Baikal Giant N - Cryptonight, Cryptonight-lite FPGA/ASIC miner

[Wananavu99]

Guys, we will have the coin to mine with Cryptonight ASICs. XMC. So it looks that this is safe to buy some cryptonight ASICs. Now, when we can buy 1 and get 5....  

BTW: https://bitcoinmagazine.com/articles/monero-just-hard-forked-and-it-resulted-four-new-projects/

It will be probably pretty much shitcoin but probably with good value

Pump and dump dude, at least you can make some monies off your miner. 

[oomurashin]

when its start 1 buy to 30??? 

[Truthchanter]

I haven't seen like any discussion on this.. the new deal for the baikal giant-b is buy 1 get 2.. anyone think this is a decent deal and getting some? like $2699 or something for 2, guessing usd wire transfer only

[Mnz00]

Monero community could be vibrant, but good question is why the monero devs didnt increased the scratchpad size from the current size (2Mb), fpga miners are not affected by this fork

I talked to another cryptonight coin developer about fpga, he said that it's cost prohibitive due to memory restrictions.  So there goes fpga out the door, cause I was thinking fpga too. 

I'm developing many algorithms for the Virtex Ultrascale+ VU9P on the VCU1525 board and this FPGA has 47MB of internal SRAM, enough for 22 instances of Cryptonight V7.  Your developer friend is gravely mistaken.
Having said that, Cryptonight V7 is not the most profitable algorithm for FPGA's at the moment, but high end FPGA's always beat GPU's, and ROI for the high end FPGA's are around 50-110 days vs. years for GPU's.
I will be releasing my bitstreams (to the public) soon, my current plan is to release them for free with a 4% mining fee.  And you would need to buy your own FPGA boards from Avnet/Digikey/Xilinx/Digilent/Hitechglobal/Bittware, etc...

Amazing. Can't wait for that. Do you have an estimated date?

First public-ready algorithms should be done around mid to late June.  However I am hesitant to 'release' the bitstreams unless numerous algorithms are available.  I wouldn't want someone to invest $10K in hardware with only 1 or 2 algorithms in case those coins profit or ROI suddenly changes.  A good investment would be to have 4-5 algorithms available, making the hardware more secure against market changes.
Anyway, once people start mining with PC's linked to numerous high end FPGA cards, they will be 'immune' to forks since the algorithms in the FPGA's can be changed as fast as developers can fork their coin, actually faster in most cases since even a poorly tested rig can still mine, whereas a coin cannot fork until the new setup is heavily tested.

Thank you for the amazing work you have been doing with FPGAs. I truly respect you for pushing the boundaries of today's mining knowledge.

So let's say that FPGAs turn out to be a healthy and profitable substitution to GPUs as we hope, so far Cryptonight is the only algo that was able to put ASICs out of the game for a while, this being said, won't ASICs just be developed for the remaining algos, beating FPGAs at what they can do best?
If so this leaves me thinking that FPGAs profitable work could then be allocated to Cryptonight, which is ASIC resistant as we speak, but as you have stated yourselfthis algo implementation is not the best mining use case for FPGAs and at that point once more accessible FPGA mining solutions are released and more FPGA hardware is purchased won't that put us back where things are now?

[leowonderful]

So, it seems all cryptonight coins going to hardfork soon or late? Bytecoin, Karbo, Electroneum etc.

Yes, nearly all of them except for some obscure coins with tiny market caps, listed on only one or two exchanges that are also tiny, that won’t be profitable when miners try to mine them. Basically RIP for CrytoNight ASIC buyers, there's no way anyone that purchased these ASICs will ever make their money back unless a miracle happens and they somehow find a way to make these things profitable again through adding a new algo or a major coin doesn't hard fork for some reason.

[tadas]

Hi there,

have just bought Baikal N , one question if it is linux based system can I just use another software to mine cryptonightv7 or it is hardware/firmware based?

