Okay, so I found a developer willing to work on new open source Nvidia mining software - djm34. His background can be seen here:
https://github.com/djm34So the goal of the bounty is as follows:
Open source Nvidia mining software that performs better than Tsivs existing software. For the record, here's a screenshot by _sp
Granted, these hashrates are for _sp's modified kernals.
djm34 apparently fiddled with tsivs code a little bit and got these numbers:
260H/s 750ti
400H/s 780ti
560H/s 980
The 750 ti usually gets 250 h/s without _sp's mod. If anyone has any other hashrates to post for comparison (780ti and 980), please take screenshots.
--- The financials ----
So, as all Monerians should know by now, the Core team has created the forum funding system, which is an attempt to move away from the conventional bounty system. The primary benefit of this system is the developer can set any milestone and timeline for payouts, and then the funders basically agree or disagree based on whether or not they fund something.
djm34 really wants an advance for his work, and this advance is not compatible with the forum funding system. I tried to convince him that a milestone payment for evidence of effort was equivalent to an advance (except that it was held in escrow by fluffypony), but apparently advances are "industry standard".
Advance: "1000 - 1500 xmr"
Total: 5000 XMR
So, as it stands, we would have to send 1k - 1.5k xmr to djm for him to start to work on the problem. After completion of the task, he would get the remaining bounty.
If anyone is willing to provide this advance, please post.
He said he could complete the task in 3 weeks, and that he didn't want to make any promises, but thinks he can get the cards on par with the AMD counterpart.
so, all details
Advance: "1000 - 1500 xmr"
Total: 5000 XMR
Timeline 3 weeks
Goal nvidia on par with amd counterpart (so around 800 h/s on a 980ti i guess)
Final products open source code on github and compiled binaries for windows and linux
Additional goals (not discussed yet) could be scaleability (the code could work on any future card, at least a maxwell card) and increased usability (getting tsivs to work in linux requires much patience). But these could be pie in the sky for all I know.
I am going to pitch this task for him on the forum funding system, with the "evidence of effort" milestone in place of the advance.
If anyone wants to pony up the advance here, that will work as well.
The rest of the bounty will be done on the forum funding system.
Also, if Wolf0 could chime in here, I think that would really help djm34's case for the advance.
edited once to add final products