Topic: 500 BTC BOUNTY FOR OPENSOURCE BITSTREAM THAT IS FASTER THAN TML *STILL OPEN* - page 2. (Read 8605 times)

So let me get this straight: you've developed a bitstream for those Spartans, but you don't want to release it because if everyone else uses your bitstream, they (customers who bought Spartans) will gain between 8000-14000 BTC (your numbers) over the next 3 months, but drive the difficulty ~2.5% up in the process. In the process, you would lose ~150BTC from the increase in difficulty.

Umm, if you've been promised > 400BTC for releasing your bitstream, then why are you so concerned about losing 150BTC in mining?

If you finished nonce distribution logic and get/return work logic you may consider to release it now. With few cores just for test purposes. Those cores may be slow but have to return valid nonces. If you do that the rest of comunity may already start to design propertiary comunication logic for varius boards.

I will explain. I expect that there's about 5'000-8'000 Spartans in circulation, working "on average" 200 Mh/s - that is 1 - 1.6 Th/s around. Giving it say 270 Mh/s would turn them into 1.35 Th/s to 2.16 Th/s. Say additional 0.35 - 0.56 Th/s to difficulty.

Current network average is about 20.5 Th/s - so added 0.35 - 0.56 Th/s would make +2.4% to 2.7% difficulty increase alone with such disclosure.
Which will cut also directly income from my existing installations of 300 Gh/s - that's DIRECT losses of 150 BTC to end of year :-) So in your offer component it can be reduced. Why I have to help competitors there ?

Then - goes about INDIRECT losses with ongoing ASICs - Tyrell one of not many who invented same tricks in bitstream as I am, he's very smart. But most actually not - i.e. friedcat's chip for example, and likely that BFL chip as well, as their stratixes would work better. So this disclosure would also result that they (and maybe you as well) can catch up and get extra 15-20% performance without doing much - i.e. just rewriting your top-level code and doing re-synthesys.

Then - gained income from today to end of year by Spartan owners by using this bitstream would be in range of 8'000 BTC to 14'000 BTC aggregated. So my offer of 1'500 BTC to 3'000 BTC looks like small share of it. And of course it will drop to zero once gain will be neglible at all and not revenues will be made from board sales anymore + people invest into backwards ASICs masksets.

BTW - also you can use signcrypted Tyrell module :-) He's almost not take charge for it - if I'd were him, I would charge 50% on additional mining generated as that would be more fair compensation. So I suggest to Tyrell to gradually increase commission to at least 30% of additional mining generated. This will be more fair to his efforts even to setup that tricone-mining servers/softwares.

Also - such disclosure would probably make tricone-mining obsolete, and that I don't like as well. He worked really hard on it.

So I think the discussion is quite pointless. As on this forum I see that people like to buy much more promises than real things. And I'm not seller of promises - sorry - you should go to BFL or buy some ASIC bonds :-)))))) Why you have to buy bitstream, if you can buy almost 0.5 Th/s hashing power with that ? :-))))))))))))))))))

Glasswalker:

I actually do have the stock now to provide a dev board

see me about a delivery address and I will express overnight you one

thanks

Tom

Hi Bitfury,

No dis-respect  , but I dont see how an exta 70-105 Mh/s jump per chip is really this valuable? or this dangerous to Bitcoin that it requires this kind of approach.

Honest question here Im not trying to be a smart ass - maybe im missing something - can you explain this to me please?
Thank you
Tom

For sure, I mainly haven't reached out to ngzhang yet as I have been busy up until now with CM1, and haven't had time to directly support multiple boards.

And yes it should be a fairly simple matter for me to support Icarus, and Lancelot with this bitstream as well...

My end goal is to have a single bitstream, on a central codebase, that is maintained by the community, and which can run on all the boards out there. Get good solid protocol in place, and get mining app support of that protocol. basically establish a standard.

Enterpoint supported me with dev boards, and kicked me a couple bonus "free" boards as well for my work, so of course I helped them out first. Now that it's stable, I'm willing to take my own time to port the bitstream to any other board where I can get the hardware for free to do testing. (in the case of MMQ I'm jumping the gun a bit as cablepair hasn't been able to provide a board yet due to stock issues, but the incentive of the bounty helps motivate in this case). But ultimately I need the hardware to develop for the board properly. So if ngzhang (or one of his customers) wants to provide me with a board, I'll take the time to port it actively. I will still port it to other boards later if I stumble onto "free" time. And of course the community is welcome to port it themselves as it's fully opensource But my time is valuable these days, so if I'm going to invest my time, I'll need at least enough support to make the job much easier on me (free board so I can dev/test directly on hardware).

I have two bitstreams - one that gives 300-305 Mh/s with power consumption of 13 W per chip (that is first SERIAL bitstream) clocked at 240 Mhz.
and another one slower that gives 270 Mh/s with about 9.5 W per chip (that is second PARALLEL bitstream - 4 hashers), where rounds are clocked
at 135 Mhz while round expanders clocked at 270 Mhz.

Tools however for both predict much better timing - for first one - 337 Mhz for second one - 350 Mhz (the second one is actually failure to get more than 300 Mh/s by lowering power consumption - so my guess about power consumption was not right).

I would anyway disclose them opensource as soon as network hashing power will be more than 100-200 Th/s for educational purposes and for those who are curious. Also it contains elements that are useful for ASIC development (i.e. math optimizations). So - this is why I am not really willing to disclose it. As I would like first people to invest well into ASICs and then find that they developed obsolete hardware :-)

But - I can disclose parallel bitstream for 1'500 BTC and both of them for 3'000 BTC as now on "AS IS" terms - i.e. without bringing and testing them to your interfaces as I simply have no time for that, and I doubt that your offer could counter-bet conditions that I have now. So - it will be up to you to implement protocol communication part and run days thru compilation, and your/your engineers time used.

Please consider that 300 Mh/s bitstream is unlikely useful for you as powering this small chip with 13 W is a challenge, and looking at your hardware I suppose it would just fry chips But I wrote disclosure price, because rolled round approach is especially useful for ASIC devices, where typically with unrolled approach yield would be affected.

About price - if that is too much - then just please wait. You will finally get it anyway for free, when you would not be able to make money using it.

And finally - 2 Tyrell: - If you have objections against this disclosure in terms that it will significantly affect at this stage competition asic development, please stop me. No matter pay - if you veto this deal it won't happen. I hereby give you explicit right here. As potentially loss can be far more than these 1-2 kBTC. You have everything you need to estimate impact ;-)

Please take into account Icarus boards too, since CM1 board is its hardware clone (the only difference is RX/TX lines AFAIK). Probably you can build all-purpose CM1/Icarus bitstream, where RX/TX lines direction will be selected with a DIP-switch or something else?

Default vcore is 1.2v if that's what they are running.

Most rolled up (pipelined) hashing cores (like all of the opensource ones are built around now) basically do 1 hash per clock. so 200Mhz = 200MHash/s

Bitfury is "sea of hashers" style core. many many tiny hashing cores that each take somewhere around 66-70 clocks to finish one hash. So their relationship from clock speed to hashing speed is not as "direct".

Also their source (last time I checked) isn't open, so no way for anyone else to use their bitstream (also they used custom tools of their own design for placement and routing of the hashing cores in the chip).

My new HashVoodoo core for my bitstream is also sea of hashers design, very similar in design to theirs. But it's unfinished. I'm hoping to finish it very soon. When I do it will be an opensource bitstream hopefully capable of achieving similar speeds (but likely slightly lower than bitfury). I expect anywhere from 250Mhash/s to 300Mhash/s initially, with the ability to maybe push it a bit further than that with some optimization effort. But for now that's all speculation. I have to finish the damn thing first

Also BTW, my bitstream is opensource. It currently only targets the Enterpoint CM1 boards, because that's all I have for dev on-hand. But I've already started talking to cablepair and the other involved parties about releasing builds of HashVoodoo for the ModMiner Quad boards too. MMQ uses different communications methods/protocols though, so It will take a bit of work to adapt the bitstream to play nice with the JTAG comms that MMQ uses. Then of course we'll need to get support for it in the mining clients added.

mining software dynamically clocks the fpga's which is the proper way to mine with fpga for many reasons. Usually on the MMQ they run anywhere from 190-210Mhz.

As far as the vcore I dont know the exact number its just whatever the stock vcore is for the chip. We do not increase the amount of normal voltage to the chip.

In the area of 200MHz I believe. Not sure on the vcore.

Out of curiosity, what MHz and Vcore do the MMQ run at currently?

this is experimental stuff

we are not going to be messing around with VCORE's

although MMQ would be ideal if someone wanted to take one and test it out with this

we used almost exclusively the Industrial grade chips on them.

300 Mh/s @ 240 Mhz @ 1.25Vcore for around 12W per chip. Not bad.

BTW bitfury has similar high-speed bitstream for XC6LX150:
http://www.bitfury.org/xc6slx150.html

You couldn't be more wrong about the whole situation.

This is not to get people to buy more ModMiner Quad's in fact I am no longer advertising them, and Highly doubt I will sell many more of them at all.

My customers are awesome, they have been very supportive since the beginning, I simply want to take a little of the profits and give back to them and make their equipment better.

There is quite litterally no possible way for me to make money from this, in fact I am losing money from this - but its not for me - its for the customers who already purchased FPGA and are worried about getting a ROI, my customers have been really good to me and I just want to give something back to them.

You are right, bu$ine$$ are bu$ine$$, sometimes I let go by utopias when I read a letter with such fair intentions.

Thank you for putting my feet on the ground.

Regards

What else? You mistake the arguments which were used to try to convince eldentyrell with the motivations which were quite obvious from the start and shouldn't need any explanation: make MMQ more profitable to users so that they buy more of them.
Reread the initial proposition, it was 200 + 200 = 400 BTC (200 out of cablepair's pockets, 200 out of BTCFPGA sales).
Didn't see any wolf but I smell a troll.

I think the trick wasn't successfully, you use good words, like open source, contribute to the bitcoin community, etc, all in a public letter to force eldentyrell to accpept it or to seems an ungrateful. Having to thank you a donation of 200BTC, and now we face the truth, the was a hiding purpose, bussiness.

Now you pay the double for anyone to make the same or better job? A wolf under the skin of a lamb, maybe this algorithm that make your fpga get a 25% more of hash power can do your bASICS get better performance compared to the competence?

Money talks

Re the revised bounty: should clarify if the goal is TLM speeds before or after the future-fee is taken out.