Zcash Open Source Miner Winners?

When are the going to be announced? When is the software for it going to be released too?

1 Like

Going to bump this. When are these going to be released?

1 Like

The submissions were released back in October, and several of them are already being used in both zcashd and various community miners:

The winners will be announced on December 2nd.

hmm let me guess? xenocat, tromp, and mbevand?
its dec 2nd now so it shoulod get announced today?

1 Like

I'm curious as well. Any advantages ?

We'd know if there were advantages when they're released... I haven't heard anything about them, that's what this thread is here. The $40k bounty program was a pretty big deal, no one is talking about it though.

Here's the announcement: https://z.cash/blog/open-source-miner-winners.html

@smith88 @Bensam123 It's a misunderstanding on your part: the submissions were being released right away, back in October, and similarly they were being updated in public in their GitHub repos afterwards. So there's nothing new to "release" now, and no new "advantages". The prize fund is $30k.

Rather, the Challenge was primarily about improving the public's understanding of the Equihash problem, making the optimizations equally known to all. I think (note: I was one of 3 judges) the Challenge is a success.

Before the Challenge, we had "cloud mining" providers claim/offer speeds that were like a hundred of times in excess of zcashd's, and there were almost no other publicly available solvers/miners for the 200,9 parameters (well, there was a revision of Equihash designers' code making it work for those parameters, and there was aabc's sample submission to the Challenge, but these were only moderately faster than zcashd's, and they were a side-effect of our work on preparing the Challenge).

By the Challenge's suggested early publication date of October 14, @xenoncat's submission (now CPU Winner) demonstrated a CPU solver speed that is still nearly the best now, along with documentation on the algorithm and source code (even if in a macro assembler). This has prompted @tromp to promptly open source and submit his light C++ code and his CUDA code as well, and to improve his CPU solver later to bring its speed almost on par with xenoncat's; unlike other submissions, it also specifically targets the alternate 144,5 parameter set, which might be relevant in the future. Now this is the first Runner Up. Also noteworthy and the second Runner Up is @morpav's submission, which reached similar speeds (and latest may even be slightly faster than xenoncat's now, per my testing, but it's more C++'ish and requires a very recent compiler to achieve that speed).

By Challenge end, we also got Marc Bevand's SILENTARMY submission (now GPU Winner), a very fast and portable OpenCL solver for Equihash and soon (after Challenge end) also miner for Zcash (until then, the solver was integrated and working in zogminer, which previously used David Jaenson's solver for a little while - a solver that was also submitted to the Challenge and won the last Runner Up prize).

We are aware that faster proprietary miners for GPUs exist now (hopefully, temporarily, until SILENTARMY catches up or they are open sourced or someone else releases a new open source solver/miner). Some people would think this makes the Challenge "irrelevant". I strongly disagree: I think the Challenge has served its purpose remarkably well. It improved the understanding on Equihash optimizations, it provided much faster (than would be available otherwise) solvers and miners by and in the first few days after Zcash launch, it narrowed the performance gap between open source and proprietary miners (from ~100x to ~2x), it influenced other solver/miner developers (including proprietary ones, by providing them with an easier to understand problem description and with more ideas to build upon), it influenced the "market" (people trying to sell proprietary miners for big $ before vs. asking for a 2.5% developer fee now), and it helped the mining decentralization (relatively less mining done by "cloud" providers).

1 Like

It can't be a 'misunderstanding' if they never explained that. The post you quoted is literally from today. They never announced the winners, not sure why they'd wait a month and a half before announcing the winners if the challenge basically completed when the challenge closed.

It seems as though this was not thoroughly thought out. Why would you have a challenge close months ago and only today announce winners of the challenge if the code was released months ago? Why not keep the challenge open till today if it's to encourage long term development? If that wasn't the case and the winners were basically decided a month and a half ago, why announce it today and not when the challenge closed?

In hindsight there should've been a separate bounty for CUDA as well as Nvidia cards lack the development effort that is being put into AMD as well.

I agree the challenge and bounty narrowed the gap and development of mining software has come a long way as a result,the community has benefited greatly and I for one am thankful.

1 Like

Feel bad that Zogminer didn't win anything. THey spents days and days developing their miner and didn't get much donations or prizes.

3 Likes

They certainly did explain "that" - the Challenge timeline was published, and not altered, since before the first submission was accepted. Yes, the winners were just announced today. Per the timeline, they should have been announced yesterday, and this minor delay does make me unhappy. But it's no big deal. The Challenge was not completed until now (and in fact, I think the paying out of prizes is still pending, as it's the weekend now, but being just a judge I am not involved in that). The judges decided on the winners just a few days ago, roughly per the timeline.

I certainly think it was thoroughly thought out. Even my little contribution in planning for this Challenge took quite some effort. One of the primary goals of the Challenge was to encourage early availability of better Equihash solvers, and cross-pollination between them. This did happen. Fair judging does take time and effort. One of the judging criteria is "further potential", in which we considered what impact the submissions made in the month after Zcash launch and how they evolved (e.g., SILENTARMY did evolve a lot). For a solver that could have appeared, say, a week ago, we wouldn't know yet. BTW, note the solver vs. miner difference: at the time of Zcash launch, it'd be too much to require fully functional miners.

I can see that you'd also want to encourage solver+miner development during November, and I guess during December and on too, but there was just this one Challenge sponsored by Zcash Co. It would be wrong, and unfair to early submitters, to mix in the ~2x improvement in miners into the same Challenge that achieved the much needed and I think more important ~100x improvement in solvers pre-launch.

I disagree there should have been a separate bounty for CUDA. You might not realize just how good NVIDIA's OpenCL is compared to their CUDA. For most tasks, it's just as good, just more portable. And if you really want e.g. PTX inline assembly, that's available via NVIDIA's OpenCL too. No need for CUDA these days, most of the time. To illustrate this: two well-known offline password cracker programs (attacking hashes, encrypted archives, etc.) - Hashcat and John the Ripper - both have dropped CUDA support recently, in favor of continuing with OpenCL only. For both, CUDA support was legacy for a while before. Yet both support NVIDIA cards to the same extent as AMD ones (in fact, the stability on NVIDIA tends to be better). Disclosure: I am involved in John the Ripper project.

As to a separate bounty for NVIDIA vs. AMD, regardless of implementation language, that would make some sense, but please recall the situation we were in back then and the limited prize fund we had. To announce a separate NVIDIA prize would mean no Runner Up prizes. So we chose to leave ourselves more flexibility; if someone submitted something really great specifically for NVIDIA, that would in fact win them a Runner Up prize (and @tromp's CUDA support may have contributed to him receiving the largest Runner Up prize, although 144,5 was another very important factor in that).

Regardless, Marc Bevand did really good work with SILENTARMY post-Challenge, including adding NVIDIA support (in OpenCL, and that's how it should be, short of going to assembly level) and achieving decent performance on those cards as well. I do hope he, or someone else, will find time for even further improvements as well, but it's out of scope of this Challenge, and it couldn't reasonably have been part of it.

One way to be supportive of them would be to throw some sol/s into zogpool and encourage others to do the same. It's good to have committed devs active and supported in the ecosphere, and I like to think this is one way to do that.

2 Likes

I did bring this issue up with Zcash Co privately, and I hope some token(s) of appreciation will be sent to Zogminer team. There's no way Zogminer could have won any prize in a Challenge focused on Equihash solvers, with published judging criteria for such, when their submission never included a complete Equihash solver of their own, nor even original ideas added on top of someone else's. I think it's not a "blow", but just fair judging, and any other decision would be unfair to others. I think Zogminer developers would agree, and are not offended by them not winning this solver Challenge.

There were also other miner projects that didn't even submit their code to the Challenge (because they correctly understood it'd be out of scope), and they were right in not doing so. If we were to award a prize to a miner, we'd offend those other projects.

As I understand, the reason Zogminer was submitted was the team's intent to complete an own solver. So it was fair play on their behalf. They just never did complete this, and suitable third-party solvers for them to use appeared sooner, so they did use those (that's great, and I did thank them for that kind of community participation privately).

2 Likes

Hello all!

Just wanted to say I feel the prizes handouts are completely fair and I am very pleased to see all those who won getting their well deserved bounties. Please support Silentarmy and eXtremal and David. They are all really awesome devs and nice folks too.

I wanted to work on zogminer before there was a contest in response to what I felt was centralized mining. I don't have the skills that Marc and others had to achieve what they did and always knew that. Zog was a community effort and I made some friends I hope will last forever and learned a ton.

Thanks to the zcash team for putting this together.

7 Likes

Respect................

I agree it's good to support Zogminer developers, and eXtremal. I did notice eXtremal's major post-deadline contribution to SILENTARMY. It would have been wrong (unfair to others) to use any of the Challenge prize money to reward that, yet I do feel (and I informed Zcash Co) that his contribution should be awarded at least some token of appreciation as well.

4 Likes