DIY FPGA Mining rig for any algorithm with fast ROI

[DigitalCruncher]

dudes prolly affiliated in some kind of way and gets % of the sale when you mark him as your referrer.

writing HDL that performs 10x better than a high-end GPU is questionable

but yes, a well done FPGA should perform better than a 1080ti, but not 10x

+1

I wanted to make such a project, but my investors make decisions very slowly.

There is an opinion that the GPU is too complicated to be implemented at the FPGA, and therefore it is impossible to make a good mining on the FPGA. This is a wrong opinion. In the FPGA you can make specialized instructions that are not available in the GPU. You do not need to implement the ALU with double precision and other unnecesassary stuff.

VU09P has at least x10 advantage in speed and power consumption in comparison with 1080ti. It had 1 month of ROI in late 2017-beginning 2018.

[1713939233]

1713939233

[1713939233]

1713939233

[1713939233]

1713939233

[gameboy366]

This is definitely possible but inaccessible outside US and algos mostly contain low grade shitcoin. I see a lot of coins that were considered good might die this year. It will improve once people start coming but still i will until Nvidia next gen comes out.

[kgeorgiev]

I have another question - what do you think is the residual value of these cards and is there any secondary market for them? Of course I'd love to mine with them for years but you need to have some expectations for the entire investment cycle which means thinking about the scenario when these are not profitable and you have to get rid of them somehow.

I know that you should never invest anything you can't lose in this field but still, with GPU rigs for example you always have expectation for at least 50% residual value in the worse scenarios and there is a well-established secondary market. So I'm curious what's the situation with these, as somebody who hardly knows anything about the FPGA market.

Probably zero once they have been run for a few years, almost like an asic as there is no gamer market to buy the old equipment.

Given that the ROI is in theory as low as 3-4 months, I definitely wouldn't care about resale value if I've run it for a few years. The question applies to the scenarios where everything crashes in 2-4 months or so when the device is almost new.

These are clearly not gaming devices but they obviously do exist outside of crypto mining and thus there are people using them for something. Which means there must be some secondary market and so my question was where is that market and how active is it.

[crypto4pizza]

Break even on high-end mining equipment like these and ASICs are tricky since hash will grow exponentially once there's mass adoption.

My question would be how these cards run over longer periods of time.  I believe some FPGA programs in the past ran into errors and stability issues.  Also, another concern of mine would be how these FPGA's handle being pushed hard w/crypto mining.  Would hate for such an expensive investment to go kaput after a short period of time.  I supposed the early adopters will have to be the lab rats to determine that.

[gameboy366]

This is not the first time they are being used. People made FPGA farms for bitcoin and then ASIC came out to destroy them. The cycle repeats again.

[HardwareCollector]

dudes prolly affiliated in some kind of way and gets % of the sale when you mark him as your referrer.

writing HDL that performs 10x better than a high-end GPU is questionable

but yes, a well done FPGA should perform better than a 1080ti, but not 10x

+1

I wanted to make such a project, but my investors make decisions very slowly.

There is an opinion that the GPU is too complicated to be implemented at the FPGA, and therefore it is impossible to make a good mining on the FPGA. This is a wrong opinion. In the FPGA you can make specialized instructions that are not available in the GPU. You do not need to implement the ALU with double precision and other unnecesassary stuff.

VU09P has at least x10 advantage in speed and power consumption in comparison with 1080ti. It had 1 month of ROI in late 2017-beginning 2018.

I am very open minded but at the same time, a little in the way of evidence doesn’t hurt.

Also, the blanket statement that “VU09P has at least x10 advantage in speed and power consumption in comparison with 1080ti…” is simply false.  Some algorithms will be better suited for a VU09P FPGA while others are better suited for a GTX 1080 Ti.

In any event, I am very interested and will wait for proof of hash rate before placing any large orders.

[Estex]

Stunning. But still waiting for proof

[DigitalCruncher]

I am very open minded but at the same time, a little in the way of evidence doesn’t hurt.

Also, the blanket statement that “VU09P has at least x10 advantage in speed and power consumption in comparison with 1080ti…” is simply false.  Some algorithms will be better suited for a VU09P FPGA while others are better suited for a GTX 1080 Ti.

In any event, I am very interested and will wait for proof of hash rate before placing any large orders.

Yes, you are right in this statement, but I do not want to write a long story. It seems that Ethash is only algorithm that is not suitable for FPGA.

[dhouse]

This is definitely possible but inaccessible outside US and algos mostly contain low grade shitcoin. I see a lot of coins that were considered good might die this year. It will improve once people start coming but still i will until Nvidia next gen comes out.

Lux, bayyybeeeeeeeeee!!

[rem26]

dudes prolly affiliated in some kind of way and gets % of the sale when you mark him as your referrer.

writing HDL that performs 10x better than a high-end GPU is questionable

but yes, a well done FPGA should perform better than a 1080ti, but not 10x

Why not?  GPU's are using DDR.  DDR sucks.

[HardwareCollector]

Yes, you are right in this statement, but I do not want to write a long story. It seems that Ethash is only algorithm that is not suitable for FPGA.

Not so fast, there is a guy who claimed he can do Ethash at 421.41 MH (FPGA Nallatech 385A). I am still waiting for proof, so it may be possible. Or I may be waiting for a very long time.

https://github.com/Maetti79/OpenCL_FPGA
https://github.com/Maetti79/ethminer

[saint78]

So will there not be anything in the future for the Altera/Intel variants? Anything for Stratix 10/V or Aria 10/V devices planned? @whitefire990 I can loan one for testing if need be. Just PM me

[saint78]

Yes, you are right in this statement, but I do not want to write a long story. It seems that Ethash is only algorithm that is not suitable for FPGA.

Not so fast, there is a guy who claimed he can do Ethash at 421.41 MH (FPGA Nallatech 385A). I am still waiting for proof, so it may be possible. Or I may be waiting for a very long time.

https://github.com/Maetti79/OpenCL_FPGA
https://github.com/Maetti79/ethminer

Ive tried to contact Maetti79 about that number. The only response I received was theyre making huge progress and he would give more info soon. That was about a month or so ago. It seems people get too caught up in making money to help others. I commend whitefire990 for what he's working to accomplish

[nsummy]

I like how all of these threads devolve into demanding proof and insinuating that OP is lying.  He said the software will no be released until May 30.  Considering his  revenue is based off of the devfee I don't think he would lie.  I think he did a pretty good job of recognizing that these are expensive and not for everybody, and even offering cheaper alternatives.  People should be more appreciative.  I'm guessing half the people bitching are tire kickers who couldn't even afford to buy one of these anyway.

[nsummy]

If you can show us the screenshots of your miner hashing at these rates, then I will pull the trigger and place an order for eight of these units.

For $35,000 why not fly out and meet, see it for yourself, or something of that nature?

Because he doesn't have that much money to spend on these to begin with.  No one demands proof and then spends that kind of dough by simply seeing a screenshot.  The heavy hitters already know what sort of capabilities are out there.

[TonyVH]

Following your progress OP. Thanks for sharing, ignore the nay sayers.

[Ginzink]

This is a interesting topic! Cant say i like it too much considering i have 8x 1080 ti on the way lol
But il just have to hope it does not become a too big success for a decent amount of time. Cant buy the FPGA as im in Europe, or il have ot use Digikey if that works

[hacko86]

This is a interesting topic! Cant say i like it too much considering i have 8x 1080 ti on the way lol
But il just have to hope it does not become a too big success for a decent amount of time. Cant buy the FPGA as im in Europe, or il have ot use Digikey if that works

You are not alone, but the fact that these are not kept in stock scares me the most. It could take forever to have them without a group buy.

I suggest again, someone in US could buy for Europe and ship (with escrow even from a trusted member of this forum if possible)

If anyone is interested let me know, i'd like to discuss further

[HardwareCollector]

If you can show us the screenshots of your miner hashing at these rates, then I will pull the trigger and place an order for eight of these units.

For $35,000 why not fly out and meet, see it for yourself, or something of that nature?

Because he doesn't have that much money to spend on these to begin with.  No one demands proof and then spends that kind of dough by simply seeing a screenshot.  The heavy hitters already know what sort of capabilities are out there.

Oh please spare me.

I’ve been tracking and renting these VU09P FPGAs ever since Amazon AWS had them for rental. They are currently renting for approximately $0.50/hr (f1.2xlarge -> 1 FPGA) and $4.00/hr (f1.16xlarge -> 8 FPGAs). So far, it hasn’t made any sense to actually purchase one vs renting for what I use them for.

But if someone had an FPGA bitstream which made renting these FPGAs less desirable, then purchasing would be preferred.

By the way, requesting evidence even something as trivial as a screenshot, doesn’t imply that someone is lying.   

[TonyVH]

This is a interesting topic! Cant say i like it too much considering i have 8x 1080 ti on the way lol
But il just have to hope it does not become a too big success for a decent amount of time. Cant buy the FPGA as im in Europe, or il have ot use Digikey if that works

You are not alone, but the fact that these are not kept in stock scares me the most. It could take forever to have them without a group buy.

I suggest again, someone in US could buy for Europe and ship (with escrow even from a trusted member of this forum if possible)

If anyone is interested let me know, i'd like to discuss further

I am guessing they are not in large stock is due to the fact Xilinx made these to be for development purpose. When the engineer is happy with his design then it goes to production with highly specialized custom board for their specific needs.