Pages:
Author

Topic: Will BFL reconfigure the FPGA's to mine LTC and then resell them? (Read 2508 times)

hero member
Activity: 540
Merit: 500
COINDER
I don't even think they know what they are going to do with them yet.

they could turn them in to fpga mini rig thats doing 25gh but sell them not for 15k but for lower price...

Just guessing...maybe they will get hundreds returnd just hang them on a pool some were ... Huh Huh Huh

At their power ratings vs speed, the power will cost more than they'd make after just a couple months of people using ASICs.
true .. but still a couple months making BTC

i also think own 1 gpu@400mh now.. will become the same as own 1 SC@60gh in half a year IF ASIC are for real, only positive is the network security nothing else will change nothing... Wink Wink Wink
member
Activity: 112
Merit: 10
I don't even think they know what they are going to do with them yet.

they could turn them in to fpga mini rig thats doing 25gh but sell them not for 15k but for lower price...

Just guessing...maybe they will get hundreds returnd just hang them on a pool some were ... Huh Huh Huh

At their power ratings vs speed, the power will cost more than they'd make after just a couple months of people using ASICs.

+1, I don't think there will be much use in the bitcoin world, but I am assuming they found some third party that uses them.. or maybe they will daisy chain them all and try to brute force wallets with high bitcoin amounts (jk)
sr. member
Activity: 392
Merit: 250
I don't even think they know what they are going to do with them yet.

they could turn them in to fpga mini rig thats doing 25gh but sell them not for 15k but for lower price...

Just guessing...maybe they will get hundreds returnd just hang them on a pool some were ... Huh Huh Huh

At their power ratings vs speed, the power will cost more than they'd make after just a couple months of people using ASICs.
hero member
Activity: 540
Merit: 500
COINDER
I don't even think they know what they are going to do with them yet.

they could turn them in to fpga mini rig thats doing 25gh but sell them not for 15k but for lower price...

Just guessing...maybe they will get hundreds returnd just hang them on a pool some were ... Huh Huh Huh
full member
Activity: 181
Merit: 100
I don't even think they know what they are going to do with them yet.
hero member
Activity: 568
Merit: 500
Quote
They must know what they are doing, since they are taking so many back, and for the same price they sold them??

Half the price they originally sold them for, and 1 for 1 basis, they're not stupid.

I've always wondered if the NSA would be buying them to get them out of hackers hands, I know I would, if I was bothered about national security. Grin

I assume they could also be used to crack bitcoin wallets too so not good for the bitcoin system either to have all this reprogrammable hashing power out there.
Half the price? A BitForce Single is $599, they give a $600 return for it, when buying a $1299 SC for $699.
Or was that Bitforce S. ever $1200?
full member
Activity: 126
Merit: 100
Quote
They must know what they are doing, since they are taking so many back, and for the same price they sold them??

Half the price they originally sold them for, and 1 for 1 basis, they're not stupid.

I've always wondered if the NSA would be buying them to get them out of hackers hands, I know I would, if I was bothered about national security. Grin

I assume they could also be used to crack bitcoin wallets too so not good for the bitcoin system either to have all this reprogrammable hashing power out there.
staff
Activity: 4284
Merit: 8808
This does not work because the BFL's have not enough memory (RAM) onboard.
And you need that for LTC mining.
No, LTC mining intentionally used parameters _far_ smaller than the scrypt recommendations—  You only need ~128k.
I'm sure that there is enough distributed sram on those FPGAs for that, although there may be routing constraints or not enough room for them and the hashing engines.

The reason the GPU scrypt miners use a lot of memory is that they must run many instances in parallel to get high performance (to hide the memory latency and to use all execution units).  A typical FPGA Bitcoin miner runs just one instance unrolled.
hero member
Activity: 540
Merit: 500
COINDER
they will use them for there OTHER BUSINESS SPECIALTYS THEY CLAIME THEY HAVE

quote; Butterfly Labs manufactures a line of high speed encryption processors for use in research, telecommunication and security applications

cheers
legendary
Activity: 1270
Merit: 1000
Wasnt BTC cpu mined at first? How did that change to GPU? Can the same be applied to LTC with the FPGA's ?

Yes of course, but the proof of work is different... it's been designed to need fast access to large amounts of memory, so you'd be looking at an entirely new pcb - probably with memory on it - and usb likely wouldn't push enough bandwidth to keep up with 1 (and certainly not 5) fpga ltc miners.

No, thats not correct. A single LTC hashing core needs only a little over 128kB for working. Thats why it can run 'fast' on modern CPUs with  enough on chip Cache. The fact that LTC mining with GPUs needs so much money comes from the way GPUs work. (stream processing), so there will filled a large block of memory with task to compute since GPUs are Vector engines that don't work well with branching (and loops).

I am not really sure how much memory the EP3SL150 has, i think the would be 2-4 Hashing cores per chip possible.

All in all it should be possible to construckt a bitstream but it would be not so efficient as for BTC-hashing.

BTW, bitfury did some calculation on LTC Asics, but i have no idea about the state of the project.
sr. member
Activity: 392
Merit: 250
CPu mining was the original way to mine bitcoins but that's because it was the easiest and the computer handled running the instructions 99% on its own.  Someone had to figure out how to interface with Nvidia and AMD cards using their own crappy language in their own crappy SDK to write a custom driver that would send data in an efficient manner to the GPU for processing then interpret the results.  That took a while Tongue
sr. member
Activity: 420
Merit: 250
Wasnt BTC cpu mined at first? How did that change to GPU? Can the same be applied to LTC with the FPGA's ?

Yes of course, but the proof of work is different... it's been designed to need fast access to large amounts of memory, so you'd be looking at an entirely new pcb - probably with memory on it - and usb likely wouldn't push enough bandwidth to keep up with 1 (and certainly not 5) fpga ltc miners.

hero member
Activity: 605
Merit: 500
Wasnt BTC cpu mined at first? How did that change to GPU? Can the same be applied to LTC with the FPGA's ?




I kinda breezed over the theory behind LTC but they state it's optimized for a CPU's architecture plus a ton of RAM.  That means for them to make a faster solution, they'd have to outdesign Intel and AMD and use the most modern (patented) memory controller technology.  The best of the best belongs to AMD at a blazing 1866MHz native speed for their new trinity APU, unless you count GDDR5 on graphics cards but I think that works sort of differently.

In other words, CPUs could do bitcoin stuff, GPUs did it way better, and ASICs are exactly what you need to run bitcoin calculations at the fastest possible speed.  With LTC, CPUs are perfect for it so to make something better, you have to design a better CPU than say an i7 Tongue  At least that's what their wiki seems to suggest.
sr. member
Activity: 392
Merit: 250
I kinda breezed over the theory behind LTC but they state it's optimized for a CPU's architecture plus a ton of RAM.  That means for them to make a faster solution, they'd have to outdesign Intel and AMD and use the most modern (patented) memory controller technology.  The best of the best belongs to AMD at a blazing 1866MHz native speed for their new trinity APU, unless you count GDDR5 on graphics cards but I think that works sort of differently.

In other words, CPUs could do bitcoin stuff, GPUs did it way better, and ASICs are exactly what you need to run bitcoin calculations at the fastest possible speed.  With LTC, CPUs are perfect for it so to make something better, you have to design a better CPU than say an i7 Tongue  At least that's what their wiki seems to suggest.
hero member
Activity: 616
Merit: 500
This does not work because the BFL's have not enough memory (RAM) onboard.
And you need that for LTC mining.

yup. GPU-z reports a total of like 40mb of ram used between BOTH my 7950's when btc mining.

3gb total used (1.5 each) for LTC mining
sr. member
Activity: 420
Merit: 250
They must know what they are doing, since they are taking so many back, and for the same price they sold them??

Their customer service hardware is almost too good to be true. Maybe thats why so many people are sceptical confused?

FTFY - but in all fairness, they're turning over a new leaf with Josh joining the company, and he's gotten more random stuff to enhance customer service done since he started than they did the entire year before he was there.

hero member
Activity: 605
Merit: 500
They must know what they are doing, since they are taking so many back, and for the same price they sold them??

Their customer service is almost too good to be true. Maybe thats why so many people are sceptical?
sr. member
Activity: 560
Merit: 256
I mean what will they do with so many FPGA's?  Huh

1. Sell them to .edu, .org institutions that have a need for computing power
2. Keep them as back-up to secure/mine the network in case let's say algorithm changes and all ASICs are made obsolete.
hero member
Activity: 576
Merit: 500
Under their drivers page it lists the following:

Drivers:
Computational research
Medical imaging
Packet integrity verification
Generic fingerprinting engine

So I imagine they can be reprogrammed for stuff like this.
sr. member
Activity: 420
Merit: 250
I mean what will they do with so many FPGA's?  Huh

Interesting question - maybe they'll apply them to their other potential markets (listed on the bfl main page) being programmable would be an advantage if they're trying to develop into another market - then they could make an asic produce release that and repeat.
Pages:
Jump to: