DIY FPGA Mining rig for any algorithm with fast ROI

[colomine]

200 day ROI is still very good IMHO.
Also people should realize the low power usage for this return. My 24*1070ti eats (and heats) up 2600w/h
Two BCU will make more money than my whole setup for around 300w...

@melpheos,
As one of those people ) who should understand ), I would disagree with you. I understand that anybody paying $0.12+/kwh  will finally see some gains and FPGAs are a big hope. That is not in question. I  myself because of my personal situation  almost ended up ordering FPGAs to an EU address. But later decided not to. And of course, a  GPU mining in EU at $0,18+ per kwh is not an option and FPGA is a dealbreaker here.
But sometimes there are better options available. It all depends on your electricity costs and the hardware prices to make profits even  on GPU mining. For my current location, if I buy GPUs at todays dropped prices or second hand, it will result in the same or better ROI as the current ROI on 0xBTC we are seeing on FPGAs. Which is around 10 USD per FPGA card.  But GPUs have better warranty and are explored well enough to pedict what to expect whereas the FPGAs have some drawbacks and there is more uncertainty there.  
In this situation, ofcourse I have hopes for a better ROI than with GPUs. I hope you could follow my thought.

I agree venture capitalists look for a ROI under 3 years.

@UOTT2
That is why even the GPU mining on a larger scale is still profitable, provided the prices are right. And this is exactly why it is a concern when such an expensive piece of hardware as BCU1525 has a only  90 days warranty. It is a bigger risk than a regular GPU when used for mining. One of the thoughts that crossed my mind was that it is safer to mine on GPUs because of that.
@whitefire990,
Yes of course, we are all having big hopes for FPGAs, and clearly the drop in profitability has to do with altcoins course drop. 0xBTC alone went from $0,82 2 days ago to #0,60 something.  
I look forward to those new bitstreams you entioned on discord to be released in September. I only wish the warranty was comparable to that of GPUs. I would do a immersion cooling if not the warranty. But not I do not want to risk hosting on my own befor having tested it with mineority

[1715122597]

1715122597

[1715122597]

1715122597

[1715122597]

1715122597

[1715122597]

1715122597

[1715122597]

1715122597

[1715122597]

1715122597

[wevieNL]

Any comments on the following proposed set of changes?

https://github.com/monero-project/monero/pull/4218

[whitefire990]

Okay I posted a major update:
http://zetheron.com/index.php/downloads/

Now posted are bitstreams for Bittware cards (XUPP3R, XUPVV4) and I also posted an incredible 773MHz 17GH/s 0xToken bitstream with instructions on how to run it.

There is also a chart of upcoming algorithms & projected dates.  Most, but not all, upcoming algorithms are still being held secret.

There is also a detailed current table posted on the downloads page that shows the core current vs. different configurations.

[whitefire990]

Any comments on the following proposed set of changes?

https://github.com/monero-project/monero/pull/4218

These changes would/will significantly decrease the speed gain on an FPGA vs. a Vega-64 on the Cryptonight algorithms, which is why I (personally) am not investing any of my team's time into cryptonight algorithms.  My goal for September is to have a total of 7 algorithms that can profitably support over 6,000 FPGA's without resorting to any cryptonight or cryptonight variant algorithms.

Profit on 0xBTC is dropping because the coin is dropping; the current set of algorithms being developed for September (if launched today) would generate $12-$20 per FPGA per day, despite the extremely poor market state.

[fr4nkthetank]

"Only in the crypto world is an ROI in under a year for equipment "too long."

Yeah...theres a good reason for that   Anyways once everyone gets their 5000 BCUs then good luck with the difficulty on the coins.  At least they will have some more mainstream algos because all the small ones will have their profit reduced to rubble in the end.

[melpheos]

200 day ROI is still very good IMHO.
Also people should realize the low power usage for this return. My 24*1070ti eats (and heats) up 2600w/h
Two BCU will make more money than my whole setup for around 300w...

@melpheos,
As one of those people ) who should understand ), I would disagree with you. I understand that anybody paying $0.12+/kwh  will finally see some gains and FPGAs are a big hope. That is not in question. I  myself because of my personal situation  almost ended up ordering FPGAs to an EU address. But later decided not to. And of course, a  GPU mining in EU at $0,18+ per kwh is not an option and FPGA is a dealbreaker here.
But sometimes there are better options available. It all depends on your electricity costs and the hardware prices to make profits even  on GPU mining. For my current location, if I buy GPUs at todays dropped prices or second hand, it will result in the same or better ROI as the current ROI on 0xBTC we are seeing on FPGAs. Which is around 10 USD per FPGA card.  But GPUs have better warranty and are explored well enough to pedict what to expect whereas the FPGAs have some drawbacks and there is more uncertainty there.  
In this situation, ofcourse I have hopes for a better ROI than with GPUs. I hope you could follow my thought.

I agree venture capitalists look for a ROI under 3 years.

@UOTT2
That is why even the GPU mining on a larger scale is still profitable, provided the prices are right. And this is exactly why it is a concern when such an expensive piece of hardware as BCU1525 has a only  90 days warranty. It is a bigger risk than a regular GPU when used for mining. One of the thoughts that crossed my mind was that it is safer to mine on GPUs because of that.
@whitefire990,
Yes of course, we are all having big hopes for FPGAs, and clearly the drop in profitability has to do with altcoins course drop. 0xBTC alone went from $0,82 2 days ago to #0,60 something.  
I look forward to those new bitstreams you entioned on discord to be released in September. I only wish the warranty was comparable to that of GPUs. I would do a immersion cooling if not the warranty. But not I do not want to risk hosting on my own befor having tested it with mineority

I hear you and have absolutely no issue with your argumentation

[wudafuxup]

"Only in the crypto world is an ROI in under a year for equipment "too long."

Yeah...theres a good reason for that  Anyways once everyone gets their 5000 BCUs then good luck with the difficulty on the coins.  At least they will have some more mainstream algos because all the small ones will have their profit reduced to rubble in the end.

A huge part of that logic is the fact that technology advances so fast. Making ROI very difficult for older architectures/nm processes. Look at the antminer D3, many people bought at the wrong time and never ROIed since the tech train never stops moving.

Look at the Baikal Giant N for example. They were selling them each for 3800USD in the first couple weeks. A boat load of people bought when all the CryptoNight ASICs were being launched and will never ROI (not even close). They are selling for 200-400USD now 3 months later since the networkhash went up 10X on certain coins.

I highly doubt FPGAs are the future of mining. There's a reason it hasn't been so for the past decade. I do however think FPGAs are in a pretty good spot as of today (we are in the middle of a transition point from GPU to ASIC) but once coins stop forking against ASICs (ASICs are the inevitable future) will be the end of FPGAs. I personally don't see the hype behind it, if anything the bitstream devs are going to be the real winners here.

Look at the top 20 coins today compared to 2 years ago.. Almost every single one that is mineable is an ASIC algo. A couple years ago it was about half and half. ASICs are clearly taking over. I think FPGAs will shine for a short lived while.

[colomine]

 @ developers
I was doing some aritmetics finding out what algorithms could  remain reasonably profitable to mine when mined with 5000+FPGAs. According to the Zetheron website  currently the only algorithm that can support first 2 batches of cards (10 000 cards cominп out this year) is Cryptonight. Digital Cruncher  also mentioned on discord that coins that are not mined by FPGA farms and ASICs would be best for FPGAs to mine. Can somebody  give some hint or provide a  link regaridng what coins or algorithm that would apply to with ROI of 200 % or better.  I rememberthere was a formula on the Zetheron webpage but I can not find it now.

PS:I am also trying to think how this all would work with so many FPGAs mining.   Are we going to form a professional union )) or do it the wild west style- "who dies first looses". Or will people come up with software that switches alorythms on the fly (if it is even possible to do in timely manner on an FPGA in the way it is on a GPU).
 
 

[philipma1957]

"Only in the crypto world is an ROI in under a year for equipment "too long."

Yeah...theres a good reason for that  Anyways once everyone gets their 5000 BCUs then good luck with the difficulty on the coins.  At least they will have some more mainstream algos because all the small ones will have their profit reduced to rubble in the end.

A huge part of that logic is the fact that technology advances so fast. Making ROI very difficult for older architectures/nm processes. Look at the antminer D3, many people bought at the wrong time and never ROIed since the tech train never stops moving.

Look at the Baikal Giant N for example. They were selling them each for 3800USD in the first couple weeks. A boat load of people bought when all the CryptoNight ASICs were being launched and will never ROI (not even close). They are selling for 200-400USD now 3 months later since the networkhash went up 10X on certain coins.

I highly doubt FPGAs are the future of mining. There's a reason it hasn't been so for the past decade. I do however think FPGAs are in a pretty good spot as of today (we are in the middle of a transition point from GPU to ASIC) but once coins stop forking against ASICs (ASICs are the inevitable future) will be the end of FPGAs. I personally don't see the hype behind it, if anything the bitstream devs are going to be the real winners here.

Look at the top 20 coins today compared to 2 years ago.. Almost every single one that is mineable is an ASIC algo. A couple years ago it was about half and half. ASICs are clearly taking over. I think FPGAs will shine for a short lived while.

If asics are the inevitable future of all mining and asic builders flood all coins with endless gear mining dies.

So your prediction above is bleak.

If FPGA gain hold and build out crushing all coins mining dies.

Thinking that gpus get pushed to the curb completely is the same as thinking mining will die.

Monero development team. Will be forking again. Thank god at least one development team understands that gpu mining is essential for overall crypto business to survive.

This does not mean all mining needs to be gpu.

It means some mining needs to be gpu.

This is simple economics .

It also means some coins should be FPGA

And some need to be asic.

Most of the worlds  society’s  have bicycles motorcycles cars and trucks

All have a purpose to move people and things.

Tossing out gpu Mining will break this system much the same that tossing bicycles out would hurt society.

So far BTG and Monero did forks
And now monero is doing a second fork.

We need this to encourage gpus to mine. 

[whitefire990]

The upcoming Monero fork is being done by smart people who really understand how to make stuff 'good for GPU's.'

At the same time, there is a way for all ASICs to be defeated forever.  Since X16R is faster on an FPGA than any ASIC could ever be (at least from a hardware investment point of view), then if every coin forked to X16R, all ASICs would die.  However, I agree that we still need GPU mining.  So some coins should do what Monero is doing.  And I stand by what I have said before that if a coin wants be fpga-friendly and asic-proof, X16R is the way to go.

Another algorithm that is highly-friendly to GPU's (and extremely hard on fpga/asic) is Wild Keccak.

[wevieNL]

The upcoming Monero fork is being done by smart people who really understand how to make stuff 'good for GPU's.'

At the same time, there is a way for all ASICs to be defeated forever.  Since X16R is faster on an FPGA than any ASIC could ever be (at least from a hardware investment point of view), then if every coin forked to X16R, all ASICs would die.  However, I agree that we still need GPU mining.  So some coins should do what Monero is doing.  And I stand by what I have said before that if a coin wants be fpga-friendly and asic-proof, X16R is the way to go.

Another algorithm that is highly-friendly to GPU's (and extremely hard on fpga/asic) is Wild Keccak.

What about the upcoming HBM2 FPGA's? Does that drastically change the playing field ?

[mike_la_jolla]

What about the upcoming HBM2 FPGA's? Does that drastically change the playing field?

Probably not.  You get 20x performance improvement on the DDR4<->FPGA interface but no improvement anywhere else.  So the algorithm you are accelerating must require high amounts of sequential access to a large memory bank.  And any algorithm could be made resistant to acceleration by an HMB/FPGA merely by requiring >16GB of memory.

... And the present price of the HBM devices is *shocking* ...  Better from Intel/Altera than Xilinx, but still more than this market can tolerate.  Also, Intel/Altera has not executed well in the last 2-3 years.  Getting a Stratix10 device has been a bit of a joke in the FPGA world.

[mfurman]

BCU-1525 delayed to the fourth quarter

[COINKING7]

BCU-1525 delayed to the fourth quarter

Very disappointed

[COINKING7]

Will they be offering refunds?

This is a direct message I received prior to buying 1 unit:
''Shipping starts in August. You should have it mid-late August depending on transfer times.''

Now we don't even have a date???

This is ridiculous.

[gotminer]

Glad I didn't buy into the FPGA stuff a month ago, to be honest.  I mean, I like the fact that people are trying ... These guys need to keep trying.  The headway that you are making isn't ready for mass mining market, that's for sure.

[gotminer]

Not sure if i remember correctly but you might have some luck checking out the FPGA discord group: (https://discord.gg/25DpDCS). Not sure if thats where i saw it, but i once saw downloads links to bitstreams from that or some other discord group.

Anyway, just stumbled upon the Zetheron website recently. Congratulations to whitefire, senseless, GPUhoarder, and everyone else that has pushed this to succeed. Hopefully this is a step towards further decentralization away from ASICS.

I'll check it, of course, thanx. Don't get me wrong, but I hope that 4% fee is high enough to make all possible downloads easily accessible for each potential donator (aka cryptoFPGA end user)...

Should be way less.  Like a lot less.

[COINKING7]

They have proven that they can't meet deadlines...sad

[whitefire990]

They have proven that they can't meet deadlines...sad

Xilinx makes FPGA's, not circuit boards.  They make the best FPGA's but they are not used to making the boards that actually use those FPGA's.  They usually let skilled 3rd parties like Bittware make the actual boards.  In this case Xilinx (I guess) figured they could make more by just making the boards themselves, and didn't realize you actually have to do some planning first.  The Bittware group order is in the works as we speak, they are used to making actual circuit boards, although it is unlikely the Bittware group order will arrive earlier than the Xilinx group order which had a big head start.

[lutsifer]

I am looking very long time where to buy a good fpga, can anyone help me please?

Thank you!