Thanks

[smoolae]

Hi there,

have just bought Baikal N , one question if it is linux based system can I just use another software to mine cryptonightv7 or it is hardware/firmware based?

Thanks

CryptonightV7 is not minable by ASICs, that's the reason it was implemented - to get ASICs out of Cryptonight.

[treanski]

Hi there,

have just bought Baikal N , one question if it is linux based system can I just use another software to mine cryptonightv7 or it is hardware/firmware based?

Thanks

try the new doorstopper coin

[smoolae]

Hi there,

have just bought Baikal N , one question if it is linux based system can I just use another software to mine cryptonightv7 or it is hardware/firmware based?

Thanks

try the new doorstopper coin

 

[tadas]

I am not saying it is worthless, NiceHash works fine. Just looking to the future, if suddenly all coins will fork to v7

[tadas]

They do not show what coin they are mining it only shows CryptoNight as only information.

[snoopydoggie]

Gosh! Looks like the trend is changing like crazy...I guess GPU would gradually back off and FPGA is the future of mining

[DigitalCruncher]

..
So let's say that FPGAs turn out to be a healthy and profitable substitution to GPUs as we hope, so far Cryptonight is the only algo that was able to put ASICs out of the game for a while, this being said, won't ASICs just be developed for the remaining algos, beating FPGAs at what they can do best?
If so this leaves me thinking that FPGAs profitable work could then be allocated to Cryptonight, which is ASIC resistant as we speak, but as you have stated yourselfthis algo implementation is not the best mining use case for FPGAs and at that point once more accessible FPGA mining solutions are released and more FPGA hardware is purchased won't that put us back where things are now?

Let me answer. All the answers to your questions - yes. But you can do a lot before these problems arise.

I think that ASIC miners will appear for all more or less good crypto-currencies that exist now. But until then, the FPGA will bring income of 100-200% for 1-2 years, it is quite real.

There are several ways to keep FPGA solution, inclding:
1. Adaptation to reality every 3...6 month by implementing a new algorithm.
2. Creating FPGA-solver, that is impossible to reproduce in ASIC more effectively, than in FPGA.
3. Creating of altcoin with POW function that would be usable in some real-life application (genome, networking, etc). See Amazon cloud instance F1 with Virtex Ultrascale+, it's price ~1 USD/h..

Speaking of current problems, it is much more simple to implement miners by using existing hardware. Ironically, people who have FPGA (and programming skills) do not trust in cryptocurrency and do not consider such activity. So we have a chance.

[whitefire990]

Thank you for the amazing work you have been doing with FPGAs. I truly respect you for pushing the boundaries of today's mining knowledge.
So let's say that FPGAs turn out to be a healthy and profitable substitution to GPUs as we hope, so far Cryptonight is the only algo that was able to put ASICs out of the game for a while, this being said, won't ASICs just be developed for the remaining algos, beating FPGAs at what they can do best?
If so this leaves me thinking that FPGAs profitable work could then be allocated to Cryptonight, which is ASIC resistant as we speak, but as you have stated yourselfthis algo implementation is not the best mining use case for FPGAs and at that point once more accessible FPGA mining solutions are released and more FPGA hardware is purchased won't that put us back where things are now?

An expertly programmed FPGA can even compete with an ASIC on the same algorithm, although the ROI is not spectacular, it is still worthwhile to mine.  For example, the VCU1525 can make a decent profit even while mining Myriad-Groestl (Baikal X10 ASIC), so even competing against a dedicated ASIC, the FPGA can still mine with profit, although the ROI will not be maximized (ROI when competing against ASIC's is 1-2 years; ROI when mining non-ASIC coins is 50-150 days).

However, some algorithms like Timetravel8, Timetravel10, X11Evo, X16R, X16S, they change the algorithm every block.  An FPGA can rapidly reconfigure itself to optimize itself each block; an ASIC cannot.  Therefore these algorithms are truly ASIC-proof, in that no ASIC can beat an FPGA (ever) on those coins, unless the ASIC itself is just designed as an FPGA which would be a waste of money since you can just buy ready-made FPGA's.

