http://serason.com/BFL1.jpg
http://serason.com/BFL2.jpg
Topic: FPGA - profitability vs effeciency? (Read 1963 times)
The BFL FPGAs in this shot are doing between 830-877Mh/s that's pretty good, better than any of my GPUs
http://serason.com/BFL1.jpg
http://serason.com/BFL2.jpg
http://serason.com/BFL1.jpg
http://serason.com/BFL2.jpg
The Blackarrow version has already the "overvolt" modify, change the upper fan with a bigger one (92mm) add a lower fan for the backside. And use a good PSU. With the V42 you shouldn't get anything less than 460-480.
Unfortunately Cgminer can't measure the speed correctly, so you have to count the accepted units (actually more than 3.6 at diff 2).
Keeping stock cooling is worth only an increase from 2.8 to 3.2 (still good) even with a 10% of errors
Unfortunately Cgminer can't measure the speed correctly, so you have to count the accepted units (actually more than 3.6 at diff 2).
Keeping stock cooling is worth only an increase from 2.8 to 3.2 (still good) even with a 10% of errors
The 5 columns on the far right represent average difficulty through the year from 8%-28%
I have all this laid out if anyone hasn't seen: https://bitcointalksearch.org/topic/mining-equipment-manufacturers-209750
Now of course this changes if you are preordered already, or the times change again... There isn't really anything you can buy at the moment (until the avalon chip order comes in) that will be a better investment than just buying btc.
Now of course this changes if you are preordered already, or the times change again... There isn't really anything you can buy at the moment (until the avalon chip order comes in) that will be a better investment than just buying btc.
Where does this chart take into account the different possible(or any real possible) network hash rate increase. This is the biggest factor people need to see.
Would still rather at the least a free graphics card come the transition...
The comparison to one year ago is interesting, meanwhile the Btc today is exchanged at more than $100 so the break even is the same, with just less coins.
It's obvious that on paper an Fpga cannot compete with ASIC technology, the problem is that the vast majority of us got only preorders
And they don't mine anything
In my opinion Fpga mining is still profitable if you can setup everything -today- and squeeze the hardware until the ASIC "tsunami".
ASIC mining -today- with the first generation devices is expensive (actually more than Fpga, look at asicminer prices) and will suffer the same problem of everyone else with just a bit smaller electricity bill. With the difference that you have to wait in line losing money very other day.
It's obvious that on paper an Fpga cannot compete with ASIC technology, the problem is that the vast majority of us got only preorders
And they don't mine anything
In my opinion Fpga mining is still profitable if you can setup everything -today- and squeeze the hardware until the ASIC "tsunami".
ASIC mining -today- with the first generation devices is expensive (actually more than Fpga, look at asicminer prices) and will suffer the same problem of everyone else with just a bit smaller electricity bill. With the difference that you have to wait in line losing money very other day.
Agreed, people who are buying ASICminer items at current pricing are more foolish than the current FPGAers.
I have all this laid out if anyone hasn't seen: https://bitcointalksearch.org/topic/mining-equipment-manufacturers-209750
Now of course this changes if you are preordered already, or the times change again... There isn't really anything you can buy at the moment (until the avalon chip order comes in) that will be a better investment than just buying btc.
Now of course this changes if you are preordered already, or the times change again... There isn't really anything you can buy at the moment (until the avalon chip order comes in) that will be a better investment than just buying btc.
The comparison to one year ago is interesting, meanwhile the Btc today is exchanged at more than $100 so the break even is the same, with just less coins.
It's obvious that on paper an Fpga cannot compete with ASIC technology, the problem is that the vast majority of us got only preorders
And they don't mine anything
In my opinion Fpga mining is still profitable if you can setup everything -today- and squeeze the hardware until the ASIC "tsunami".
ASIC mining -today- with the first generation devices is expensive (actually more than Fpga, look at asicminer prices) and will suffer the same problem of everyone else with just a bit smaller electricity bill. With the difference that you have to wait in line losing money very other day.
It's obvious that on paper an Fpga cannot compete with ASIC technology, the problem is that the vast majority of us got only preorders
And they don't mine anything
In my opinion Fpga mining is still profitable if you can setup everything -today- and squeeze the hardware until the ASIC "tsunami".
ASIC mining -today- with the first generation devices is expensive (actually more than Fpga, look at asicminer prices) and will suffer the same problem of everyone else with just a bit smaller electricity bill. With the difference that you have to wait in line losing money very other day.
Hopefully not to much hashing power has not been bought. It seems to many people don't understand the large amount of ASICs that are going to come online and skyrocket the hash rate. Back before the block halving when GPU's and FPGA were dominant, it seemed 5-6 months was the average break even time frame. No matter how much you spent, so long as it was on effecient hardware MHs/$. Obviously FPGAs were breaking earlier on the lower end(5 months) due to less power consumption. Once the block reward halved, that initially changed to about 10-11 months, then moved to about 9-10 months as the price got up to $13-14 range. And since then, I haven't watched it as closely but a quick calculation(not consider a ASIC) is about 5 months. But this is quickly climbing even with the current slow introduction of ASICs.
Hopefully at worst it stabilizes in the 9-10 month range again as this seems like that is the limit for what miners are willing to risk. And now with ASICs the investment risk is much higher as these devices are now 'Application Specific' and the price per device is many times higher now. These big investors with these centralized setups can really kill the break even time though. I for one am strapped in and hoping for the best.
Hopefully at worst it stabilizes in the 9-10 month range again as this seems like that is the limit for what miners are willing to risk. And now with ASICs the investment risk is much higher as these devices are now 'Application Specific' and the price per device is many times higher now. These big investors with these centralized setups can really kill the break even time though. I for one am strapped in and hoping for the best.
The math still doesn't work in FPGA's favor at that price. In order for the math to work, FPGAs need to be $200-$300ish for a ~400m/h unit...
I agree with you completely - I don't think a 400 mh unit should cost more than 1 btc at today's diff. Even at that, it still might take a long time to break even. I was trying to say that you can't put a 7970 in a mom & pop desktop with a 300w psu and expect it to work (tho I agree you don't need a quad core CPU and 8 gb of ram either - mom & pops CPU would do just fine). Blockchain.info has a very telling chart that shows the growth in diff - its climbing at a fairly steep rate! Even the mythical Jalipeno might not break even a year from now (tho I certainly hope that's not the case!).
I'm a lucky owner of some Blackarrow Lancelots, with the slight tunes they made to the original design (overvolt basically) you can run the V42 firmware and go over 500MHash/s with nothing more than a good PSU.
Quite better than some currently popular ASIC
Quite better than some currently popular ASIC
How is 500 mh/s better then any asic? Even the USB block eruptors pull around 350 mh/s @ 1.5 W and those are pretty much the most expensive asic's there are.
You hit the nail on the head: efficiency =/= profitability. I can undervolt my GPUs to all hell and get some efficiency numbers that come close to FPGAs, but it's more profitable to overclock (well, it's actually moving towards being more profitable to balance the two, but that's besides the point 
).
I've never understood why you would pay $1,100 for a unit that hashes ~20% faster than a $400 GPU. I don't care how little power it uses, it would take years to compensate for the extra $700 in power savings. Even now, you can buy 400Mh/s for $350, or a 700MH/s GPU for the same price.
The only ones that made sense to me were the BFL Singles, and the Cairnsmore1 units. They were the only ones priced at a point that didn't seem ridiculous. Sure there was a premium over GPUs for the same hashrate, but the simplicity of setting them up and the power savings actually made it worth it.
Now, I wouldn't bother. ASICs are what's next, and both FPGAs and GPUs are soon to be left making pennies per day.
But what do I know? It's 3am and I gotta be at work at 9:00. After that, however, I'm on vacation, and I couldn't give 2 shits if I wanted to.
).I've never understood why you would pay $1,100 for a unit that hashes ~20% faster than a $400 GPU. I don't care how little power it uses, it would take years to compensate for the extra $700 in power savings. Even now, you can buy 400Mh/s for $350, or a 700MH/s GPU for the same price.
The only ones that made sense to me were the BFL Singles, and the Cairnsmore1 units. They were the only ones priced at a point that didn't seem ridiculous. Sure there was a premium over GPUs for the same hashrate, but the simplicity of setting them up and the power savings actually made it worth it.
Now, I wouldn't bother. ASICs are what's next, and both FPGAs and GPUs are soon to be left making pennies per day.
But what do I know? It's 3am and I gotta be at work at 9:00. After that, however, I'm on vacation, and I couldn't give 2 shits if I wanted to.
This depends pretty heavily on other factors though. You have to consider that some people here on the forum live in countries where Power is several times more expensive then for example mainland US where I am ($0.06/kwh). I have heard of people paying as high as $0.35/kwh, at which point running a 200W GPU starts really cutting into profits.
The math still doesn't work in FPGA's favor at that price. In order for the math to work, FPGAs need to be $200-$300ish for a ~400m/h unit...
Which is within the realm of what the Blackarrow Lancelot FPGAs are going for ($269 - $350). I agree with you though, a lot of the FPGAs that are being sold are grossly overpriced - like I can't understand buying the ModMiner for over $1000, when two Lancelots will give the same or better hash rate for less. Heck, you could get four Lancelots (if you buy on the group manufacturing deal) for about the same price as one ModMiner, and you would have 1.6 GH/s.
Once ASICs begin shipping in mass quantities though, there will be no reason to purchase FPGAs any longer. Right now, I think they can still be (mildly) profitable.
FPGAs made a lot more sense when the difficulty was lower and BTC were $6 each. Then it was barely profitable to run GPUs if you lived where power was expensive.
I'm a lucky owner of some Blackarrow Lancelots, with the slight tunes they made to the original design (overvolt basically) you can run the V42 firmware and go over 500MHash/s with nothing more than a good PSU.
Quite better than some currently popular ASIC
Quite better than some currently popular ASIC
The math still doesn't work in FPGA's favor at that price. In order for the math to work, FPGAs need to be $200-$300ish for a ~400m/h unit...
Which is within the realm of what the Blackarrow Lancelot FPGAs are going for ($269 - $350). I agree with you though, a lot of the FPGAs that are being sold are grossly overpriced - like I can't understand buying the ModMiner for over $1000, when two Lancelots will give the same or better hash rate for less. Heck, you could get four Lancelots (if you buy on the group manufacturing deal) for about the same price as one ModMiner, and you would have 1.6 GH/s.
Once ASICs begin shipping in mass quantities though, there will be no reason to purchase FPGAs any longer. Right now, I think they can still be (mildly) profitable.
There is one factor that often gets overlooked. Sure, you can get a GPU that has twice the hash rate for half of the money, but you also have to have a high-end PSU to drive it as well as a motherboard to hook it to. What you need is a decent gaming rig since your average BestBuy desktop doesn't have the specs to support that high-end GPU (much less 2 or more at a time!). That all adds to the cost (and is also why crate rigs are so popular with the hard-core miners).
While hash/watt is a major factor going for FPGA/ASIC, the fact that you can run them (though maybe not directly power them) from any old box you have collecting dust (or even a RasPI if you are adventurous) also makes them attractive. So what we need to really do is look at the whole system cost - not just the hasher, but everything you need to get the hasher to hash. That's what the ROI figures should be based on...
Personally I'd be happy to have a 'mere' 1-2 GH unit (or two) that will pay for itself in a month or two that will then help me get some use of all the spare cycles I have laying around the house
While hash/watt is a major factor going for FPGA/ASIC, the fact that you can run them (though maybe not directly power them) from any old box you have collecting dust (or even a RasPI if you are adventurous) also makes them attractive. So what we need to really do is look at the whole system cost - not just the hasher, but everything you need to get the hasher to hash. That's what the ROI figures should be based on...
Personally I'd be happy to have a 'mere' 1-2 GH unit (or two) that will pay for itself in a month or two that will then help me get some use of all the spare cycles I have laying around the house
High end PSU?
Gaming rig? Any old piece of shit single-core processor machine will do. If you need to run multiple cards, you can upgrade the PSU, but you can easily get a suitable PSU to run multiple cards for $100ish.So the old crappy PC that EVERYBODY still has laying around unused, and maybe $100 to get the PSU up to snuff lets you run $120-$180 cards that hash at the same rate as an $800 FPGA...
The math still doesn't work in FPGA's favor at that price. In order for the math to work, FPGAs need to be $200-$300ish for a ~400m/h unit...
There is one factor that often gets overlooked. Sure, you can get a GPU that has twice the hash rate for half of the money, but you also have to have a high-end PSU to drive it as well as a motherboard to hook it to. What you need is a decent gaming rig since your average BestBuy desktop doesn't have the specs to support that high-end GPU (much less 2 or more at a time!). That all adds to the cost (and is also why crate rigs are so popular with the hard-core miners).
While hash/watt is a major factor going for FPGA/ASIC, the fact that you can run them (though maybe not directly power them) from any old box you have collecting dust (or even a RasPI if you are adventurous) also makes them attractive. So what we need to really do is look at the whole system cost - not just the hasher, but everything you need to get the hasher to hash. That's what the ROI figures should be based on...
Personally I'd be happy to have a 'mere' 1-2 GH unit (or two) that will pay for itself in a month or two that will then help me get some use of all the spare cycles I have laying around the house
While hash/watt is a major factor going for FPGA/ASIC, the fact that you can run them (though maybe not directly power them) from any old box you have collecting dust (or even a RasPI if you are adventurous) also makes them attractive. So what we need to really do is look at the whole system cost - not just the hasher, but everything you need to get the hasher to hash. That's what the ROI figures should be based on...
Personally I'd be happy to have a 'mere' 1-2 GH unit (or two) that will pay for itself in a month or two that will then help me get some use of all the spare cycles I have laying around the house
ASIC's doesnt exists in the real world, or they do, but not as much as it should do, and thet are pretty stupid of themselves, becasue they can only do one thing.....
Huh? They do exist, and they only do one thing, but they do that one thing quite nicely, with very little power! You jelly? The difference is, nothing can really replace ASICs. CPUs were great, but GPUs were better at parallel computing. GPUs were great, but FPGAs could do it with less power. All of those are being replaced with ASICs, but those are the end-all. We'll see incremental upgrades, but nothing as drastic as what we're seeing now (100x improvements in MHs/Watt).
The fact that they can't do anything besides mine BTC is a little redundant. That's what they were designed to do, so that's all they're good for.
ASIC's doesnt exists in the real world, or they do, but not as much as it should do, and thet are pretty stupid of themselves, becasue they can only do one thing.....
Huh? They do exist, and they only do one thing, but they do that one thing quite nicely, with very little power! You jelly? ...newbs are funny. You do know they easily make up 50-60% of the current networks hashrate. And this is only a hint of whats to come.
Yeah, of course they do it nicely, but when they are used? My point is that what happends when all of the BFL rigs at 5 or 6GHash/s hit the market at fully, are they even powerful enough or are they like todays gpu ming rigs at some GHash/s then? CPU Mining is dead, and when ASIC's hit the market at full speed the GPU will die (my opportunity) The good thing is that you can still do alots of many other things with you GPU or FPGA device! Thats my POINT!
ASIC's doesnt exists in the real world, or they do, but not as much as it should do, and thet are pretty stupid of themselves, becasue they can only do one thing.....
Huh? They do exist, and they only do one thing, but they do that one thing quite nicely, with very little power! You jelly? ...newbs are funny. You do know they easily make up 50-60% of the current networks hashrate. And this is only a hint of whats to come.
Jump to: