DIY FPGA Mining rig for any algorithm with fast ROI

[yrk1957]

Well, not so bright expectations, I assume...

Realistically, Monero/CNV7 is the only hope for these cards. And that too only if Monero does not fork again and they are able to do 40+KH/s at the minimum. Otherwise, the batch of 5000 FPGAs, plus other group buys, will just kill each other on other algorithms.

Once we finish work on our shell requirements for shipping (Expected to be finished by monday -- we are in testing now and everything looks good), we can continue working on bitstreams. We should be able to get lyra2z to 70-80Mh/s pretty quickly, Then we will focus on either neoscrypt (expecting 30mh/s +/-) or phi2 (pretty wide range here, around 100Mh/s starting is a good bet).

This is the first time I am hearing Neoscrypt numbers. 30MH/s on Neo seems pretty good, about 15-20 1080Tis, with much lower power. But, the Neo total network is only 20,000 MH/s, which is equivalent to 700 odds FPGAs, so pretty small target.

[yrk1957]

Well, not so bright expectations, I assume...

Realistically, Monero/CNV7 is the only hope for these cards. And that too only if Monero does not fork again and they are able to do 40+KH/s at the minimum. Otherwise, the batch of 5000 FPGAs, plus other group buys, will just kill each other on other algorithms.

Monero will fork in october and the algo has been specifically changed to lower ROI on FPGA's. I am not sure what percentage it will drop though, I think it was 20% but don't quote me on that.

Is the new algorithm fixed? It will be interesting to see the hit on FPGAs.

[Hueristic]

Well, not so bright expectations, I assume...

Realistically, Monero/CNV7 is the only hope for these cards. And that too only if Monero does not fork again and they are able to do 40+KH/s at the minimum. Otherwise, the batch of 5000 FPGAs, plus other group buys, will just kill each other on other algorithms.

Monero will fork in october and the algo has been specifically changed to lower ROI on FPGA's. I am not sure what percentage it will drop though, I think it was 20% but don't quote me on that.

Is the new algorithm fixed? It will be interesting to see the hit on FPGAs.

There is a discussion on it on git I think you'll have to google it.

[COINKING7]

Bitmain is going public and will control every Algo soon. Most cryptos are deciding that forking isn't worth it, [ETH, ETC, SIA, DECRED, ZEN, etc.] Monero seems to be one of the only major projects dodging ASICs. So does 6k justify the ability to mine a fork? I'm not a huge Bitmain fan, but let's face it, they have the market cornered and have the ability to undercut competition and install the latest chips. What happens when the VU13P chip becomes perhaps overtaken by the Stratix 10 next year? Will ONE 6k FPGA ever give customers their ROI? You can get a Z9 mini on Ebay for about $700 with PSU [1 sold at that price 4 days ago], and net about $5-7/day overclocked. In addition, ASICs are plug-and-play, has anyone read through the PDF on the downloads page to get these FPGA's up and running???

[yrk1957]

Bitmain is going public and will control every Algo soon. Most cryptos are deciding that forking isn't worth it, [ETH, ZEN, etc.] Monero seems to be the only major project dodging ASICs. So does 6k justify the ability to mine a fork? I'm not a huge Bitmain fan, but let's face it, they have the market cornered and have the ability to undercut competition and install the latest chips. What happens when the VU13P chip becomes perhaps overtaken by the Stratix 10 next year? Will ONE 6k FPGA ever give customers their ROI? In addition, ASICs are plug-and-play, has anyone read through the PDF on the downloads page to get these FPGA's up and running???

I doubt Bitmain will go for small market-cap algorithms (coins) like on Neoscrypt or Lyra2z. Only when they become big like Decred.

But that’s the catch also, FPGA’s will be fine until the coins are small, so low returns overall, as they become bigger, they will attract ASICs.

[yrk1957]

Is the new algorithm fixed? It will be interesting to see the hit on FPGAs.

There is a discussion on it on git I think you'll have to google it.

I see it’s ready.

 I vaguely remember a discussion on discord where the FPGA Devs were laughing at the change of introducing a square root to slow down FPGAs. I suppose Monero Devs will get smarter in the fork after October fork.

[Hueristic]

Is the new algorithm fixed? It will be interesting to see the hit on FPGAs.

There is a discussion on it on git I think you'll have to google it.

I see it’s ready.

 I vaguely remember a discussion on discord where the FPGA Devs were laughing at the change of introducing a square root to slow down FPGAs. I suppose Monero Devs will get smarter in the fork after October fork.

You guys get to laugh all the way to the bank after you ROI.

I considered buying one just for shits and grins but I don't have a rig setup anymore and didn't want to dust it off.

[Iamtutut]

Well, not so bright expectations, I assume...

Realistically, Monero/CNV7 is the only hope for these cards. And that too only if Monero does not fork again and they are able to do 40+KH/s at the minimum. Otherwise, the batch of 5000 FPGAs, plus other group buys, will just kill each other on other algorithms.

Monero will fork in october and the algo has been specifically changed to lower ROI on FPGA's. I am not sure what percentage it will drop though, I think it was 20% but don't quote me on that.

Compared to now
ASIC hashrate: divide by 16
FPGA hashrate divide by 4
AMD GPUs (Polaris and Vega) will lose a little more hashrate compared to NVIDIA.

[Hueristic]

Well, not so bright expectations, I assume...

Realistically, Monero/CNV7 is the only hope for these cards. And that too only if Monero does not fork again and they are able to do 40+KH/s at the minimum. Otherwise, the batch of 5000 FPGAs, plus other group buys, will just kill each other on other algorithms.

Monero will fork in october and the algo has been specifically changed to lower ROI on FPGA's. I am not sure what percentage it will drop though, I think it was 20% but don't quote me on that.

Compared to now
ASIC hashrate: divide by 16
FPGA hashrate divide by 4
AMD GPUs (Polaris and Vega) will lose a little more hashrate compared to NVIDIA.

Do you have a link? IIRC Intel CPUs will actually increase their rate.

[whitefire990]

Monero is irrelevant to FPGA's.  I have never considered it a viable option and I am not doing any development on it or any of its variants.  Notice that I didn't even list it in my original April 30 post.  There are enough algorithms (without Monero or any cryptonight variant) to support 25000 FPGA's now, and at least one top-40 coin can make >$30/day with a VCU1525/BCU1525.  However, unlike April 30th, I will no longer post information about any 'good' coins until the algorithms are actually ready for download.  I don't expect this type of information to cause anyone to buy an FPGA, but I will say those people who took the risk will be VERY HAPPY THEY DID.  By the time the super profitable algorithms are released, it will be too late for the naysayers to acquire hardware, as the hardware will all be owned by the early adopters.

The same goes for people who bought Bittware CVP-13's in the group order.  A risky choice, but they will be very happy they took the risk.

If you consider that a BCU1525 can hash certain ASIC algorithms like Skein at almost 3 times the rate of a Baikal X/X10, and also do very well on things like Bytom and the equihash variants, there are tons of future possibilities.

[cryptolumos]

Monero is irrelevant to FPGA's.  I have never considered it a viable option and I am not doing any development on it or any of its variants.  Notice that I didn't even list it in my original April 30 post.  There are enough algorithms (without Monero or any cryptonight variant) to support 25000 FPGA's now, and at least one top-40 coin can make >$30/day with a VCU1525/BCU1525.  However, unlike April 30th, I will no longer post information about any 'good' coins until the algorithms are actually ready for download.  I don't expect this type of information to cause anyone to buy an FPGA, but I will say those people who took the risk will be VERY HAPPY THEY DID.  By the time the super profitable algorithms are released, it will be too late for the naysayers to acquire hardware, as the hardware will all be owned by the early adopters.

The same goes for people who bought Bittware CVP-13's in the group order.  A risky choice, but they will be very happy they took the risk.

If you consider that a BCU1525 can hash certain ASIC algorithms like Skein at almost 3 times the rate of a Baikal X/X10, and also do very well on things like Bytom and the equihash variants, there are tons of future possibilities.

shhhh people will catch on, im still saving for another cvp-13 <3 keep up the awesome work Zeth!

[COINKING7]

Monero is irrelevant to FPGA's.  I have never considered it a viable option and I am not doing any development on it or any of its variants.  Notice that I didn't even list it in my original April 30 post.  There are enough algorithms (without Monero or any cryptonight variant) to support 25000 FPGA's now, and at least one top-40 coin can make >$30/day with a VCU1525/BCU1525.  However, unlike April 30th, I will no longer post information about any 'good' coins until the algorithms are actually ready for download.  I don't expect this type of information to cause anyone to buy an FPGA, but I will say those people who took the risk will be VERY HAPPY THEY DID.  By the time the super profitable algorithms are released, it will be too late for the naysayers to acquire hardware, as the hardware will all be owned by the early adopters.

The same goes for people who bought Bittware CVP-13's in the group order.  A risky choice, but they will be very happy they took the risk.

If you consider that a BCU1525 can hash certain ASIC algorithms like Skein at almost 3 times the rate of a Baikal X/X10, and also do very well on things like Bytom and the equihash variants, there are tons of future possibilities.

Just to be realistic, I am sure you are brilliant but Bitmain can crush anything you are doing. Especially with with continued delays. They have billions of funds and 100's of talented coders/programmers.

In addition, you are saying no hardware will be available after these batches? That just gives Bitmain another chance to leapfrog you.

Each delay gives Bitmain a chance to create their own plug and play hardware, and on each delay, mining difficulty will rise and profits will fall. So any numbers being put out know are irrelevant, especially when 5,000 units hit the market.

[Hueristic]

Monero is irrelevant to FPGA's.  I have never considered it a viable option and I am not doing any development on it or any of its variants.  Notice that I didn't even list it in my original April 30 post.  There are enough algorithms (without Monero or any cryptonight variant) to support 25000 FPGA's now, and at least one top-40 coin can make >$30/day with a VCU1525/BCU1525.  However, unlike April 30th, I will no longer post information about any 'good' coins until the algorithms are actually ready for download.  I don't expect this type of information to cause anyone to buy an FPGA, but I will say those people who took the risk will be VERY HAPPY THEY DID.  By the time the super profitable algorithms are released, it will be too late for the naysayers to acquire hardware, as the hardware will all be owned by the early adopters.

The same goes for people who bought Bittware CVP-13's in the group order.  A risky choice, but they will be very happy they took the risk.

If you consider that a BCU1525 can hash certain ASIC algorithms like Skein at almost 3 times the rate of a Baikal X/X10, and also do very well on things like Bytom and the equihash variants, there are tons of future possibilities.

Just to be realistic, I am sure you are brilliant but Bitmain can crush anything you are doing. Especially with with continued delays. They have billions of funds and 100's of talented coders/programmers.

In addition, you are saying no hardware will be available after these batches? That just gives Bitmain another chance to leapfrog you.

Each delay gives Bitmain a chance to create their own plug and play hardware, and on each delay, mining difficulty will rise and profits will fall. So any numbers being put out know are irrelevant, especially when 5,000 units hit the market.

You are ignorant of semiconductor construction processes and The industries ability to supply. Bitmain relies on Fabs like TSMC and they supply the largest players in the world, there is a waiting list. You don't run down to walmart and say dude give me some 7nm (not the intel 10nm) wafers from this tape out I did last night. KK be back at lunch and make sure I get 90% + yields cause like particles are a real bitch ya know.

[whitefire990]

Monero is irrelevant to FPGA's.  I have never considered it a viable option and I am not doing any development on it or any of its variants.  Notice that I didn't even list it in my original April 30 post.  There are enough algorithms (without Monero or any cryptonight variant) to support 25000 FPGA's now, and at least one top-40 coin can make >$30/day with a VCU1525/BCU1525.  However, unlike April 30th, I will no longer post information about any 'good' coins until the algorithms are actually ready for download.  I don't expect this type of information to cause anyone to buy an FPGA, but I will say those people who took the risk will be VERY HAPPY THEY DID.  By the time the super profitable algorithms are released, it will be too late for the naysayers to acquire hardware, as the hardware will all be owned by the early adopters.

The same goes for people who bought Bittware CVP-13's in the group order.  A risky choice, but they will be very happy they took the risk.

If you consider that a BCU1525 can hash certain ASIC algorithms like Skein at almost 3 times the rate of a Baikal X/X10, and also do very well on things like Bytom and the equihash variants, there are tons of future possibilities.

Just to be realistic, I am sure you are brilliant but Bitmain can crush anything you are doing. Especially with with continued delays. They have billions of funds and 100's of talented coders/programmers.

In addition, you are saying no hardware will be available after these batches? That just gives Bitmain another chance to leapfrog you.

Each delay gives Bitmain a chance to create their own plug and play hardware, and on each delay, mining difficulty will rise and profits will fall. So any numbers being put out know are irrelevant, especially when 5,000 units hit the market.

You obviously haven't been reading financial news.  Bitmain is almost bankrupt.  Their supposed IPO is now considered an exit scam.  Their mining machines sell for basically nothing, except for the Z9 series.  They are bleeding cash and lost billions by converting their money into Bitcoin Cash which then crashed.  Innosilicon has a better chance to survive.

[Iamtutut]

Do you have a link? IIRC Intel CPUs will actually increase their rate.

Found it ! Didn't read correctly the first time though.

Source: https://github.com/SChernykh/xmr-stak-cpu/blob/master/README.md

"On the other side, ASIC/FPGA which use external memory for scratchpad will get 4 times slower due to increased bandwidth usage. ASIC/FPGA which use on-chip memory for scratchpad will get ~15 times slower because of high latencies introduced with division and square root calculations: they just don't have enough on-chip memory to hide these latencies with many parallel Cryptonight calculations."

[huntingthesnark]

Monero is irrelevant to FPGA's.  I have never considered it a viable option and I am not doing any development on it or any of its variants.  Notice that I didn't even list it in my original April 30 post.  There are enough algorithms (without Monero or any cryptonight variant) to support 25000 FPGA's now, and at least one top-40 coin can make >$30/day with a VCU1525/BCU1525.  However, unlike April 30th, I will no longer post information about any 'good' coins until the algorithms are actually ready for download.  I don't expect this type of information to cause anyone to buy an FPGA, but I will say those people who took the risk will be VERY HAPPY THEY DID.  By the time the super profitable algorithms are released, it will be too late for the naysayers to acquire hardware, as the hardware will all be owned by the early adopters.

The same goes for people who bought Bittware CVP-13's in the group order.  A risky choice, but they will be very happy they took the risk.

If you consider that a BCU1525 can hash certain ASIC algorithms like Skein at almost 3 times the rate of a Baikal X/X10, and also do very well on things like Bytom and the equihash variants, there are tons of future possibilities.

Just to be realistic, I am sure you are brilliant but Bitmain can crush anything you are doing. Especially with with continued delays. They have billions of funds and 100's of talented coders/programmers.

In addition, you are saying no hardware will be available after these batches? That just gives Bitmain another chance to leapfrog you.

Each delay gives Bitmain a chance to create their own plug and play hardware, and on each delay, mining difficulty will rise and profits will fall. So any numbers being put out know are irrelevant, especially when 5,000 units hit the market.

You obviously haven't been reading financial news.  Bitmain is almost bankrupt.  Their supposed IPO is now considered an exit scam.  Their mining machines sell for basically nothing, except for the Z9 series.  They are bleeding cash and lost billions by converting their money into Bitcoin Cash which then crashed.  Innosilicon has a better chance to survive.

Amen to that. Also BM is in a dangerous position because it is the crypto market leader - going after small cap coins with asics just isn't going to pay the increasingly hefty bills. Meanwhile the really big guys (Samsung for example) could annihilate BM from a technical and purchasing firepower point of view. Interesting times.

[yrk1957]

Monero is irrelevant to FPGA's.  I have never considered it a viable option and I am not doing any development on it or any of its variants.  Notice that I didn't even list it in my original April 30 post.  There are enough algorithms (without Monero or any cryptonight variant) to support 25000 FPGA's now, and at least one top-40 coin can make >$30/day with a VCU1525/BCU1525.  However, unlike April 30th, I will no longer post information about any 'good' coins until the algorithms are actually ready for download.  I don't expect this type of information to cause anyone to buy an FPGA, but I will say those people who took the risk will be VERY HAPPY THEY DID.  By the time the super profitable algorithms are released, it will be too late for the naysayers to acquire hardware, as the hardware will all be owned by the early adopters.

The same goes for people who bought Bittware CVP-13's in the group order.  A risky choice, but they will be very happy they took the risk.

If you consider that a BCU1525 can hash certain ASIC algorithms like Skein at almost 3 times the rate of a Baikal X/X10, and also do very well on things like Bytom and the equihash variants, there are tons of future possibilities.

Well the April assessment was for 25000 FPGAs with 12000 odd on Monero. Since then coins are down multiple times, and if you take Monero also out, the figure if 25000 still being supported is hard to believe. Though if you are adding Equihash, or BTG with its Equihash variant...

So maybe 50+ KH/s on Equihash/variant?

[yrk1957]

Monero is irrelevant to FPGA's.  I have never considered it a viable option and I am not doing any development on it or any of its variants.  Notice that I didn't even list it in my original April 30 post.  There are enough algorithms (without Monero or any cryptonight variant) to support 25000 FPGA's now, and at least one top-40 coin can make >$30/day with a VCU1525/BCU1525.  However, unlike April 30th, I will no longer post information about any 'good' coins until the algorithms are actually ready for download.  I don't expect this type of information to cause anyone to buy an FPGA, but I will say those people who took the risk will be VERY HAPPY THEY DID.  By the time the super profitable algorithms are released, it will be too late for the naysayers to acquire hardware, as the hardware will all be owned by the early adopters.

The same goes for people who bought Bittware CVP-13's in the group order.  A risky choice, but they will be very happy they took the risk.

If you consider that a BCU1525 can hash certain ASIC algorithms like Skein at almost 3 times the rate of a Baikal X/X10, and also do very well on things like Bytom and the equihash variants, there are tons of future possibilities.

Just to be realistic, I am sure you are brilliant but Bitmain can crush anything you are doing. Especially with with continued delays. They have billions of funds and 100's of talented coders/programmers.

In addition, you are saying no hardware will be available after these batches? That just gives Bitmain another chance to leapfrog you.

Each delay gives Bitmain a chance to create their own plug and play hardware, and on each delay, mining difficulty will rise and profits will fall. So any numbers being put out know are irrelevant, especially when 5,000 units hit the market.

You obviously haven't been reading financial news.  Bitmain is almost bankrupt.  Their supposed IPO is now considered an exit scam.  Their mining machines sell for basically nothing, except for the Z9 series.  They are bleeding cash and lost billions by converting their money into Bitcoin Cash which then crashed.  Innosilicon has a better chance to survive.

Amen to that. Also BM is in a dangerous position because it is the crypto market leader - going after small cap coins with asics just isn't going to pay the increasingly hefty bills. Meanwhile the really big guys (Samsung for example) could annihilate BM from a technical and purchasing firepower point of view. Interesting times.

They already lost $500 million on chip design. Which apparently is part and parcel for chip design process, but still. They won't be going after any small coin/algo.

[roc-noc]

For the BCU-1525 it appears that there are 4 heatsink options and I have been asked to complete a survey to chose my cooling options.

As a group buy purchaser back in June, I m really confused on the options and what performance benefits we will see.

I ordered the $3600 USD models with RAM and "mods and alterations".

The 4 options are:

1) stock factory BCU Xilinx heatsink
2) SQRL upgraded passive heatsink
3) SQRL upgraded active heatsink
4) SQRL water block

Does my $3600 model include the #2 heatsink?

What will the mining performance variance be for the 4 different options?

Can the active air cooled units mine as quickly as the water cooled units assuming 4 or 8 cards in a standard GPU mining rack with external fan blowing?

Thnx!

[m5]

For the BCU-1525 it appears that there are 4 heatsink options and I have been asked to complete a survey to chose my cooling options.

As a group buy purchaser back in June, I m really confused on the options and what performance benefits we will see.

I ordered the $3600 USD models with RAM and "mods and alterations".

The 4 options are:

1) stock factory BCU Xilinx heatsink
2) SQRL upgraded passive heatsink
3) SQRL upgraded active heatsink
4) SQRL water block

Does my $3600 model include the #2 heatsink?

What will the mining performance variance be for the 4 different options?

Can the active air cooled units mine as quickly as the water cooled units assuming 4 or 8 cards in a standard GPU mining rack with external fan blowing?

Thnx!

You can get "2)" for free, but there is some delay. So you either get "1)" now or wait 1-2 weeks for "2)".
"3)" is "2)" with fitter powerful fan.
"4)" is a water block. You can either have "1)" shipped now and water block later, which you will install yourself and void warranty, or wait until end of Oct and have water block installed at factory and shipped with card.

"2)" can run 350-400 W with good airflow. I don't think you can cool with air the same as with water, however the difference in water is not so huge as it is for VCU1525.

The whole survey is because they got the cards before the heatsinks, so you have option to wait or not to wait for them...
All this info is from the Discord.