Once I publish the website for the DIY-FPGA mining, I will strongly push new crypto coins to go with one of those dynamically changing algorithms, since they are the only ASIC proof algorithms and they would guarantee decentralization in the future world of DIY-FPGA mining.

I will publish an X16R bitstream which can do Ravencoin at an incredible rate; however these dynamically changing algorithms are extremely complicated to implement in FPGA's so it will be early 2019 by the time I have that one ready.

[cangliang1121]

Hi, guys. Some friends told me that they can update the Baikal Giant N to support the Cryptonight V7. I just have no idea how it is possible... I think the Baikal N is ASIC, not FPGA. Updating the firmware only cannot support new algos. Right? Maybe they have some other way?

[stash2coin]

However, some algorithms like Timetravel8, Timetravel10, X11Evo, X16R, X16S, they change the algorithm every block.  An FPGA can rapidly reconfigure itself to optimize itself each block; an ASIC cannot.  Therefore these algorithms are truly ASIC-proof, in that no ASIC can beat an FPGA (ever) on those coins, unless the ASIC itself is just designed as an FPGA which would be a waste of money since you can just buy ready-made FPGA's.

Correct me if im wrong but i think that algos like X16R still can be attacked with ASICs, you dont have to implement all of the algos that this algo is cycling through you just have to pick the once that are for example 3-4 times even more faster on ASIC compared to a GPU with a much less electricity consumed then you mine when blocks with the implemented algos are coming.

[cangliang1121]

A lot of people have asked me about the FPGA rig that I am building, and make a long story short, I am hoping eventually to release bitstreams for a couple of very common FPGA boards; in this fashion, the average person can just buy one to 30 of these 'stock' FPGA boards, connect them to their PC by USB cables and PCIe-to-USB3 cards, and start mining with the publicly available bitstreams for a 2% fee.  The ROI on high end FPGA boards right now is 15 to 90 days depending on the algorithm and the board.  This setup is almost immune to 'forking', and in my opinion, GPU's will gradually be replaced by FPGA's and I believe stock-hardware FPGA mining with publicly available bitstreams will replace the current set up of stock-GPU's with publicly available mining software.

When the transition from GPU's to FPGA's is complete, true-ASIC rigs will not be that attractive.  They will offer only a moderate hash rate increase, for high risk.

Very very interested. When will you have them ready? And where is it possible to buy those high end fgpa? Thanks!

Other than myself I know of at least two other people who are also working on open-platform FPGA rigs, which means in reality there must be even more than that.  Likely several will be released around the same time.

High end cards are available from digikey.com, avnet.com, hitechglobal.com, bittware.com, xilinx.com.  The lowest end card that can ROI in around 30 days is the $490 Digilent Nexys Video [Xilinx XCA7200T] (available from Digilent.com, Avnet.com, Digikey.com).  However, the Nexys card is limited in which algorithms you can mine, and personally I believe the future of open platform FPGA rigs is in the high end cards which can run almost every altcoin algorithm.  The high end cards cost around $4K to $6K each, which is around the same price of custom mining rigs, with much better ROI's, more flexibility, and none of the 'screw-you-over' attitude of the big mining companies.  With full open access to your own hardware, there are no 'secret' or 'locked' algorithms which are out of your control.

I'm working full time on this project, I might have something publicly available by June or July.

I forked tpruvot's CPU miner, the miner works the same on the command line, with -a specifying the algorithm, and the PC mining software loads the correct bitstream into the FPGA card right before you start mining.  If you want to run profit switching, the software just reconfigures the FPGA in a few seconds and then switches algorithms.

The other tremendous gain is the low power consumption.  A high end FPGA card burns around 150-200W and makes $40 to $270 per day.  Which means you can 'live' off mining revenue without a complicated cooling system, 220V circuits, and all the other headaches of GPU's.

Very interesting indeed, thanks for sharing. I have a few questions, but for folks doing searches, the correct part for the Nexys is XC7A200T

So, on to the questions, and please correct me wherever my assumptions are wrong; not an expert in FPGA by an means.

1. The low-end fpga in your example, seems to only come with 512MB 800MHz DDR3, which also means that it would not be able to mine algos that require 2GB of memory or more, correct?
2. Do you have a few high-end fpga example boards to link to?
3. Do you have a working proof-of-concept, what algo is it running, on which board, and is it stable?
4. Further to #3 above, what are the temperatures like at full blast?

The high end board I am developing on has 64GB of DDR4 and there are 8 boards in the rig (512GB), however another board I am working with has 288MB of QDRII+ RAM which is way faster than DDR4/GDDR for random access.  Still, generally the best ROI is for algorithms that are not that memory intensive and can be done internally on the FPGA without external memory accesses.

I am building hash functions one by one (functions, not algorithms, like X11 = 11 different hash functions), and no, I have not finished enough hash functions to run a full multi-function algorithm yet.  However based on the speed of the individual hash functions it is possible to calculate the hash rate of the full chain well in advance.

Anyway, time to get back to work.

Nice work! I think you may firstly profile the C version algo, then find out the hotspot. After that, you may try to split the algo into CPU part and FPGA part. Lastly, implement the FPGA part with RTL. Right? I am interested in implementating the CN heavry algo with FPGA. Just cannot start the project. :-)

[M5M400]

 For example, the VCU1525 can make a decent profit even while mining Myriad-Groestl (Baikal X10 ASIC), so even competing against a dedicated ASIC, the FPGA can still mine with profit, although the ROI will not be maximized (ROI when competing against ASIC's is 1-2 years; ROI when mining non-ASIC coins is 50-150 days).

any idea where one can easily buy a VCU1525? digikey has very, very long lead time.

[Freddy973]

 For example, the VCU1525 can make a decent profit even while mining Myriad-Groestl (Baikal X10 ASIC), so even competing against a dedicated ASIC, the FPGA can still mine with profit, although the ROI will not be maximized (ROI when competing against ASIC's is 1-2 years; ROI when mining non-ASIC coins is 50-150 days).

any idea where one can easily buy a VCU1525? digikey has very, very long lead time.

How it possible to pay this fpga around 1000 usd when the italian distributor make the crazy price of 4000€ plus taxes i speak about the vcu1525

[whitefire990]

However, some algorithms like Timetravel8, Timetravel10, X11Evo, X16R, X16S, they change the algorithm every block.  An FPGA can rapidly reconfigure itself to optimize itself each block; an ASIC cannot.  Therefore these algorithms are truly ASIC-proof, in that no ASIC can beat an FPGA (ever) on those coins, unless the ASIC itself is just designed as an FPGA which would be a waste of money since you can just buy ready-made FPGA's.

Correct me if im wrong but i think that algos like X16R still can be attacked with ASICs, you dont have to implement all of the algos that this algo is cycling through you just have to pick the once that are for example 3-4 times even more faster on ASIC compared to a GPU with a much less electricity consumed then you mine when blocks with the implemented algos are coming.

You should re-read the X16R white paper.  X16R uses 16 algorithms on every single block, done in sequence, in a different order.  You need to understand how pipelining works.  An ASIC capable of all 16 algorithms would not be able to pipeline the algorithms because the pipeline requires the order to be the same.  Titanic multiplexers might solve the problem for Timetravel8 and Timetravel10, but X16R has the extra change that some times you do one algorithms 3 times in the same set, making an FPGA the only method that could run the algorithms in a pipeline.

Regarding the VCU1525; the price is $3995 - $4600 USD, this is an incredible deal considering the FPGA on the board costs $40,000.  Since Xilinx makes the FPGA they can sell the board for whatever price they want.  I am trying to make a deal with Avnet+Xilinx to establish a custom channel for crypto miners to buy VCU1525 at short lead time.  Stay tuned in the next few days.