CCminer(SP-MOD) Modded GPU kernels.

[sp_]

I have decided to release my optimized rainforest v2 miner for NVIDIA

https://github.com/sp-hash/microbitcoin/releases/tag/1.0

The miner is around +400-500% faster than the DJM34 miner

Gtx 1060 3gb 5.2 MHASH ($0.81 per day)
RTX 2060 SUPER 14 MHASH ($2.18 per day)

[1714833513]

1714833513

[1714833513]

1714833513

[1714833513]

1714833513

[1714833513]

1714833513

[1714833513]

1714833513

[HaloGenius]

I have decided to release my optimized rainforest v2 miner for NVIDIA

Linux build ?

[pallas]

I have decided to release my optimized rainforest v2 miner for NVIDIA

https://github.com/sp-hash/microbitcoin/releases/tag/1.0

The miner is around +400-500% faster than the DJM34 miner

Gtx 1060 3gb 5.2 MHASH ($0.81 per day)
RTX 2060 SUPER 14 MHASH ($2.18 per day)

Where is the code?

[sp_]

Where is the code?

Why do you need it? 

microbitcoin has successfully forked to a new pow algo. Yespower. (antigpu)

[WhyMe]

So why publish an unusable miner ?

[sp_]

It was usable for 8 hours.

I published it because I wanted to show that I beat the djm34 +500% faster, in a memory algo that has been minable for 6 months.

A few years back you could take a CPU minable algo, convert to gpu code and make it +5-600% faster than a cpu. I don't see why this couldn't be done again...

[scryptr]

It was usable for 8 hours.

I published it because I wanted to show that I beat the djm34 +500% faster, in a memory algo that has been minable for 6 months.

A few years back you could take a CPU minable algo, convert to gpu code and make it +5-600% faster than a cpu. I don't see why this couldn't be done again...

YOU ARE JUST TOO GOOD--

So good that they decided to fork...

Not bad at all, really.  I downloaded it.  Does the miner mine anything else in the CCminer inventory?  Haven't  actually tried it yet.       --scryptr

[reb0rn21]

rvf2 is explotable, jericko has been mining with his private miner for months with more speed then sp now do

sp_ started a week+ mining or so, him or jericho are behind the pump you now see on MBC

[ivakar]

rvf2 is explotable, jericko has been mining with his private miner for months with more speed then sp now do

sp_ started a week+ mining or so, him or jericho are behind the pump you now see on MBC

so you mean there is even faster miner then sp just shown?
interesting 
but not for majority of miners I think
the coin is old, but if you have pool, or has a lot of hashing power could mine it solo..
it's a pity I have nothing of it 

[sp_]

sp_ started a week+ mining

nope. I've been there for a long time.

[sp_]

jericko has been mining with his private miner for months with more speed then sp now do

More speed? His rigs are probobly bether than mine. I still have gtx 750ti cards (2mhash @ 38 watt)

[pallas]

jericko has been mining with his private miner for months with more speed then sp now do

More speed? His rigs are probobly bether than mine. I still have gtx 750ti cards (2mhash @ 38 watt)

then you should replace them, it's a waste of power and, thus, money.
morover, they are very difficult to sell now, as they are out of warranty.
think how much more money you'd made, using the same electricity power, if you sold those some years ago.

[sp_]

then you should replace them, it's a waste of power and, thus, money.

Most of my old cards are replaced. The rainforestv2 miner did 1.7-2 MHASH@ 38watt on the gtx 750ti. and DJM34's fork did 2MHASH on the gtx 1070@150 watt.
So by optimizing the code, I turned on the old horses again with a great profit.

[scryptr]

IS THERE CURRENTLY ANY COIN UTILIZING RAINFORESTV2? --

Apparently the tiny volume of MicroBitCoin (MBC, now on Power2b algo) makes it a long shot.  Is there any other coin on RainForest V2 (rfv2)?

--scryptr

[ivakar]

IS THERE CURRENTLY ANY COIN UTILIZING RAINFORESTV2? --

Apparently the tiny volume of MicroBitCoin (MBC, now on Power2b algo) makes it a long shot.  Is there any other coin on RainForest V2 (rfv2)?

--scryptr

yesterday I was googling exact same question too and the results were dissapinting
the only one coin which use this algo is MicroBitCoin.
and because it switched to new algo as well, there are no pool for rainforest algo
so in order to mine it - you have to do it solo

[C3Sector]

hi there sp will there be power2b supported ccminer by you ? thanks in advance.

[scryptr]

IS THERE CURRENTLY ANY COIN UTILIZING RAINFORESTV2? --

Apparently the tiny volume of MicroBitCoin (MBC, now on Power2b algo) makes it a long shot.  Is there any other coin on RainForest V2 (rfv2)?

--scryptr

yesterday I was googling exact same question too and the results were dissapinting
the only one coin which use this algo is MicroBitCoin.
and because it switched to new algo as well, there are no pool for rainforest algo
so in order to mine it - you have to do it solo

MICROBITCOIN IS ON ZERGPOOL--

But the only exchange for MBC has the wallet in maintenance.  You can mine MBC on the pool, but there may be a wait for payment.       --scryptr

EDIT:  The exchange wallet at StocksEchange is online for MBC, and ZergPool is live with MBC and Power2b algorithm.       --scryptr

[sech1]

Yes, I am working on RandomX and some other cpu algos like Yespower.
A few years back you could take a CPU minable algo, convert to gpu code and make it +5-600% faster than a cpu. I don't see why this couldn't be done again...

A pow kernel evolution normally move from hardware to hardware:

Timeline
----------------------------->
CPU->GPU->FPGA->ASIC


Competing with the CPU algos are easier and probably more profitable than competing with the FPGA, but the coins are normally smaller and the profit could be limited to a few rigs.

You will fail with RandomX. Period. You can try to beat my CUDA/OpenCL code however, it would be interesting to see how much improvement you can achieve. But I doubt it can be more than 10% speedup over my code.

[sp_]

But I doubt it can be more than 10% speedup over my code.

Why don't you generate ptx assembly per block and let the cuda compiler compile it before execution. (JIT style).

1. Generate a ptx .txt file in memory 1 time per block instead of parsing and executing the instructions realtime for every hash.
2. Use the buildt in runtime functions to execute the ptx. (compiling and caching of the generated cubin is done in the driver)

[sech1]

But I doubt it can be more than 10% speedup over my code.

Why don't you generate ptx assembly per block and let the cuda compiler compile it before execution. (JIT style).

Because the code changes 8 times per hash for every hash. Too much to compile and all hashes will run different code in the end, so no place for parallel execution. It's not Cryptonight/R where there was only one code sequence per block - yes, I implemented JIT compiler there without any problems. With RandomX it's either direct machine code generation (not PTX) or VM emulation. NVIDIA only allows PTX and keeps their GPU machine code proprietary.