FPGA miner for altcoin

[rem26]

I would consider it, if it was well supported.  Xilinx or Alterra?

[1714606069]

1714606069

[1714606069]

1714606069

[1714606069]

1714606069

[1714606069]

1714606069

[arm_race]

I would consider it, if it was well supported.  Xilinx or Alterra?

Thank you for the interest. we are now custom designing pcb board for xilinx fpga.

[smoolae]

FPGA miners sound cool and scary for me at the same time (being a GPU miner). I know you can use FPGA to mine almost anything when you know how to generate a bitstream (?) for the algo you want to mine.
FPGA seems the way to go if the:

Price is normal (everyone is able to buy, not like ASICs. Keeps the network decentralized)
Hashrate is high
Power usage is really low

[arm_race]

FPGA miners sound cool and scary for me at the same time (being a GPU miner). I know you can use FPGA to mine almost anything when you know how to generate a bitstream (?) for the algo you want to mine.
FPGA seems the way to go if the:

Price is normal (everyone is able to buy, not like ASICs. Keeps the network decentralized)
Hashrate is high
Power usage is really low

the power consumption is almost guaranteed to be low due to the advancement of fpga hardware technology in recent years. between hashrate and price you need to strike a balance, however, but to me i think to make it cheap and accessible to people is more important.

[dhouse]

the usb one sounds fun as hell imo

[DigitalCruncher]

the power consumption is almost guaranteed to be low due to the advancement of fpga hardware technology in recent years. between hashrate and price you need to strike a balance, however, but to me i think to make it cheap and accessible to people is more important.

I think the board should have a PCI Express interface, at least x1. This makes easy integration with ccminer and others.
Also VCCINT's current should be at least 10A per 10000 slices, and strong decoupling.

[arm_race]

the power consumption is almost guaranteed to be low due to the advancement of fpga hardware technology in recent years. between hashrate and price you need to strike a balance, however, but to me i think to make it cheap and accessible to people is more important.

I think the board should have a PCI Express interface, at least x1. This makes easy integration with ccminer and others.
Also VCCINT's current should be at least 10A per 10000 slices, and strong decoupling.

thank you for the good advice on power supply. we are currently considering a 10A power supply to VCCINT and VCCBRAM. based on our initial power estimate the total fpga power consumption should be below 10w. for PCIE i am not sure it would fit on the miner-on-a-stick product? admittedly both of us has limited hw design experience, so we are trying to keep thing simple(and cheap)as possible 

[DigitalCruncher]

thank you for the good advice on power supply. we are currently considering a 10A power supply to VCCINT and VCCBRAM. based on our initial power estimate the total fpga power consumption should be below 10w. for PCIE i am not sure it would fit on the miner-on-a-stick product? admittedly both of us has limited hw design experience, so we are trying to keep thing simple(and cheap)as possible 

I would prefer a narrow and short PCI Express board. It is scalable and standard solution.

Also I have built some miners on generic development boards by Xilinx. But all of these boards never reach the performance of the professional level device. The dev kit  fails to cool and to withstand the ripple of the core's supply. It was very sad to see a good FPGA soldered to such restricted PCB.

I would start from XC7K160T device. Some coins would fit in it pretty well.

[arm_race]

thank you for the good advice on power supply. we are currently considering a 10A power supply to VCCINT and VCCBRAM. based on our initial power estimate the total fpga power consumption should be below 10w. for PCIE i am not sure it would fit on the miner-on-a-stick product? admittedly both of us has limited hw design experience, so we are trying to keep thing simple(and cheap)as possible 

I would prefer a narrow and short PCI Express board. It is scalable and standard solution.

Also I have built some miners on generic development boards by Xilinx. But all of these boards never reach the performance of the professional level device. The dev kit  fails to cool and to withstand the ripple of the core's supply. It was very sad to see a good FPGA soldered to such restricted PCB.

I would start from XC7K160T device. Some coins would fit in it pretty well.

we observed the same thing that the PMIC on those commercial development board is not capable of providing enough power to the FPGA with a fully unrolled pipeline design. that's one reason why we want our own custom pcb design with a beefy power supply.
thanks for the suggestion on pcie, we will consider adding it to the design.
btw, you said some coins may fit in fpga miner, can you give me some names? maybe we can expand our product to be more generic and available to support various mining algorithms.

Thanks!

[DigitalCruncher]

we observed the same thing that the PMIC on those commercial development board is not capable of providing enough power to the FPGA with a fully unrolled pipeline design. that's one reason why we want our own custom pcb design with a beefy power supply.
thanks for the suggestion on pcie, we will consider adding it to the design.
btw, you said some coins may fit in fpga miner, can you give me some names? maybe we can expand our product to be more generic and available to support various mining algorithms.

Thanks!

I am taking algorithms from the Tpruvot's ccminer. All algorithms that are using from 1 to 5 hashes are profitable in practice, like nist5 or groestle.

[arm_race]

we observed the same thing that the PMIC on those commercial development board is not capable of providing enough power to the FPGA with a fully unrolled pipeline design. that's one reason why we want our own custom pcb design with a beefy power supply.
thanks for the suggestion on pcie, we will consider adding it to the design.
btw, you said some coins may fit in fpga miner, can you give me some names? maybe we can expand our product to be more generic and available to support various mining algorithms.

Thanks!

I am taking algorithms from the Tpruvot's ccminer. All algorithms that are using from 1 to 5 hashes are profitable in practice, like nist5 or groestle.

this is valuable comment. there will be great potential if we can make it compatible working with ccminer and hw resource capable of supporting multiple algorithms.
Thanks.

[arm_race]

we observed the same thing that the PMIC on those commercial development board is not capable of providing enough power to the FPGA with a fully unrolled pipeline design. that's one reason why we want our own custom pcb design with a beefy power supply.
thanks for the suggestion on pcie, we will consider adding it to the design.
btw, you said some coins may fit in fpga miner, can you give me some names? maybe we can expand our product to be more generic and available to support various mining algorithms.

Thanks!

I am taking algorithms from the Tpruvot's ccminer. All algorithms that are using from 1 to 5 hashes are profitable in practice, like nist5 or groestle.

what do you think of having a on-board 4GB DDR4 SRAM? would it be useful for mining memory intensive algorithms? what other components could be useful addition to the fpga board?

[DigitalCruncher]

we observed the same thing that the PMIC on those commercial development board is not capable of providing enough power to the FPGA with a fully unrolled pipeline design. that's one reason why we want our own custom pcb design with a beefy power supply.
thanks for the suggestion on pcie, we will consider adding it to the design.
btw, you said some coins may fit in fpga miner, can you give me some names? maybe we can expand our product to be more generic and available to support various mining algorithms.

Thanks!

I am taking algorithms from the Tpruvot's ccminer. All algorithms that are using from 1 to 5 hashes are profitable in practice, like nist5 or groestle.

what do you think of having a on-board 4GB DDR4 SRAM? would it be useful for mining memory intensive algorithms? what other components could be useful addition to the fpga board?

The 4 GB DDR4 SRAM would be great, the right component for equihash and cryptonight. Also please consider the temperature monitoring or cooling of hottest parts.

The throttling mode is required at the high temperature of the FPGA's core. So consider the usage of System monitor in firmware.

[saint78]

Any progress on the fpga design? Also which fpga chips are you considering using? I personally think you are the right track and fgpas for altcoins will be a big market soon . Ive been trying to find any info i could on fgpas to build my own for the past couple weeks. which is how i ran across this thread. Keep us updated

[ciciteng]

I am one of the supporter to make FPGA mining comes alive again.
Owned bunch of Xilinx leftover from projects and really need to make a good use out of it 
If you need hand for testing or helping you out with debugging purposes do not hesitate to PMed me. I will be happy to give some hands ;-)

[nsummy]

I think this is a cool idea and I'm glad this got bumped.  Just like any new invention, the cost of entry will be expensive at first and the results won't be as high as they could be, but you have to crack a few eggs to make an omelette.  I see a lot of people referencing AMD GPUs here.  I'm sure the FPGA idea would be a pipe dream if there was an overabundant supply of RX 580s for $200 each.  Live by the GPU, Die by the GPU.  Smart miners look for ways to diversify.  I would definitely buy one of these.  An idea for OP would be to offer some kind of incentive in the form of future discounts and pre-releases to early adopters.

[arm_race]

Hello Everyone, sorry for lack of updates. we have been busy working all the time and paid little(no) attention to the public relationship and marketing aspects.  

We have made major breakthrough in both the HW and SW development. it has been found that the power supply circuit on any of the off-the-shelf FPGA development boards is way insufficient to support any crypto mining algorithm running at a decent speed. we made a board modification to use different power supply chip which can deliver up to 25A of current to FPGA device. with this mighty power circuit mod, we managed to achieve the following fpga design goals on our development board:

4/21 - Developed and verified two Altcoin mining algorithms, both can run at 2/3 hashrate of a mid-class GPU (use RX470 as a reference), with only 1/10 of power consumption. The whole board power consumption is measured at 12W. But this is my big hack-up board. Our final slimmed down version board would consume even less power.

Currently, we have been testing on a development board with early Engineering Sample FPGA device. This specific FPGA device is scheduled to be in full production in second half of 2018. We are actively negotiating with the sales and distributors trying to get this chip sooner and at an affordable price for customers.

Remember our design goals are:
1. Affordable(cheap) and accessible to EVERYONE, not the selected few. We'd like to contribute to the decentralization of the crypto, not the other way. because we believe crypto belongs to the people not big mining farms.
2. Active after-launch development to continuously add support for multiple Altcoin mining algorithms based on the crowd demand.
3. Home mining friendly. not an excessive heat radiator or noise generator.. like those antminers (i used to have a few in my basement and traumatized by that experience )
4. Light wight and mobility. plug in and mine anywhere with your laptop.

Most importantly, I'd like to hear what you want, give us some ideas from your pipe dream and we will try our best to design it in.   Thanks.

[MA3A]

Talking about ASIC's and FPGA's - not sure what SAMSUNGis producing and for whom, but INTEL, just announced they got FASTEST IN THE WORLD FPGA:
https://newsroom.intel.com/news/intel-chip-performs-10-trillion-calculations-per-second/

- get those properly programmed and stacked up in 1 miner and i bet it will beat anything thats out there right now. Interesting point that we may be seeing high end FPGA based miners to come out soon, which can be adopted to forks. BEWARE Wink

[DigitalCruncher]

Most importantly, I'd like to hear what you want, give us some ideas from your pipe dream and we will try our best to design it in.   Thanks.

Very good progress.
It would be nice to implement PCI Express x1 at least, and allow the development of third-party firmware.

[b9ron]

Exactly :

- PCIe like HPC accelerator card but without network port at affordable price : $300-$400
- And with OpenCL support too.