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

[Miningtry]

I have serious question here, why devs from cryptonight didn't fork algo months ago when they sense sudden hashrate increase.

[bitcoinexplorer]

i am not buying unless they bring down price and remove MOQs

[Mattthev]

I have serious question here, why devs from cryptonight didn't fork algo months ago when they sense sudden hashrate increase.

Monero hardforks are every 6 months...

[bitcoinexplorer]

I have serious question here, why devs from cryptonight didn't fork algo months ago when they sense sudden hashrate increase.

Monero hardforks are every 6 months...

so will baikal update every 6 months to match monero?

[Mattthev]

I have serious question here, why devs from cryptonight didn't fork algo months ago when they sense sudden hashrate increase.

Monero hardforks are every 6 months...

so will baikal update every 6 months to match monero?

They haven't respond if they wil update firmware. If it's FPGA it should be possible.
We still need to wait for final algo changes, I saw some changes on XMR-STAK PR from someone, but Monero team probably haven't released anything final.

Edit: XMRig accept PR and made needed changes for the algo change.

[sylabis]

I really wish people would stop calling FPGA's ASICs.  Baikal (so far) has only made FPGA mining rigs.  There is a huge difference between an FPGA and an ASIC.  An FPGA is not that much different from a GPU.  Anyone can get one.  Anyone can buy a Zynq FPGA board from digikey for $89-$199, or a higher end one for more money, and if you take a little while to learn how to program it, you can hash any algorithm except equihash & ethash.  Furthermore, your ROI will be better than a GPU in almost every case, in some cases dramatically better (as Baikal showed with the X10 and Giant-B).

Are you really sure that Baikals X10/B are FPGA? (Which FPGA, by the way?) While it is enterely possible, technical characteristics of these products are not typical for FPGA. Especially the low power consumption.

As for ROI I wouldn't expect good ROI from entry and mid-range FPGA boards because of their weak power supplys. We need propertly designed professional grade DC/DC for core voltage, that is rarely seen in practice.

There is nothing suspicious about the Giant-N.  I was already working on an FPGA cryptonight miner before the Giant-N was announced (and obviously I am now focusing on other algorithms).  The power of 60W is realistic for one FPGA accessing many external SRAM's.  Unlike DRAM, SRAM consumes very little power.  The fundamental nature of Cryptonight is that it uses almost no number crunching (by design).  A single FPGA just accesses many parallel SRAM's and these memory accesses do not consume a great deal of power.  FPGA's consume way less power than other mining devices already.  Consider the X10 burns 250-500W and makes the same amount per day as a 2000W GPU rig.  Some algorithms burn more, some burn less, and algorithms that have no number crunching (like Cryptonight) burn the least.  The reason a Vega 56/64 burns so much power on cryptonight is because it is using high bandwidth external memory, a totally different approach than using many SRAM's in parallel.

FPGA's can be reconfigured very quickly.  It is true that certain PCB designs and part selections are better at some algorithms than others.  But it doesn't matter if Monero does a hard fork, you can still just use an FPGA to mine the new algorithm, ad infinitum.  As I mentioned before, only Ethash is truly resistant to FPGA's.  As Baikal has more and more FPGA miners on the market with different types of FPGA's and RAM (Giant-B, X10, Giant-N), a coin which 'forks' would have to know the exact internal configuration of every FPGA mining rig on the market to 'avoid' a new algorithm which could be efficiently mined by them.  To give an example, there is a decent chance that Monero's new algorithm could be (accidentally) mineable by the Giant-B or Giant-X10 or Giant-N, and all Baikal has to do is release new bitstreams (firmware for the SD card) that would update those rigs to mine the new algorithm.

As the number of different FPGA rigs on the market continues to increase, it would be very difficult to fork to an algorithm that would be immune to those rigs, unless you pick an Ethash style algorithm.  Furthermore if you add in all the cheap FPGA boards available from companies like Digikey, Avnet, Xilinx and Intel, then there is ALREADY a mass produced FPGA board that can do any algorithm efficiently except Ethash.

FYI the Monero ASIC statement is specific to ASIC's.  They specifically say they want to avoid ASICs mining their coin (they speak of FPGA's more favorably, and separately from ASICs).  Since the Giant-N is an FPGA rig, it doesn't actually fall into the category of something they would fork away from.  Furthermore, the Giant-N hash rate is not devastating to GPU's.  It has a slightly better ROI than Vega's, but in no way do Vega's become obsolete.  Baikal would have to ship out 100,000 Giant-N's to truly disrupt the Cryptonight networks, which is unlikely.

(BTW I bought 2 Giant-N from a local reseller in Vancouver.  The units are supposed to arrive on Monday.)

If the Baikal Giant N is an FPGA that could resynthesize from a hardware update and be effective after the V7 Fork, then why are Baikal only selling them for $3600 per unit? And why are they no longer marketing them as being compatible with Monero? Is the public sale of these units weeks before the V7 fork just a coincidence? For DIY setups like you're talking about then FPGA is the only viable option, however, if you have millions to invest into mass production then the cost per unit to produce an ASIC is going to be much less than an FPGA.  Baikal have their own propriety development, they don't just make FPGAs, they make multi algo ASICS. 

Also, from the  2018-03-04 Dev meeting:

12:22 PM <knifeofpi> somebody made an FPGA
12:22 PM <knifeofpi> but ASICs...?
12:22 PM <@fluffypony> somebody taped out an ASIC
12:22 PM <knifeofpi> when did this happen?
12:22 PM <medusa_> the reddit post you mean ?
12:22 PM <rehrar> link?
12:22 PM <@fluffypony> sidechannelled to me, not public
12:22 PM <@fluffypony> a handful of others have had similar confirmation via via
12:22 PM <psychocrypt> @fluffypony: are this ASICS or FPGAs?
12:23 PM <@fluffypony> were it just me I would find it suspicious
12:23 PM <@fluffypony> psychocrypt: ASICs, not FPGAs

[Tim.Yoshi]

https://www.hongkongcompanygo.com/hongkong?utm_term=Seariver-Industrial-Limited&utm_source=2586774

What is most suspicious: "SEARIVER INDUSTRIAL LIMITED was incorporated on[ 03-OCT-2017 as a Private company limited"
So it's literally "just" created company. Maybe they are just avoiding taxes this way?

[damNmad]

This is total RIPOFF for all the GPU miners, these money mongers will do anything they possibly can to make the cryptos centralised.

Cryptonight is the only algo many small miners, like a PC holders are able to mine it, but these guys already made ASICS out of them, their agenda is simple, make rich richer.

[tboy32c]

I really wish people would stop calling FPGA's ASICs.  Baikal (so far) has only made FPGA mining rigs. ...

How can you tell the difference?

You probably can't without physically inspecting the chips used.

This is true, and besides that, the fact that Baikals can theoretically be flashed to hash other algorithms gives it away. The Baikal miners being FPGAs also meant that they were less efficient than the D3 X11 ASICS when the D3s first came out. FPGAs are typically weaker than ASICs.

As far as I know the Giant series have never actually been "upgraded" to support additional algorithms, is that right? I realize Baikal advertises the X10 as having that ability, but that's the only hint I've seen that they may be FPGAs. Seems to me like they're ASICs unless proven otherwise by an update from Baikal or possibly a decapping inspection by someone with the knowledge and equipment to do so.

[Spill]

I have serious question here, why devs from cryptonight didn't fork algo months ago when they sense sudden hashrate increase.

It wasn't sudden, it was a slow roll since september,   I noticed it when i stopped mining monero back in october when the unknown mined blocks was only like 40% back then,  but it made monero far less profitable so i switched back to duel mining eth.  For the past 3 months Baikal has held it
at 75% of all the new blocks,  I guess they thought no one would notice.   They are only selling them now cause of the monero fork
coming in 2 weeks,  hopefully  the fork will make these miners worthless space heaters.

[bovineplane]

I really wish people would stop calling FPGA's ASICs.  Baikal (so far) has only made FPGA mining rigs. ...

How can you tell the difference?

You probably can't without physically inspecting the chips used.

This is true, and besides that, the fact that Baikals can theoretically be flashed to hash other algorithms gives it away. The Baikal miners being FPGAs also meant that they were less efficient than the D3 X11 ASICS when the D3s first came out. FPGAs are typically weaker than ASICs.

As far as I know the Giant series have never actually been "upgraded" to support additional algorithms, is that right? I realize Baikal advertises the X10 as having that ability, but that's the only hint I've seen that they may be FPGAs. Seems to me like they're ASICs unless proven otherwise by an update from Baikal or possibly a decapping inspection by someone with the knowledge and equipment to do so.

They never upgrade. If the upgrade exists it will not be upgraded until it becomes less or not profitable to use it themselves.

[JohnSch]

I really wish people would stop calling FPGA's ASICs.  Baikal (so far) has only made FPGA mining rigs.  There is a huge difference between an FPGA and an ASIC.  An FPGA is not that much different from a GPU.  Anyone can get one.  Anyone can buy a Zynq FPGA board from digikey for $89-$199, or a higher end one for more money, and if you take a little while to learn how to program it, you can hash any algorithm except equihash & ethash.  Furthermore, your ROI will be better than a GPU in almost every case, in some cases dramatically better (as Baikal showed with the X10 and Giant-B).

ASIC's on the other hand are NOT available to everyone.  First you need the software (Synopsys, which costs $500K), then you need at least $3 million USD for the first batch of chips (assuming you can find a billion dollar fab that wants to run your project), and most likely the first revision fails and needs at least another $3 million for another revision.

So:
CPU's + GPU's + FPGA's = available for everyone, can be programmed by anyone with extremely low cost or free tools
ASIC = extremely expensive and not feasible for an individual

I'm working on my own FPGA rig and I suggest other people do the same.  It also allows you to stay ahead of the curve, especially on smaller altcoins.  Baikal is good at making FPGA mining equipment, good for them.  Embrace change and advancement. 

And for those who are wondering, 60W for 20,000 hash on Cryptonight is absolutely feasible for a single FPGA with multiple external SRAM's.

And for those who want to develop their own FPGA rigs, make sure to analyze the algorithm(s) you want to implement, and choose the best board for the task in terms of the amount of internal memory the FPGA has vs. the amount of logic cells & DSP slices.

I am also working on a FPGA miner,  what algorithm are you working on?  

I have taken apart both the Giant B and the Giant X10 apart myself.  The chips are definitely ASIC because they have the Baikal logo etched on to them and an unknown model number(I will post pics if anyone wants them).  Even if Baikal for some reason etched thier own logo on an FPGA, the chips are very tiny compared to any FPGA's I have ever seen.  Most high-end FPGA's are the size of a CPU. Also, FPGA's do not perform 100 times better than a GPU (like most ASICs).  They do at the best ~8 times the performance of a GPU with a $3000 FPGA.

Just because Baikal miners support multiple algorithms doesn't mean they are FPGA's.  ASICs can support multiple Different Algorithms but it takes away die space which makes the miners not perform as fast it would with 1 algorithm.  This is part of the reason Bitmain's A3 does so much better than the Baikal B (at siacoin), because Bitmain A3 has the entire ASIC die dedicated to Siacoin.   If you still don't believe me, look at Dash miners, the x11 algorithm is comprised of 11 different algorithms that can fit in one ASIC (which includes Skein).

Some of you maybe wondering how they have updated their Baikal x10 to support new algorithms.  Baikal probably has already developed them from square one (or atleast knows which one they are doing) but is waiting for the current Algorithms to be unprofitable.  I almost gurantee that one of the next algorithms for the x10 will be Nist5 becuase Nist5 uses 5 algorithms that are used in x11 except they are 512 bit. which is Keccak512, Blake512, Skein512, Gr0estl512 and JH512.

Back to the Baikal N:  I don't think a 20 kh/s cryptonight FPGA miner is feasible because there is too much latency/bandwidth involved in offloading memory into external sram's.  It would however be possible to use a high-end FPGA chip like the Virtex-7.  The Virtex-7 has about 68mb of internal block ram so you could put ~34 really fast monero cores but I don't know if it would get 20khs(but maybe).  These chips cost over $2000 each and would be unreasonable for Baikal to use them.  


Those who are mining monero with GPU, you can sleep soundly, baikal can't update their miner.

[DigitalCruncher]

I think you are right, the speculation that X10 and -B products are FPGA is not true. One can see the thread about overclocking Baikal units somewhere at this forum to find the proof.

I have built Lbry and Myriad-Groestl  algos inside of 28 nm FPGA. These allgos are hot - about 80 C and power hungry - near 40 W. The hashrate was about 200 MH/s in both cases. Unfortunately some month after that -B had been released and these two bitstreams became history.

But one thing Baikal have not done yet to bury altcoin even lower by their units. [And GPU miners should pray..]

As for Baikal-N I would not be so sure. The cryptonight is definetly feaseble in FPGA. Personally I am too lazy to implement it, because other altcoins are much easier to implement.If the hashrate will go down after the fork, it would be reasonable to build Monero in FPGA anyway.

Also the price of FPGA may become quite low in mass quantities, and even lower when using refurbished parts.

[JohnSch]

I think you are right, the speculation that X10 and -B products are FPGA is not true. One can see the thread about overclocking Baikal units somewhere at this forum to find the proof.

I have built Lbry and Myriad-Groestl  algos inside of 28 nm FPGA. These allgos are hot - about 80 C and power hungry - near 40 W. The hashrate was about 200 MH/s in both cases. Unfortunately some month after that -B had been released and these two bitstreams became history.

But one thing Baikal have not done yet to bury altcoin even lower by their units. [And GPU miners should pray..]

As for Baikal-N I would not be so sure. The cryptonight is definetly feaseble in FPGA. Personally I am too lazy to implement it, because other altcoins are much easier to implement.If the hashrate will go down after the fork, it would be reasonable to build Monero in FPGA anyway.

Also the price of FPGA may become quite low in mass quantities, and even lower when using refurbished parts.

Which FPGA do you think would run Cryptonight the best? 

[oomurashin]

Asicmarket.com have not ship and nothing replay......

[Kalyst69]

There is also this model from PinIdea

http://winkor88.en.alibaba.com/product/60742335573-0/PinIdea_RR_200.html

27KH/s for 350W

[DigitalCruncher]

Which FPGA do you think would run Cryptonight the best? 

Personally I prefer large 16 nm FPGA. You can build almost any altcoin except ETH.
I am looking at the directed graph of operations in Cryptonight algorithm, and I don't like it. The 128 bit multiplier and AES blocks have high latency and therefore the storage of the 2 MB contexts must be done in the external memory. The good news are that logical resources are almost free and it probably possible to build dual miner Monero+nist5. Also DSP blocks are suitable to build triple mode miner XMR+nist5+some another coin. So I prefer diversification, although the top-range FPGA part could be not optimal for solo XMR mining.

I have seen the messages in the discussion of the XMR V7 POW selection. Some man advises to use high bandwidth memory with serial interfaces. I would agree, but it could be complicated and expensive solution.
Maybe HyperRAM is also usable, I like low cost and simple interface of this memory,

[JohnSch]

Which FPGA do you think would run Cryptonight the best? 

Personally I prefer large 16 nm FPGA. You can build almost any altcoin except ETH.
I am looking at the directed graph of operations in Cryptonight algorithm, and I don't like it. The 128 bit multiplier and AES blocks have high latency and therefore the storage of the 2 MB contexts must be done in the external memory. The good news are that logical resources are almost free and it probably possible to build dual miner Monero+nist5. Also DSP blocks are suitable to build triple mode miner XMR+nist5+some another coin. So I prefer diversification, although the top-range FPGA part could be not optimal for solo XMR mining.

I have seen the messages in the discussion of the XMR V7 POW selection. Some man advises to use high bandwidth memory with serial interfaces. I would agree, but it could be complicated and expensive solution.
Maybe HyperRAM is also usable, I like low cost and simple interface of this memory,

I do not understand why you could not just use the internal memory located inside the FPGA.  Many new 16nm Fpga has upwards of 75 megabyte of memory.  I would think having 37 really fast cores would be better than trying to use slow external ram.

Anyways, I found a reddit post of someone who supposedly did 20 kh/s on an monero FPGA miner. https://www.reddit.com/r/MoneroMining/comments/7s9zwe/fpga_mining/

Could you give a link to the XMR V7 discussion?

[sylabis]

I really wish people would stop calling FPGA's ASICs.  Baikal (so far) has only made FPGA mining rigs.  There is a huge difference between an FPGA and an ASIC.  An FPGA is not that much different from a GPU.  Anyone can get one.  Anyone can buy a Zynq FPGA board from digikey for $89-$199, or a higher end one for more money, and if you take a little while to learn how to program it, you can hash any algorithm except equihash & ethash.  Furthermore, your ROI will be better than a GPU in almost every case, in some cases dramatically better (as Baikal showed with the X10 and Giant-B).

ASIC's on the other hand are NOT available to everyone.  First you need the software (Synopsys, which costs $500K), then you need at least $3 million USD for the first batch of chips (assuming you can find a billion dollar fab that wants to run your project), and most likely the first revision fails and needs at least another $3 million for another revision.

So:
CPU's + GPU's + FPGA's = available for everyone, can be programmed by anyone with extremely low cost or free tools
ASIC = extremely expensive and not feasible for an individual

I'm working on my own FPGA rig and I suggest other people do the same.  It also allows you to stay ahead of the curve, especially on smaller altcoins.  Baikal is good at making FPGA mining equipment, good for them.  Embrace change and advancement. 

And for those who are wondering, 60W for 20,000 hash on Cryptonight is absolutely feasible for a single FPGA with multiple external SRAM's.

And for those who want to develop their own FPGA rigs, make sure to analyze the algorithm(s) you want to implement, and choose the best board for the task in terms of the amount of internal memory the FPGA has vs. the amount of logic cells & DSP slices.

I am also working on a FPGA miner,  what algorithm are you working on?  

I have taken apart both the Giant B and the Giant X10 apart myself.  The chips are definitely ASIC because they have the Baikal logo etched on to them and an unknown model number(I will post pics if anyone wants them).  Even if Baikal for some reason etched thier own logo on an FPGA, the chips are very tiny compared to any FPGA's I have ever seen.  Most high-end FPGA's are the size of a CPU. Also, FPGA's do not perform 100 times better than a GPU (like most ASICs).  They do at the best ~8 times the performance of a GPU with a $3000 FPGA.

Just because Baikal miners support multiple algorithms doesn't mean they are FPGA's.  ASICs can support multiple Different Algorithms but it takes away die space which makes the miners not perform as fast it would with 1 algorithm.  This is part of the reason Bitmain's A3 does so much better than the Baikal B (at siacoin), because Bitmain A3 has the entire ASIC die dedicated to Siacoin.   If you still don't believe me, look at Dash miners, the x11 algorithm is comprised of 11 different algorithms that can fit in one ASIC (which includes Skein).

Some of you maybe wondering how they have updated their Baikal x10 to support new algorithms.  Baikal probably has already developed them from square one (or atleast knows which one they are doing) but is waiting for the current Algorithms to be unprofitable.  I almost gurantee that one of the next algorithms for the x10 will be Nist5 becuase Nist5 uses 5 algorithms that are used in x11 except they are 512 bit. which is Keccak512, Blake512, Skein512, Gr0estl512 and JH512.

Back to the Baikal N:  I don't think a 20 kh/s cryptonight FPGA miner is feasible because there is too much latency/bandwidth involved in offloading memory into external sram's.  It would however be possible to use a high-end FPGA chip like the Virtex-7.  The Virtex-7 has about 68mb of internal block ram so you could put ~34 really fast monero cores but I don't know if it would get 20khs(but maybe).  These chips cost over $2000 each and would be unreasonable for Baikal to use them.  


Those who are mining monero with GPU, you can sleep soundly, baikal can't update their miner.

Thanks for your post. Finally someone nails it.  The premise that Baikal only makes FPGA's is Fud.

[DigitalCruncher]

I do not understand why you could not just use the internal memory located inside the FPGA.  Many new 16nm Fpga has upwards of 75 megabyte of memory.  I would think having 37 really fast cores would be better than trying to use slow external ram.

Anyways, I found a reddit post of someone who supposedly did 20 kh/s on an monero FPGA miner. https://www.reddit.com/r/MoneroMining/comments/7s9zwe/fpga_mining/

Could you give a link to the XMR V7 discussion?

It is true about internal memory, yes, tens of megabytes and dozens of cores. But what is the overall hashrate of the solution using internal memory? Data dependancy and high latency make memory stalled almost all the time. I suppose the external memory would hide the latency of AES and multiplier.

Anyway the practice is the cretery of truth. If 20KH/s miner is possible - OK, lets build it. I have not done the IP core of XMR yet, so my estimations are rough and careful.

So I do not change my answer - top 16 nm FPGA and dual-triple coin minning in mind is reasonable choice for miner.

I have to find the the link, it is somewhere in browser history.