Topic: DIY FPGA Mining rig for any algorithm with fast ROI - page 89. (Read 99472 times)
Looks like Bittware reduced their standard warranty for the "crypto" models. I believe they typically have 1 yr mfg warranty on their products. Kind of disappointing the warranty was reduced that much for a $6k product.
I am saying it makes no financial sense. The major GPU coins already have asics, with ETH being the biggest. And those coins that have no asic on it will have asics on it before the FPGA will come close to ever breaking even.
And everytime an asic takes over an alternate coin, all the GPUs and FPGAs that were mining that coin will go to a non-asic coin. That means the remaining non-asic coins will have difficulty skyrocket and guess wat, that means your GPUs and FPGAs mining make less money. Everytime a new asic comes out, the FPGAs make less money. So, there is no niche. There is a limited amount to be earned from mining u see. If the asics keep appearing to take some of it, the switchers (FPGAs/GPUs) will have less and less.
Why not u calculate how much it would cost to make an FPGA that can do the equivalent of mining 504mh like the L3+ or 14TH like the S9.
Run your numbers and see the gap. While I dont have the numbers, I dont think it can work.
Anyways, good luck in your endeavors. Maybe I am wrong.
Just my 2 cents
While your point of view is true in general, there are a couple of exceptions: mainly space and power constraints. Let’s be very specific in our comparison to avoid meaningless comparisons, XCVU9P ($4,000 FPGA) vs GTX 1080 Ti ($800 GPU).
Now let’s take the Phi1612 algorithm and give the FPGA a more realistic 5x performance advantage against the GPU at a power consumption rate of 150 watts (0.150 kWh). I would rather run a 100 FPGA farm (17.5 kWh) vs a 500 GPU farm (87.5 kWh) any day and here’s why. Lower overall costs if you believe the 5x advantage ascribed to the FPGA.
4x FPGAs and components approx. ($16K + $1.2K) in server chassis and add 100 watts for overhead.
4x GPUs and components approx. ($3.2K + $0.65K) in frame and add 100 watts for overhead.
$430K 100 FPGA farm vs $481.25K 500 GPU farm. Can you at least agree that this makes financial sense?
Please don't take the power usage here, let's consider only the hash power and the cost between XCVU9P ($4,000 FPGA) vs GTX 1080 Ti ($800 GPU)
The difference in cost is 5x, does it apply to the hashpower or not?
How quickly does FPGA programming adapt to new algos or the change of old algos?
I am saying it makes no financial sense. The major GPU coins already have asics, with ETH being the biggest. And those coins that have no asic on it will have asics on it before the FPGA will come close to ever breaking even.
And everytime an asic takes over an alternate coin, all the GPUs and FPGAs that were mining that coin will go to a non-asic coin. That means the remaining non-asic coins will have difficulty skyrocket and guess wat, that means your GPUs and FPGAs mining make less money. Everytime a new asic comes out, the FPGAs make less money. So, there is no niche. There is a limited amount to be earned from mining u see. If the asics keep appearing to take some of it, the switchers (FPGAs/GPUs) will have less and less.
Why not u calculate how much it would cost to make an FPGA that can do the equivalent of mining 504mh like the L3+ or 14TH like the S9.
Run your numbers and see the gap. While I dont have the numbers, I dont think it can work.
Anyways, good luck in your endeavors. Maybe I am wrong.
Just my 2 cents
And everytime an asic takes over an alternate coin, all the GPUs and FPGAs that were mining that coin will go to a non-asic coin. That means the remaining non-asic coins will have difficulty skyrocket and guess wat, that means your GPUs and FPGAs mining make less money. Everytime a new asic comes out, the FPGAs make less money. So, there is no niche. There is a limited amount to be earned from mining u see. If the asics keep appearing to take some of it, the switchers (FPGAs/GPUs) will have less and less.
Why not u calculate how much it would cost to make an FPGA that can do the equivalent of mining 504mh like the L3+ or 14TH like the S9.
Run your numbers and see the gap. While I dont have the numbers, I dont think it can work.
Anyways, good luck in your endeavors. Maybe I am wrong.
Just my 2 cents
There is no point in comparing FPGAs to ASICs using Scrypt and SHA256 as examples. That train has sailed... but until ASICs are made for the more obscure algorithms - FPGA can rule there. It takes time and millions of dollars to develop ASICs.
Following your logic, GPU mining makes even less sense, because they are "inferior" against both FPGAs and ASICs. Yet it has been quite good for many years.
Then there are various possibilities of forks, specific FPGA-friendly algorithms etc...
I am saying it makes no financial sense. The major GPU coins already have asics, with ETH being the biggest. And those coins that have no asic on it will have asics on it before the FPGA will come close to ever breaking even.
And everytime an asic takes over an alternate coin, all the GPUs and FPGAs that were mining that coin will go to a non-asic coin. That means the remaining non-asic coins will have difficulty skyrocket and guess wat, that means your GPUs and FPGAs mining make less money. Everytime a new asic comes out, the FPGAs make less money. So, there is no niche. There is a limited amount to be earned from mining u see. If the asics keep appearing to take some of it, the switchers (FPGAs/GPUs) will have less and less.
Why not u calculate how much it would cost to make an FPGA that can do the equivalent of mining 504mh like the L3+ or 14TH like the S9.
Run your numbers and see the gap. While I dont have the numbers, I dont think it can work.
Anyways, good luck in your endeavors. Maybe I am wrong.
Just my 2 cents
While your point of view is true in general, there are a couple of exceptions: mainly space and power constraints. Let’s be very specific in our comparison to avoid meaningless comparisons, XCVU9P ($4,000 FPGA) vs GTX 1080 Ti ($800 GPU).
Now let’s take the Phi1612 algorithm and give the FPGA a more realistic 5x performance advantage against the GPU at a power consumption rate of 150 watts (0.150 kWh). I would rather run a 100 FPGA farm (17.5 kWh) vs a 500 GPU farm (87.5 kWh) any day and here’s why. Lower overall costs if you believe the 5x advantage ascribed to the FPGA.
4x FPGAs and components approx. ($16K + $1.2K) in server chassis and add 100 watts for overhead.
4x GPUs and components approx. ($3.2K + $0.65K) in frame and add 100 watts for overhead.
$430K 100 FPGA farm vs $481.25K 500 GPU farm. Can you at least agree that this makes financial sense?
Interesting
I was thinking FPGA will be a good thing to fight Asics, but it`s very very expensive to buy and very difficulty in a lot of countries
I was thinking FPGA will be a good thing to fight Asics, but it`s very very expensive to buy and very difficulty in a lot of countries
Just my 2 cents speculation
the point is not to compete with asics... as we have seen recently any coin with a large marketcap is going to have an asic implemented after some time. the point is that an asic will only ever work for that algorithm. an fpga can do Yes, I get the point.
But if your FPGA cost $2000 USD and can do 500 hashrate while the asic cost 400 USD and can do the same 500 hashrate, u are still screwed.
U will switch to other coins but so will all the FPGAs and GPUs. Those other coins that are not yet touched by asics will then have their mining difficulty skyrocket due to all the FPGAs and GPUs going there.
Basically, u will lose on a cost per MH/TH basis by too much even if u can match the asics efficiency. And looking at the speed in which asics appear on new coins these days, I dont see how that can work. This looks pretty sucidal.
Just my 2 cents speculation
you cant compare them. the asic will do what an asic does, so you compete on every coin without an asic and you compete vs gpus and cpus.
*edit to add, if you gain yourself a nice niche and an asic does happen to come out, you just go be the king of some other anthill of an algo that isnt asic implemented yet. theres no point trying to make your fpga be as fast or faster than an asic.
I am saying it makes no financial sense. The major GPU coins already have asics, with ETH being the biggest. And those coins that have no asic on it will have asics on it before the FPGA will come close to ever breaking even.
And everytime an asic takes over an alternate coin, all the GPUs and FPGAs that were mining that coin will go to a non-asic coin. That means the remaining non-asic coins will have difficulty skyrocket and guess wat, that means your GPUs and FPGAs mining make less money. Everytime a new asic comes out, the FPGAs make less money. So, there is no niche. There is a limited amount to be earned from mining u see. If the asics keep appearing to take some of it, the switchers (FPGAs/GPUs) will have less and less.
Why not u calculate how much it would cost to make an FPGA that can do the equivalent of mining 504mh like the L3+ or 14TH like the S9.
Run your numbers and see the gap. While I dont have the numbers, I dont think it can work.
Anyways, good luck in your endeavors. Maybe I am wrong.
Just my 2 cents
Sounds like you should probably just stick to GPUs and ASICs, leave the real profits to the people who understand this stuff.
ill ask again because people seem to not have understood me.
where
is
claymore-fpga
miner/ ?
where
is
claymore-fpga
miner/ ?
Apparently you don't understand mining or FPGA.
Claymore has nothing to do with FPGA. No such thing exists.
The FPGA will have it's own mining software/logic. It has nothing to do with claymore.
Just my 2 cents speculation
the point is not to compete with asics... as we have seen recently any coin with a large marketcap is going to have an asic implemented after some time. the point is that an asic will only ever work for that algorithm. an fpga can do Yes, I get the point.
But if your FPGA cost $2000 USD and can do 500 hashrate while the asic cost 400 USD and can do the same 500 hashrate, u are still screwed.
U will switch to other coins but so will all the FPGAs and GPUs. Those other coins that are not yet touched by asics will then have their mining difficulty skyrocket due to all the FPGAs and GPUs going there.
Basically, u will lose on a cost per MH/TH basis by too much even if u can match the asics efficiency. And looking at the speed in which asics appear on new coins these days, I dont see how that can work. This looks pretty sucidal.
Just my 2 cents speculation
you cant compare them. the asic will do what an asic does, so you compete on every coin without an asic and you compete vs gpus and cpus.
*edit to add, if you gain yourself a nice niche and an asic does happen to come out, you just go be the king of some other anthill of an algo that isnt asic implemented yet. theres no point trying to make your fpga be as fast or faster than an asic.
I am saying it makes no financial sense. The major GPU coins already have asics, with ETH being the biggest. And those coins that have no asic on it will have asics on it before the FPGA will come close to ever breaking even.
And everytime an asic takes over an alternate coin, all the GPUs and FPGAs that were mining that coin will go to a non-asic coin. That means the remaining non-asic coins will have difficulty skyrocket and guess wat, that means your GPUs and FPGAs mining make less money. Everytime a new asic comes out, the FPGAs make less money. So, there is no niche. There is a limited amount to be earned from mining u see. If the asics keep appearing to take some of it, the switchers (FPGAs/GPUs) will have less and less.
Why not u calculate how much it would cost to make an FPGA that can do the equivalent of mining 504mh like the L3+ or 14TH like the S9.
Run your numbers and see the gap. While I dont have the numbers, I dont think it can work.
Anyways, good luck in your endeavors. Maybe I am wrong.
Just my 2 cents
Just my 2 cents speculation
the point is not to compete with asics... as we have seen recently any coin with a large marketcap is going to have an asic implemented after some time. the point is that an asic will only ever work for that algorithm. an fpga can do Yes, I get the point.
But if your FPGA cost $2000 USD and can do 500 hashrate while the asic cost 400 USD and can do the same 500 hashrate, u are still screwed.
U will switch to other coins but so will all the FPGAs and GPUs. Those other coins that are not yet touched by asics will then have their mining difficulty skyrocket due to all the FPGAs and GPUs going there.
Basically, u will lose on a cost per MH/TH basis by too much even if u can match the asics efficiency. And looking at the speed in which asics appear on new coins these days, I dont see how that can work. This looks pretty sucidal.
Just my 2 cents speculation
you cant compare them. the asic will do what an asic does, so you compete on every coin without an asic and you compete vs gpus and cpus.
*edit to add, if you gain yourself a nice niche and an asic does happen to come out, you just go be the king of some other anthill of an algo that isnt asic implemented yet. theres no point trying to make your fpga be as fast or faster than an asic.
ill ask again because people seem to not have understood me.
where
is
claymore-fpga
miner/ ?
where
is
claymore-fpga
miner/ ?
Just my 2 cents speculation
the point is not to compete with asics... as we have seen recently any coin with a large marketcap is going to have an asic implemented after some time. the point is that an asic will only ever work for that algorithm. an fpga can do Yes, I get the point.
But if your FPGA cost $2000 USD and can do 500 hashrate while the asic cost 400 USD and can do the same 500 hashrate, u are still screwed.
U will switch to other coins but so will all the FPGAs and GPUs. Those other coins that are not yet touched by asics will then have their mining difficulty skyrocket due to all the FPGAs and GPUs going there.
Basically, u will lose on a cost per MH/TH basis by too much even if u can match the asics efficiency. And looking at the speed in which asics appear on new coins these days, I dont see how that can work. This looks pretty sucidal.
Just my 2 cents speculation
Just my 2 cents speculation
the point is not to compete with asics... as we have seen recently any coin with a large marketcap is going to have an asic implemented after some time. the point is that an asic will only ever work for that algorithm. an fpga can do got a quote from bittmex about the xupp3r. about $5900USD with its development kit but with no memory.
and also a price a for xupp3r with development kit and two of the qdrii+ 288 mb modules. about $12500 USD. ouch. the qdrii+ memory may not even help much with algos that are mostly compute heavy, so im wondering if its worth it. have to try and get some numbers for eth.
and also a price a for xupp3r with development kit and two of the qdrii+ 288 mb modules. about $12500 USD. ouch. the qdrii+ memory may not even help much with algos that are mostly compute heavy, so im wondering if its worth it. have to try and get some numbers for eth.
Ok now pls roughly gauge how much MH or TH u can get if u buy and fix one of those against any of the current ASIC algo.
Because lets say if u gotta spend say 2000 USD to get your FPGA able to do 504mh on scrypt, which makes it equivalent to an L3+ and your FPGA can mine say 3 other coins that are not asic linked yet, I hope u realize that the L3+ is alot cheaper at 725 USD now and it was once 450 USD.
I dont see how this can work because FPGA is more costly than asic, yes? This reminds me of how baikal miners are always more expensive than bitmains.
Compare 50 different asics that mine 5 different coins
With
50 FPGAs that can mine the 5 coins and switch to each coin, I believe the FPGAs are clear cut more expensive yes?
For the 50 FPGAs, u can maybe buy maybe 60-150 asics? And it wont matter if u can switch because when the asics mine 3 of those 5 coins, all FPGAs/GPUs will go to the last 2 coins and try to mine them. Problem is that will increase difficulty of those 2 coins so much that u will earn little anyways.
Furthermore, new asics will be designed economically and manufactured via economies of scale cheaply. I dont see how that can be more economical against a retail user buying an FPGA card from a manufacturer peacemeal. U are paying higher for the switching capability and lack of scale.
Perhaps calculate how much an FPGA that mine an existing asic algo would cost on a per MH/TH basis against one of the asics already out there to get an idea of the gap?
Just my 2 cents speculation
Compare 50 different asics that mine 5 different coins
With
50 FPGAs that can mine the 5 coins and switch to each coin, I believe the FPGAs are clear cut more expensive yes?
For the 50 FPGAs, u can maybe buy maybe 60-150 asics? And it wont matter if u can switch because when the asics mine 3 of those 5 coins, all FPGAs/GPUs will go to the last 2 coins and try to mine them. Problem is that will increase difficulty of those 2 coins so much that u will earn little anyways.
Furthermore, new asics will be designed economically and manufactured via economies of scale cheaply. I dont see how that can be more economical against a retail user buying an FPGA card from a manufacturer peacemeal. U are paying higher for the switching capability and lack of scale.
Perhaps calculate how much an FPGA that mine an existing asic algo would cost on a per MH/TH basis against one of the asics already out there to get an idea of the gap?
Just my 2 cents speculation
About X16R. Consider using an array of folded cores. All you need is 16 arrays of cores of each algorithm. Arrays should be balanced to have the same value of hashrate. The firmware is one.
https://cryptography.gmu.edu/athena/index.php?id=source_codes
https://cryptography.gmu.edu/athena/index.php?id=source_codes
the difficulty is just to have it optimized for some many algos... anyway at the end it will resort as always at proof of trust...
I don't get it. Can you explain in other words?
Those algorithms are already implemented and seems to be working. Even unoptimized firmware in proper moment of time can give much profit.
About X16R. Consider using an array of folded cores. All you need is 16 arrays of cores of each algorithm. Arrays should be balanced to have the same value of hashrate. The firmware is one.
https://cryptography.gmu.edu/athena/index.php?id=source_codes
https://cryptography.gmu.edu/athena/index.php?id=source_codes
the difficulty is just to have it optimized for some many algos... anyway at the end it will resort as always at proof of trust...
About X16R. Consider using an array of folded cores. All you need is 16 arrays of cores of each algorithm. Arrays should be balanced to have the same value of hashrate. The firmware is one.
https://cryptography.gmu.edu/athena/index.php?id=source_codes
https://cryptography.gmu.edu/athena/index.php?id=source_codes
got a quote from bittmex about the xupp3r. about $5900USD with its development kit but with no memory.
and also a price a for xupp3r with development kit and two of the qdrii+ 288 mb modules. about $12500 USD. ouch. the qdrii+ memory may not even help much with algos that are mostly compute heavy, so im wondering if its worth it. have to try and get some numbers for eth.
and also a price a for xupp3r with development kit and two of the qdrii+ 288 mb modules. about $12500 USD. ouch. the qdrii+ memory may not even help much with algos that are mostly compute heavy, so im wondering if its worth it. have to try and get some numbers for eth.
then eth fork and you are naked like a worm
or who knows how reliable this things are... I have no idea. 
got a quote from bittmex about the xupp3r. about $5900USD with its development kit but with no memory.
and also a price a for xupp3r with development kit and two of the qdrii+ 288 mb modules. about $12500 USD. ouch. the qdrii+ memory may not even help much with algos that are mostly compute heavy, so im wondering if its worth it. have to try and get some numbers for eth.
and also a price a for xupp3r with development kit and two of the qdrii+ 288 mb modules. about $12500 USD. ouch. the qdrii+ memory may not even help much with algos that are mostly compute heavy, so im wondering if its worth it. have to try and get some numbers for eth.
For Ravencoin and Bitcore, the DDR4 is not used for hashing at all, but rather to store hundreds of different FPGA configuration bitstreams, which allows the entire FPGA to rapidly reprogram itself on every block based on the algorithm sequence for that block. Cryptonight7 and Equihash, on the other hand, would use the external memory for actual hashing. Whether CN7/Equihash gain from additional RAM depends on how those algorithms are ultimately implemented.
bumped. if anyone has an idea
if you look at the timing of ddr4 avaible you will see that they have been getting slower (ie bigger from CL12-13 to 15-16) for a while... ie as soon as fpgs mfg adapted their design to ddr4....
Jump to: