To write the fastest kernels in the world you can't use the NVIDIA compliler
use this:
https://github.com/NervanaSystems/maxas
Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 428. (Read 2347659 times)
I didn't get even a single donation for the faster pascal solo miner (on which the sp-mod is based). I see that nothing changed in this front :-/
How much do you have with the gtx 1070 with your private kernel? Wiithout any power limits
Wich kernel? I have a big private farm.
I hope to have more than 1000 gpu's by the end of 2017. All my profit is invested into new hardware.
can you reach 5GH on a 1070? it would be really profitable currently
If I opensource, nothing will be profitable.
can you reach 5GH on a 1070? it would be really profitable currently
If I opensource, nothing will be profitable.
Decred is slow on pascal.
should be 6GHASH on the gtx 1080
Here is the private maxwell assembly version:
should be 6GHASH on the gtx 1080
Here is the private maxwell assembly version:
can you reach 5GH on a 1070? it would be really profitable currently
so my question is: why not simply copy and modify the sha256 part in ccminer to fit the changes made and maybe adjust the stratum to some new specs for suprnova stratum (if there are any)?
Because the sha256 in ccminer is not optimized.
The Pascalcoin seem to have a variable length block header. A 1-2 day job to fix. Klaus-t will do it for free.
1. it's crashing after some seconds
2. the hash is wrong
Screenshot
I can't seem to find the bug.
The stratum part for suprnova.cc should be ok, I think, but I can't properly test it right now.
Decred is slow on pascal.
should be 6GHASH on the gtx 1080
Here is the private maxwell assembly version:
should be 6GHASH on the gtx 1080
Here is the private maxwell assembly version:
so my question is: why not simply copy and modify the sha256 part in ccminer to fit the changes made and maybe adjust the stratum to some new specs for suprnova stratum (if there are any)?
Because the opensource sha256 in ccminer is not optimized.
The Pascalcoin seem to have a variable length block header. A 1-2 day job to fix. Klaus-t will do it for free.
many times i have read that pascal algo is basically sha256 with some minor modifications to prevent asic usage
so my question is: why not simply copy and modify the sha256 part in ccminer to fit the changes made and maybe adjust the stratum to some new specs for suprnova stratum (if there are any)?
im not an expert in those things, could someone clarify as it seems its not that simple or im missing something?
so my question is: why not simply copy and modify the sha256 part in ccminer to fit the changes made and maybe adjust the stratum to some new specs for suprnova stratum (if there are any)?
im not an expert in those things, could someone clarify as it seems its not that simple or im missing something?
I see that klaus_t has fixed some more bugs in his fork. So perhaps we can get a working stratum implementation for NVIDIA soon..
If klaust fix his fork I will merge my kernel into ccminer and optimize more.
If klaust fix his fork I will merge my kernel into ccminer and optimize more.
Pascal wallet is garbage in 5 days it crashed or stuck five times, or java proxy killed itself
cuda miner was stable but....
Mined with 25G+ for 3-4 days zero block, just wasted a lot of time on that crap
cuda miner was stable but....
Mined with 25G+ for 3-4 days zero block, just wasted a lot of time on that crap
617mhz with a gtx 1060 6gb :|
I get 617 on the gtx 1060 3gb
Very nice, now we just need to connect to the pool =S
1236 MH/s report running 1card_2threads.bat
That means 2Gh/s?
1236 MH/s report running 1card_2threads.bat
That means 2Gh/s?
Run it for a while and the speed will drop. Or perhaps the one of the threads has died. This is a beta. abit unstable. Sometimes crash on exit.
617mhz with a gtx 1060 6gb :|
Very nice, now we just need to connect to the pool =S
1236 MH/s report running 1card_2threads.bat
That means 2Gh/s?
1236 MH/s report running 1card_2threads.bat
That means 2Gh/s?
Ok, here is pascalcoin sp-mod #2 (SOLO miner)
enjoy
https://github.com/sp-hash/ccminer/releases/tag/pascal-spmod2
enjoy
https://github.com/sp-hash/ccminer/releases/tag/pascal-spmod2
Team, sorry to repost this in another thread, not getting much traction in the other one. Has anyone had success getting 6 nvidia gpus working with 2 threads each, if so can you post miner, software/driver ver and config. I am using 6 asus gtx 1060 3gb cards and when trying sgminer, i have to set thread to 1 so it doesnt error out on config. I have tried ccminer SP-MOD but cannot remember the error, I am at work now.
The only way to run the solominer in 2 threads is by using the sp-mod. You need to set up 12 gpu's in the proxy..
I found a block in the pascal lite with the sp-mod #2
Based on pallas opensource.
I apologize for my noobness, but how do I get the sp-mod, is it not part of the download from the beginning of this thread? Is there reading material so I can get mine setup correctly?
Thanks!
found it....
but getting that error Invalid new block 187: Proof of work is higher than target
but getting that error Invalid new block 187: Proof of work is higher than target
You found #1. #2 is not yet released.
You will get alot of Proof of work is higher than target before you find a block because the target is only with 8 zeros in the hash in the miner.
Team, sorry to repost this in another thread, not getting much traction in the other one. Has anyone had success getting 6 nvidia gpus working with 2 threads each, if so can you post miner, software/driver ver and config. I am using 6 asus gtx 1060 3gb cards and when trying sgminer, i have to set thread to 1 so it doesnt error out on config. I have tried ccminer SP-MOD but cannot remember the error, I am at work now.
The only way to run the solominer in 2 threads is by using the sp-mod. You need to set up 12 gpu's in the proxy..
I found a block in the pascal lite with the sp-mod #2
Based on pallas opensource.
found it....
but getting that error Invalid new block 187: Proof of work is higher than target
Jump to: