Are GPUs still relevant and for how long

[aznewsh]

I have tried to do a lot of research and used calculators etc, however there is a lot of guesswork and I would appreciate more informed opinions.

With ASICs on the horizon - If I were to invest in GPUs:

1) Am I likely to get at least my money back?

2) Are there realistic options in Alt currencies that the GPU could be useful for vs ASIC - can ASIC at this point or in the near future be used in Alt?

3) Are there any realistic ASIC options at this point, all the sellers seem rather dodgy to me?

4) What other angles could I look at other than actually buying Bitcoin of course?

Feel free to add any other pearls of wisdom that spring to mind...

Thanks for your time.

[odie158]

The manufacturers do seem rather dodgy don't they? I am personally not willing to send thousands of dollars to a website for the promise of fortune later (unless of course it's a Nigerian widow).
I believe that if ASICS are that good, mainstream companies will begin producing them and the price will drop once they are in full production. I say wait.

[kuriboh]

Well I think when you can't make profit with BTC change the mining to an alt coin with the same GPU.
Anyways you are less likely to loss completily you investment because GPU's can easily resell to a PC Gamers.

[nwoolls]

Here's what I've found in doing similar research and then crunching the numbers in available profit calculators. Your realistic options for things you can do to mine right now are:

• Add a powerful GPU to an existing PC (600 Mh/s - $280)
• Build a 4 GPU rig specifically for mining (2400 Mh/s - $1600)
• Buy an ASICMINER Blade (12000 Mh/s - $6500)
• Buy an ASICMINER USB Erupter (333 Mh/s - $275)
• Buy a Lancelot FPGA (400 Mh/s - $350)

If you plug these into any of the mining calculators I think you'll find that the most of the options break even in 3-4 months. After that you are looking at things like:

• You can re-sell the GPU's but they are very hot and draw a lot of power
• You can re-purpose the FPGA's or resell them
• An ASIC will only ever hash for Bitcoins but draws very little power

Of course then there's all the preorder mumbo jumbo with various companies, the group buys for chips that you then have to build into a working PCB, and the ASICMINER shares.

[massnerder]

Well I think when you can't make profit with BTC change the mining to an alt coin with the same GPU.
Anyways you are less likely to loss completily you investment because GPU's can easily resell to a PC Gamers.

This.  The GPU rig you can point at a profitable scrypt coin after BTC hits a difficulty level that makes it unrealistic, or sell it.

[Kuzushi]

A quick search on Ebay will show that there's a feeding frenzy going on for ASIC mining gear, and even the new Asicminer Block Erupter USB's are selling for nearly twice as much as you could buy them for on this forum.

It's a little bit of a risk, but I would buy (for example) two Asicminer Block Erupter USB's on this forum for about 2BTC with the intention of selling one of them on Ebay, and with a little luck you'll sell it for enough to bring your cost basis to nearly zero on the one that you keep, making the ROI a non-factor.

[erpbridge]

You can re-purpose the FPGA's or resell them

Do tell... how exactly can you re-purpose the FPGA's? You already had them running SHA256, so you're looking to get away from that... can't repurpose them to Scrypt, I already tried asking a question to try to figure out what path we could go down to retrofit them for memory access Scrypt needs, and was given the "here's a quarter, buy a clue" video as my first response in the thread.

SHA256 and Scrypt are the two coin models. What other profitable usage do you have that you are suggesting to repurpose the FPGA for?

[kinitex]

GPUs will be a viable option for cryptos for the next year atleast imo.

[davecoin]

I think one of the neatest things about GPU mining is the fact that you can buy them with fiat 

Then "getting your money back" is subject to the BTC exchange rate.  They also have potential resell value.

[suryc]

You can re-purpose the FPGA's or resell them

Do tell... how exactly can you re-purpose the FPGA's? You already had them running SHA256, so you're looking to get away from that... can't repurpose them to Scrypt, I already tried asking a question to try to figure out what path we could go down to retrofit them for memory access Scrypt needs, and was given the "here's a quarter, buy a clue" video as my first response in the thread.

SHA256 and Scrypt are the two coin models. What other profitable usage do you have that you are suggesting to repurpose the FPGA for?

FPGAs are programmable chips. They have many uses and they were not created for bitcoin/altcoin mining. They can be re-purposed to do a variety of things (by someone that has expertise in FGPA programming). So, they do have an intrinsic value even when they are no longer profitable to use for mining. See wiki here: http://en.wikipedia.org/wiki/Field-programmable_gate_array
Maybe not a profitable use for you, but you could sell them to someone that wanted to re-program them to do something else.

FPGAs could in fact be used for mining scrypt, the problem is that the current designs do not have onboard memory and programming them to use system RAM would cause them to be slow and inefficient. This is why GPUs are so good for scrypt, they are essentially ASICS that have fast access to high bandwidth DDR.

It would be possible to design an FPGA for mining scrypt, but it would be different from the FPGA chips that are out there now. I have heard they have been prototypes on an Altera board for mining scrypt, but cost vs performance has put it well in excess of a GPU.  GPUs are already mass-produced and their design is very close to what you would do if you wanted to design an FPGA or ASIC for Scrypt, so for the foreseeable future, GPUs will remain the most effective method for mining scrypt based alt-coins.

[BeetcoinScummer]

I'd like to see someone repurpose an FPGA that is already soldered on a board designed purely for SHA-256 hashing.

I doubt that is it would be cost effective (or even possible) for most miners unless they have hundreds of FPGAs and can contract out the work.

[Kinetic915]

why not buy and sell some of those asicminer usbs on ebay? make double your money that way

[nwoolls]

why not buy and sell some of those asicminer usbs on ebay? make double your money that way

It's a good idea but apparently Ebay is canceling auctions related to Bitcoin mining.

[erpbridge]

You can re-purpose the FPGA's or resell them

Do tell... how exactly can you re-purpose the FPGA's? You already had them running SHA256, so you're looking to get away from that... can't repurpose them to Scrypt, I already tried asking a question to try to figure out what path we could go down to retrofit them for memory access Scrypt needs, and was given the "here's a quarter, buy a clue" video as my first response in the thread.

SHA256 and Scrypt are the two coin models. What other profitable usage do you have that you are suggesting to repurpose the FPGA for?

FPGAs are programmable chips. They have many uses and they were not created for bitcoin/altcoin mining. They can be re-purposed to do a variety of things (by someone that has expertise in FGPA programming). So, they do have an intrinsic value even when they are no longer profitable to use for mining. See wiki here: http://en.wikipedia.org/wiki/Field-programmable_gate_array
Maybe not a profitable use for you, but you could sell them to someone that wanted to re-program them to do something else.

FPGAs could in fact be used for mining scrypt, the problem is that the current designs do not have onboard memory and programming them to use system RAM would cause them to be slow and inefficient. This is why GPUs are so good for scrypt, they are essentially ASICS that have fast access to high bandwidth DDR.

It would be possible to design an FPGA for mining scrypt, but it would be different from the FPGA chips that are out there now. I have heard they have been prototypes on an Altera board for mining scrypt, but cost vs performance has put it well in excess of a GPU.  GPUs are already mass-produced and their design is very close to what you would do if you wanted to design an FPGA or ASIC for Scrypt, so for the foreseeable future, GPUs will remain the most effective method for mining scrypt based alt-coins.

So again I ask... you're proposing to repurpose those FPGAs (as opposed to selling them.... repurpose implies you are using them again for your own purposes.) What way, right now, is a profitable way to use those FPGAs that you have in hand for YOUR use? As you agreed, you can't use it for Scrypt. I tried asking the question here: https://bitcointalk.org/index.php?topic=187772.msg1946610#msg1946610 ... and you can see what the answers were. In short, once that FPGA is not profitable in SHA256 mining anymore, it has no other purpose to YOU. (This is why I keep suggesting either a project to retrofit FPGAs to Scrypt, cannibalize them to make a new Scrypt FPGA, or develop a third mining algorithm that does not have the memory requirements of Scrypt, allowing FPGAs a cryptocoin life beyond SHA256 once ASICs fully trounce that.)

I would love to have the knowledge of how to make an add-on board for existing FPGAs, make it something like rather than computer-USB cable-FPGA, have computer-USBcable-memory module-FPGA, where the memory module has a USB in from computer, and a USB out to the FPGA. Or, take the FPGA's in, desolder the chips, pop them into a new board with memory, and send them back out.

But, when I try asking the question, a Hero Member gives me this: Let Josh answer your question: http://www.youtube.com/watch?v=UlWrmIqGs3Y&t=2m48s

This board is pretty hostile to anything other than BTC only, though...

[aznewsh]

Thank you all for taking the time to provide feedback - I am going to try a 3 card rig with the hope of not losing too much and will treat it as an hobby, then watch how things develop...

[pheaonix]

Thank you all for taking the time to provide feedback - I am going to try a 3 card rig with the hope of not losing too much and will treat it as an hobby, then watch how things develop...

this is a bad choice, buy klondike boards

[erpbridge]

If you're starting from scratch like I was a couple months ago, here's a roadmap I've been following:

-Utilize your current machine while you wait on orders. Your current machine can probably mine a little, sometimes.
-Choose which route you'll take, and how fast you want to start. Videocards... you can have them shipped with Amazon Prime and overnighted if you so desire. FPGAs, I can't seem to find anyone who is actively making them and shipping within 1 week from order. ASICs, again, no one seemed to be actively making one for a week order when I started... now there's Block Eruptor USB's, but they are minimum batch 100*2BTC=20 thousand bucks, or get in on a group buy and wait for them to be distributed.

I decided to go video cards, since I had a spare older machine hanging around collecting dust that had PCIe slots, and since at the time FPGAs were rare to find, and ASICs were still on horizon.

Mining Rig
----
-Make a spare miner machine. Tried mining with my active everyday desktop machine... and its not really the best option to make a profit when you're NOT mining more often that you ARE.
-Grab an old motherboard with a few PCIe slots. I had an older one hanging around, Core Quad processor, 1x PCIe 16x, 2x PCIe 1x. Its 5 or so years old.
-Buy as many video cards ASAP that you can to populate your 16x slots. My suggestion would also be to start with 7950s, but theres 17 different arguments about who/what to start with. If you want, place your order for the ones in the 1x slots, but know you're also going to wait a while for those to run as you'll have to wait for...
-Buy as many riser cables as you need, INCLUDING 16-16's. My board, one of the 1x slots was obscured by the fan shroud of my 16x card... couldn't use it until my 16-16 riser came along.

For your first couple months, if your outside world can afford it, mine as a hobby, and just record your extra electric costs, the cost that you sunk into your cards, and the cost you sunk into your hardware.

Mine some coin. Use dustcoin and coinchoose to determine which at the right time, unless you're morally opposed to Scrypt, and morally opposed to any other SHA256 than BTC. Actively trade it toward BTC at Vircurex/BTC-E/Cryptsy. Invest some to buy dividend producing securities at Bitfunder or BTCT. I'd probably suggest looking at one of the partial ASICminer stocks like TAT.asicminer, until you can afford a full asicminer stock. This will give you some additional income, but not necessarily as much as your card gives you. More importantly, its relatively constant income that you aren't paying any electricity or hardware cost to... I look at it as that my already sunk card investment is averaging a few more hashes each week, for the same electricity as I pay.... and moreso that when I get a week outage from a East Coast hurricane, that I'll still make some income.

I did this until I had a full Bitcoin into stocks, and am now saving up enough crypto to pay myself back for electric bills [I've been paying the increase out of pocket, but want the whole thing to be self sufficient over time]. My goal after that is to save some up for an actual BlockEruptor USB so that my average hash/W is better (compare 1100hash/400W, and add 300h/2.5w to make it 1400hash/402.5w.) Then, throw a little more into ASICminer shares, then start paying down my initial hardware investment.

Just my roadmap... Yours will be different.

[af_newbie]

Thank you all for taking the time to provide feedback - I am going to try a 3 card rig with the hope of not losing too much and will treat it as an hobby, then watch how things develop...

You have to treat it as a hobby.  Less stress.

Besides, mining it is not that expensive comparing to other hobbies. 

Not too many hobbies let you earn back a bit and in most cases, pay for themselves.

[GenTarkin]

they are relevant as long as you can turn a profit from them over power. They can always be sold for nearly what you paid, given they are 7xxx series...gamers will always eat em up.

[suryc]

they are relevant as long as you can turn a profit from them over power. They can always be sold for nearly what you paid, given they are 7xxx series...gamers will always eat em up.

Do you think there will still be a market for resale to gamers when AMD releases the 9xxx series later this year (they are skipping 8xxx)? It is not officially announced by AMD, but seems to be widely reported that its coming in late 2013.
http://www.mcvuk.com/news/read/amd-confirms-new-volcanic-island-gpus-for-2013-launch/0115398