to 0.9.4.1 (from cryptominingblog) which is a bit slower but submit shares.
Regarding 750 Ti's, I get like 300khs every 5th or 6th lines amd the rest is just zeros.

Did you try the opencl version with adjustable blocksize?

https://github.com/Genoil/cpp-ethereum/tree/opencl-chunks

[1714764060]

1714764060

[1714764060]

1714764060

[1714764060]

1714764060

[1714764060]

1714764060

[1714764060]

1714764060

[1714764060]

1714764060

[Genoil]

Had the same issue, reverted to 0.9.4.1 (from cryptominingblog) which is a bit slower but submit shares.
Regarding 750 Ti's, I get like 300khs every 5th or 6th lines amd the rest is just zeros.

Did you try the opencl version with adjustable blocksize?

https://github.com/Genoil/cpp-ethereum/tree/opencl-chunks

Chunks version does work on NVidia (not on AMD!), but AFAIK it doesn't do anything to avert TLB trashing.

I wonder what's the difference between these versions. I also get reports from people using a build posted on cryptomining who can't submit shares. If I'm correct, there are actually two different builds posted on Cryptomining. I'll try them both.

To rule out a problem with the algo, try running the miner for a while using the -S flag instead of -F. It simulates hashing at increasing difficulty until a block time of 15 seconds is reached. Using a GTX9x0, you should be able to get to a difficulty of about 30 (1 << 30) in a relatively short amount of time. If you get stuck at 20 or 25, the algo is most likely broken. Don't be fooled if it makes your 750ti tick hash faster on Win7/Linux than it does now as it uses the epoch 0 DAG.

[pallas]

SP will you be incorporating the new vanilla blake algo into ccminer?  If so, when?

Blake is already included. ccminer -a blake --benchmark or ccminer -a blakecoin --benchmark

I don't think it'll work.
You should import a little patch which was submitted to tpruvot ccminer a couple days ago.

EDIT:
"Algo is the same Blake256 8-rounds. It's just different from Blakecoin who's using single sha256 for the merkle root generation. Vanilla will use standard double sha256."

[lemp]

Samsung and Elpida give 20-22MH/s, hynix 17-19MH/s in windows 7. I am not sure about windows 8.1.

You need to get your card to P0 state:

1) open cmd.exe with the admin privileges
2) cd C:\Program Files\NVIDIA Corporation\NVSMI
3) nvidia-smi -q -d SUPPORTED_CLOCKS -i 0
4) nvidia-smi -ac 350X,14XX -i 0 (where 350X is that very first number printed after supported clocks-memory and 14XX one of those numbers printed after graphics)

5) Overclock your memory, +500 generally works. Should give you memory speed of 4005MHz
6) Overclock your core, doesn't make big difference mining ETH.
7) If you wan't to save power, reduce your power limit. Even 60% works suprisingly well.

Someone should open new topic for nvidia and ethereum... 

Thanks, now my gtx970 card gives ~21Mh
Overclock: nvidia-smi -ac 3505,1505
Startup parametrs eth miner: ethminer --cuda-grid-size 8192 --cuda-block-size 128 --cuda-turbo

[Genoil]

Perhaps compile the ether miner for 32 bit's will help? Cached Pointersizes will go from 64bit to 32 (and double the tlb limit?) You need to remove the cpu verfication code because it use 64bit libraries I think..

Thought of that but it's going to be troublesome. You only have a 4GB address space, with windows already sucking up ~half. Then you have to load the 1.3 GB DAG from disk, and allocate 1.3GB of GPU RAM (which, AFAIK sits in the same space, although it isn't pinned to host). This doens't fit. So then you would have to read the DAG from disk in small chunks and copy it cover to GPU RAM. And when that's all done, you will have to pass on all solutions to a special light version of ethminer, that does light verification, is it can't load a DAG into RAM for the same reasons. Or you simply don't verify and risk some Boo's.

Then, when that's all done, you're not even sure if it fixes the problem. You could try getting a 32-bit version of dagSimCL to work.

I believe Epsylon3/tpruvot has been trying to get a 32-bit version of ethminer to work a while back. Can't find the source anymore.

[djm34]

Perhaps compile the ether miner for 32 bit's will help? Cached Pointersizes will go from 64bit to 32 (and double the tlb limit?) You need to remove the cpu verfication code because it use 64bit libraries I think..

Thought of that but it's going to be troublesome. You only have a 4GB address space, with windows already sucking up ~half. Then you have to load the 1.3 GB DAG from disk, and allocate 1.3GB of GPU RAM (which, AFAIK sits in the same space, although it isn't pinned to host). This doens't fit. So then you would have to read the DAG from disk in small chunks and copy it cover to GPU RAM. And when that's all done, you will have to pass on all solutions to a special light version of ethminer, that does light verification, is it can't load a DAG into RAM for the same reasons. Or you simply don't verify and risk some Boo's.

Then, when that's all done, you're not even sure if it fixes the problem. You could try getting a 32-bit version of dagSimCL to work.

I believe Epsylon3/tpruvot has been trying to get a 32-bit version of ethminer to work a while back. Can't find the source anymore.

I don't remember it was released either (I remember that I tested it though...) I think it was on a dropbox or something , don't think he created a repo (?)

[sp_]

I believe Epsylon3/tpruvot has been trying to get a 32-bit version of ethminer to work a while back. Can't find the source anymore.

I tried as well to integrate the ether miner into ccminer.. But lost interest because the 64bit nonce and other non standard code. I can do another try and release the result.
NO cpu verification or DAG generation. Only to test and mod the gpu code.. (--benchmark mode)

[sp_]

I don't remember it was released either (I remember that I tested it though...) I think it was on a dropbox or something , don't think he created a repo (?)

I found the code on github:

https://github.com/tpruvot/ccminer/tree/ethash

ethash test to finish...
remains the mixHash thing to compute, and to check the submitted nonce64 endian/value

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>

[pokeytex]

SP will you be incorporating the new vanilla blake algo into ccminer?  If so, when?

Blake is already included. ccminer -a blake --benchmark or ccminer -a blakecoin --benchmark

No - they have modified the file somehow to be vanilla-blake.

xCore - ccminer GTX9xx/750 (based on last sp-hash's github): https://talk.vanillacoin.net/uploads/files/tools/ccminer-1.7.1-blake256-standard-uncompressed-sp.zip

-pokeytex

[sp_]

ok I added it now. You have to build yourself..

The gtx 970 is doing 2.3GHASH
Run with -a vanilla

[pokeytex]

ok I added it now. You have to build yourself..

The gtx 970 is doing 2.3GHASH
Run with -a vanilla

Thank you - what is the software used to build from source again?

[sp_]

1. Download cuda 7.5
2. Download and install visual studio 2013 express
3. Download the source: https://github.com/sp-hash/ccminer/archive/windows.zip
3. Open the project. select release, win32
4. Build.

[sp_]

ok I added it now. You have to build yourself..

The gtx 970 is doing 2.3GHASH
Run with -a vanilla

Gtx 970: 2.6GHASH
Gtx 960: 1.65GHASH
gtx 750ti: 0.94GHASH

[sp_]

xCore - ccminer GTX9xx/750 (based on last sp-hash's github): https://talk.vanillacoin.net/uploads/files/tools/ccminer-1.7.1-blake256-standard-uncompressed-sp.zip

Version 1.7.1 is not my fork. My version is 1.5.78

[crsminer]

Perhaps compile the ether miner for 32 bit's will help? Cached Pointersizes will go from 64bit to 32 (and double the tlb limit?) You need to remove the cpu verfication code because it use 64bit libraries I think..

Thought of that but it's going to be troublesome. You only have a 4GB address space, with windows already sucking up ~half. Then you have to load the 1.3 GB DAG from disk, and allocate 1.3GB of GPU RAM (which, AFAIK sits in the same space, although it isn't pinned to host). This doens't fit. So then you would have to read the DAG from disk in small chunks and copy it cover to GPU RAM. And when that's all done, you will have to pass on all solutions to a special light version of ethminer, that does light verification, is it can't load a DAG into RAM for the same reasons. Or you simply don't verify and risk some Boo's.

Then, when that's all done, you're not even sure if it fixes the problem. You could try getting a 32-bit version of dagSimCL to work.

I believe Epsylon3/tpruvot has been trying to get a 32-bit version of ethminer to work a while back. Can't find the source anymore.

Hi Genoil,

This is a small contribution from me to your work on ethminer --- 0xc3e7bda79b60fb4f34fbe467a8c3c3084e9008bbdc9014d8ade8e8c22cd1352a

I am using your miner and making some ETH and I thought that you deserve part of it (as I did with all devs improving the miners we are using).

Keep up the good work!

[kangjooe]

xCore - ccminer GTX9xx/750 (based on last sp-hash's github): https://talk.vanillacoin.net/uploads/files/tools/ccminer-1.7.1-blake256-standard-uncompressed-sp.zip

Version 1.7.1 is not my fork. My version is 1.5.78

ask:

why nicehash accepted share is always lower than the reported? for example, Lyra2 is accepted around 9600~ kH/s but nicehash only accepted 8.9 mH/s, no matter how long i mine, it always lower than my mining speed.  

and what is the best intensity for lyra2 in 78 release? i'm using 970.

[sp_]

xCore - ccminer GTX9xx/750 (based on last sp-hash's github): https://talk.vanillacoin.net/uploads/files/tools/ccminer-1.7.1-blake256-standard-uncompressed-sp.zip

Version 1.7.1 is not my fork. My version is 1.5.78

ask:
why nicehash accepted share is always lower than the reported? for example, Lyra2 is accepted around 9600~ kH/s but nicehash only accepted 8.9 mH/s, no matter how long i mine, it always lower than my mining speed.  
and what is the best intensity for lyra2 in 78 release? i'm using 970.

Wich version of ccminer are you using?

You should mine lyra2v2 and not lyra2.

[sp_]

xCore - ccminer GTX9xx/750 (based on last sp-hash's github): https://talk.vanillacoin.net/uploads/files/tools/ccminer-1.7.1-blake256-standard-uncompressed-sp.zip

Version 1.7.1 is not my fork. My version is 1.5.78

ask:
why nicehash accepted share is always lower than the reported? for example, Lyra2 is accepted around 9600~ kH/s but nicehash only accepted 8.9 mH/s, no matter how long i mine, it always lower than my mining speed.  
and what is the best intensity for lyra2 in 78 release? i'm using 970.

Wich version of ccminer are you using?
You should mine lyra2v2 and not lyra2.

Download version 74 here:

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

[kangjooe]

xCore - ccminer GTX9xx/750 (based on last sp-hash's github): https://talk.vanillacoin.net/uploads/files/tools/ccminer-1.7.1-blake256-standard-uncompressed-sp.zip

Version 1.7.1 is not my fork. My version is 1.5.78

ask:
why nicehash accepted share is always lower than the reported? for example, Lyra2 is accepted around 9600~ kH/s but nicehash only accepted 8.9 mH/s, no matter how long i mine, it always lower than my mining speed.  
and what is the best intensity for lyra2 in 78 release? i'm using 970.

Wich version of ccminer are you using?

You should mine lyra2v2 and not lyra2.

i'm using your version. 1.5.78


this is my bat:

Code:

ccminer.exe -a lyra2v2 -o stratum+tcp://lyra2rev2.jp.nicehash.com:3347 -u 1AwFc6m5evP5a8csDWsb6S3joSaBx8ia1e -p d=4 -i 16

is 1.5.74 give better hash rate for lyrav2 than 78?

[sp_]

xCore - ccminer GTX9xx/750 (based on last sp-hash's github): https://talk.vanillacoin.net/uploads/files/tools/ccminer-1.7.1-blake256-standard-uncompressed-sp.zip

Version 1.7.1 is not my fork. My version is 1.5.78

ask:
why nicehash accepted share is always lower than the reported? for example, Lyra2 is accepted around 9600~ kH/s but nicehash only accepted 8.9 mH/s, no matter how long i mine, it always lower than my mining speed.  
and what is the best intensity for lyra2 in 78 release? i'm using 970.

Wich version of ccminer are you using?
You should mine lyra2v2 and not lyra2.

i'm using your version. 1.5.78
this is my bat:

Code:

ccminer.exe -a lyra2v2 -o stratum+tcp://lyra2rev2.jp.nicehash.com:3347 -u 1AwFc6m5evP5a8csDWsb6S3joSaBx8ia1e -p d=4 -i 16

is 1.5.74 give better hash rate for lyrav2 than 78?

Replace d=4 with x. and remove the intensity setting.

ccminer.exe -a lyra2v2 -o stratum+tcp://lyra2rev2.jp.nicehash.com:3347 -u 1AwFc6m5evP5a8csDWsb6S3joSaBx8ia1e -p x

You should get full hashrate.

Release 78 was compiled with cuda 7.5 and I think it is slower than release 74 but not sure. If you have time, try all releases from 68 and up to 78 and check wich version has the best hashrate for your card.

I also have a private version that improves the speed alot. (from release 74) 4-10%  (it cost 0.1 BTC in donation)