[Bounty] New GPU Miner - Scrypter

[DarkKnight]

It occurs to me that I posted my bounty in the wrong forum initally, so I am cross-posting it here, in the hopes that in the right place it will get more visibility. After careful consideration I'm also altering the goals. I expect this will require a larger bounty to accomplish, so I've added more to the pot and I'm hoping for donations to that end as well.

Any donations to this address will go 100% towards the bounty, which will be awarded to the first individual to complete all the objectives. I have sent .12BTC to that address, and it can be monitored so everyone can be sure that the complete balance goes toward payout.

If you want to donate, but want a refund if the bounty terms aren't met, PM me your donation amount and a refund address *before* you send your donation.

To be clear, you will only receive a refund on your donation if the bounty is closed without having been fulfilled.

The Bounty is closed.


General goals:

Create a new miner: Scrypter by forking CGMiner 3.7.2 and narrowing code for support of GPUs only.

Minimum Requirements:

• Miner must be able to properly control VDDC, Engine, Mem, Powertune, and fans for all the unlocked R9 280x cards in a given rig.
• Enable Scrypt by default, no CLI switch should be needed
• Remove ASIC/FPGA code, drivers, and CLI options from the code base.
• All code must be open sourced, prior to claiming the bounty, so it can be inspected.
• Windows Binaries must be provided.
• Stable Code - must run minimum 24 hours without crashing

Optional: Support Scrypt-Jane.

Notes:

Some GPUs are voltage locked, and will not be covered under this task. The cards that I have: Asus R9 280x DCII & Sapphire R9 280x Toxic can be controlled via OEM tools such as Afterburner, Trixx, & GPU Tweak. Those cards must be able to be VDDC controlled via the miner. Currently CGMiner 3.7.2 can control the Engine, Mem, & Fans. Powertune can be controlled, but it doesn't seem reliable.

To claim the bounty, you must be the first to PM me a link to your complete github source AND a link to your Windows binary (hosted via Mega/dropbox or other 'no wait' download provider).  

There will be a 24-48 hour inspection and testing period to determine if the objectives have been correctly met before the bounty is considered won.

[DarkKnight]

reserved

[padrino]

Recommendation, the reliability will depend upon the characteristics of the clocking on the GPUs, and general health of the GPUs themselves. You might wnat to limit the claim of relability to default engine/memory clocks for the purposes of qualifying it... In addiiton you are correct it might take more then what you are offering..

Outside of the VDDC control I have 3.7.2 running for days/weeks at a time without one issue, none of the ASIC/FPGA code is compiled in (by default they are off) so it's all good.. Now focusing on improvements would be another matter entirely and perhaps this is a good way to kick that off..

[DarkKnight]

Recommendation, the reliability will depend upon the characteristics of the clocking on the GPUs, and general health of the GPUs themselves. You might wnat to limit the claim of relability to default engine/memory clocks for the purposes of qualifying it... In addiiton you are correct it might take more then what you are offering..

Outside of the VDDC control I have 3.7.2 running for days/weeks at a time without one issue, none of the ASIC/FPGA code is compiled in (by default they are off) so it's all good.. Now focusing on improvements would be another matter entirely and perhaps this is a good way to kick that off..

My rigs are running stable enough at this point where they only time they stop is when I run a maintenance cycle. The crashing I was referring to, would not mean sick/dead devices from overclocking and/or undervolting.

Also, I'm hoping for donations toward the cause to spur development. I'll leave it open for a few days and if no one else wants to make a meaningful donation toward development, I'll close the bounty and simply keep controlling VDDC via OEM tools.

I too would like to see speed improvements, however I think it's best that those come on the foundation of a clean code base already specialized to AMD GPUs. Further, based on other topics, the real speed-ups will come from improvements to the kernel, which isn't being developed anymore either. I'm also unwilling to be the solo donor to spurring kernel developments that would give speed improvements for everyone, but I will donate for a ASIC/FPGA free codebase.

[Luke-Jr]

What's the point to removing ASIC/FPGA/SHA2 support?
If someone wanted an enhanced scrypt miner, the logical move would be to join the BFGMiner team and improve it there...

[DarkKnight]

What's the point to removing ASIC/FPGA/SHA2 support?
If someone wanted an enhanced scrypt miner, the logical move would be to join the BFGMiner team and improve it there...

What was the point in removing GPU support from CGMiner after 3.7.2?

The point is to promote GPU mining software development, with a GPU specialized mining software tool.

BFGMiner - a modular ASIC/FPGA Bitcoin miner
bfgminer.org/‎
BFGMiner. a modular ASIC/FPGA Bitcoin miner. Last version: 3.9.0. Windows 32bit - Windows 64bit. Ubuntu: PPA mantained by "unit3" Arch: pacman -S ...

Why would the logical move for an individual, that uses GPUs for mining, to help support an application (BFGMiner) that bills itself as an ASIC/FPGA Miner? ASIC/FPGAs quite literally killed SHA256 GPU mining. Its a competition. Keep Scrypt GPU hash rates/watt high enough, FPGAs look much less attractive, slowing sales and development; keeping Scrypt GPU only, longer. That to me is the logical move.

[Shadow_Runner]

Must choose second GPU without errors by command line for users with nVidia + AMD installed

[Spiffy_1]

Keeping an eye on this.. cgminer and bfgminer have made me plenty of money.  I did find a few tweaks to strip out unneeded code to get my 7950's running a bit better(10 khps).  Still using 3.2.2.  Would be nice to update to something a little faster but it seems that as the authors went asic crazy, the scrypt performance went down.

[Luke-Jr]

What's the point to removing ASIC/FPGA/SHA2 support?
If someone wanted an enhanced scrypt miner, the logical move would be to join the BFGMiner team and improve it there...

What was the point in removing GPU support from CGMiner after 3.7.2?

The point is to promote GPU mining software development, with a GPU specialized mining software tool.

Who knows. cgminer isn't exactly a sensible choice in the first place.
BFGMiner 3.9.0 still has CPU and GPU support.

BFGMiner - a modular ASIC/FPGA Bitcoin miner
bfgminer.org/‎
BFGMiner. a modular ASIC/FPGA Bitcoin miner. Last version: 3.9.0. Windows 32bit - Windows 64bit. Ubuntu: PPA mantained by "unit3" Arch: pacman -S ...

Why would the logical move for an individual, that uses GPUs for mining, to help support an application (BFGMiner) that bills itself as an ASIC/FPGA Miner? ASIC/FPGAs quite literally killed SHA256 GPU mining. Its a competition. Keep Scrypt GPU hash rates/watt high enough, FPGAs look much less attractive, slowing sales and development; keeping Scrypt GPU only, longer. That to me is the logical move.

You're quoting from a Bitcoin forum post.
scrypt is not Bitcoin. Someone who takes over maintaining the scrypt side for BFGMiner is welcome to advertise it differently.

[DarkKnight]

Luke: You don't CPU mine on a GPU rig. Most GPU rigs use low end budget CPUs to begin with, and even tying up two cores of an I7 920 @ 4GHz is enough to negatively affect hash rates of the GPUs. I've tested this myself.

I any case, if one did have the desire to mine on his CPU while using his GPUs, you could always run a dedicated CPU mining program as well. You don't need a program that does both, because you don't CPU mine GPU coins, and you don't GPU mine CPU coins (yet). It's bee clearly demonstrated that once any significant speedup is achieved on a different flavor of hardware CPU->GPU->FPGA->ASIC it will quickly raise the network difficulty enough to make mining with lower tier hardware unprofitable.

Did you even read the quote? The quote about BFGMiner was pulled directly from Google's first search result of BFGMiner. It's not a bitcointalk thread, it's directly from BFGMiner.org, and says so in the quote. It is BFGMiner.org's own advertisement. Why would someone bother maintaining a 'Scrypt side' of BFGMiner instead of just building a pure Scrypt mining engine? Bitcoin may have the market share, but Scrypt coins are taking over. The GPU market is huge, and deserves it's own dedicated software.

I'm not going to keep arguing the point. If you want to keep supporting BFGMiner's 'scrypt' side, feel free to do so. In the mean time, I will continue to pursue my interests in getting a dedicated GPU miner written.

Shadow_Runner: As to choosing which GPU (AMD/Nvidia) to run without errors, I don't see that as a necessary requirement at this stage. It's not terribly complicated to pick the platform you want to use when initializing the miner. Further, the intended goal is to deeply specialize in AMD hardware. I suppose eventually it should just disregard other platforms and autoselect AMD, but that's not a sticking point for me at this stage.

Spiffy_1: It's made a lot of people a lot of money. Where's the giving back to the community for it? We could have ended up with crappy mine-share software like the GPU miner for CPU coins that force mines something like 12% for the author. It's one thing to want to make some money on your work, but that's freaking robbery if you ask me. If I were into those coins, I'd put up a much larger bounty for a GPU miner just to make his miner pointless to use, on principal.

Just 1% of everything made on mining with CGMiner using GPUs would buy a lot of damn programmers, and possibly keep us competitive longer. Instead, everyone hoards their coins and does little to work together for something better. I've put up a lot more in that bounty than 1% of what I've taken in since joining the Scrypt community.

[fran2k]

Great initiative.

[Notanon]

Given that there is a Nvidia-optimised scrypt miner in the form of CUDAminer, might be worth focusing on AMD cards to try and find ways of optimising their output instead of trying to develop a catch-all program for both. Worst case, you can incorporate elements of both down the track into a combined GPU Scrypt miner program.

On a side note, has anyone tried developing an Intel HD Graphics miner? Might not be able to get much of it, but would help to squeeze a bit more hashing power out of some desktop PCs.

[slavo]

Luke: You don't CPU mine on a GPU rig. Most GPU rigs use low end budget CPUs to begin with, and even tying up two cores of an I7 920 @ 4GHz is enough to negatively affect hash rates of the GPUs. I've tested this myself.

I CPU mine with all my rigs with i7 3770k; with 3.5 cores for 7 threads; and it does not affect my gpus (4 r9280x)... Just to say 

[barruka]

.12 btc? its a joke?

you are not a developer, aren't you?

you cant even imagine how much time is needed to do what you are asking for

I think you must reward at least 5 btc for that job

[Luke-Jr]

Did you even read the quote? The quote about BFGMiner was pulled directly from Google's first search result of BFGMiner. It's not a bitcointalk thread, it's directly from BFGMiner.org, and says so in the quote. It is BFGMiner.org's own advertisement.

BFGMiner.org is a fan site and completely unofficial.

Why would someone bother maintaining a 'Scrypt side' of BFGMiner instead of just building a pure Scrypt mining engine?

Eh, you have that backward. Maintaining scrypt support for BFGMiner is far less work and more rational, than building a pure scrypt miner.

Bitcoin may have the market share, but Scrypt coins are taking over. The GPU market is huge, and deserves it's own dedicated software.

lolwut, this is just delusional. scrypt is a failed experiment, used only by scam "altcoins".

[DarkKnight]

Luke: You don't CPU mine on a GPU rig. Most GPU rigs use low end budget CPUs to begin with, and even tying up two cores of an I7 920 @ 4GHz is enough to negatively affect hash rates of the GPUs. I've tested this myself.

I CPU mine with all my rigs with i7 3770k; with 3.5 cores for 7 threads; and it does not affect my gpus (4 r9280x)... Just to say  

Maybe your cards aren't running particularly fast. I can provide SS if needed, but in my experience it slows them down. In any case, the bounty is for a GPU only tool. If mining with both is important to you, I suggest sticking with your current program.

.12 btc? its a joke?

you are not a developer, aren't you?

you cant even imagine how much time is needed to do what you are asking for

I think you must reward at least 5 btc for that job

No, I'm not a developer. I am a miner, putting up funds to improve OSS; and asking for donations to that end. I don't expect it to be accomplished for .12BTC, however, I think for 5BTC, I would just get a custom client coded for myself -- which would have to be damn good to payback that investment. Why would any individual invest that kind of coin for an OSS tool? Be realistic yourself. If the bounty fails, maybe I will do just that.

Did you even read the quote? The quote about BFGMiner was pulled directly from Google's first search result of BFGMiner. It's not a bitcointalk thread, it's directly from BFGMiner.org, and says so in the quote. It is BFGMiner.org's own advertisement.

BFGMiner.org is a fan site and completely unofficial.

Why would someone bother maintaining a 'Scrypt side' of BFGMiner instead of just building a pure Scrypt mining engine?

Eh, you have that backward. Maintaining scrypt support for BFGMiner is far less work and more rational, than building a pure scrypt miner.

Bitcoin may have the market share, but Scrypt coins are taking over. The GPU market is huge, and deserves it's own dedicated software.

lolwut, this is just delusional. scrypt is a failed experiment, used only by scam "altcoins".

I can't tell if you feel threatened or are trying to troll me. Keep it up, I'll take the free bumps. It's also entertaining watching you argue; so No U.

Edit: Just looked up some of your posts. You're a BFGMiner developer? That's so adorable watching you try to protect it. 

[Spiffy_1]

Why the hell does it matter if he is a bfgminer developer?  Why does it matter that you want it done on cgminer?  The important thing is a continued interest in scrypt and the continued increase in GPU optimization.  Take the good from both and work together.  We all love scrypt coins and all have a vested interest in their success.  CGminer stopped scrypt support a few versions ago.  If bfgminer is still supporting scrypt why not take the scrypt code from both, analyze and put back into the projects still under development.  I'm not a coder but I do know that different people have different ways to do things.  some work better than others.

[hipstacoin]

Keeping an eye on this.. cgminer and bfgminer have made me plenty of money.  I did find a few tweaks to strip out unneeded code to get my 7950's running a bit better(10 khps).  Still using 3.2.2.  Would be nice to update to something a little faster but it seems that as the authors went asic crazy, the scrypt performance went down.

Hey man, that sounds interesting. Care to elaborate on those tweaks?

[hipstacoin]

Luke: You don't CPU mine on a GPU rig. Most GPU rigs use low end budget CPUs to begin with, and even tying up two cores of an I7 920 @ 4GHz is enough to negatively affect hash rates of the GPUs. I've tested this myself.

I CPU mine with all my rigs with i7 3770k; with 3.5 cores for 7 threads; and it does not affect my gpus (4 r9280x)... Just to say 

Really? What hashrates are you getting from the CPU miner and what are you using?

[nonny12]

.12 btc? its a joke?

you are not a developer, aren't you?

you cant even imagine how much time is needed to do what you are asking for

I think you must reward at least 5 btc for that job

Is this really necessary? The original cgminer was created by a programmer 'in his spare time' - there was ZERO bounty.