Bitcoin Forum

As difficulty ramps up, folks look for more powerful and efficient ways to mine.  At some point in the past someone when "You know...if I use my GPU instead of my CPU I could really hammer out some block."

So, what do you all think the next 'big thing' will be?  Will it be FPGA's?  Will it just be huge GPU farms taking over?  Something out of the blue?

Let's try not to start a flame war and see if we can have a civil conversation about this. :)

New GPUs will be competing with ASICs and FPGAs.

I propose wind/solar powered mining farms, which would make mining free. ;)

LOL, I put in a search for "best hardware for calculating hash rate", and all I am getting is info on ganga !!

My money would be on the Chinese with massive GPU farms.

I think some guys will Buy some Big Server Farms only to get Bitcoins.
I also think that some big companys will start Mining.

You really think so?  I'm interested to see the new AMD 7000 series cards as well, but seeing that AMD didn't really improve mining performance from the 5000 to 6000 series, I'm wondering if the 7000's will actually improve mining.  Anyone seen anything that leads them to believe it might?


I guess this isn't unexpected.  If you look at something comparable like the Gold Rush in the US, at first it was great for even little guys, but eventually it got to the point that the individual prospector couldn't compete and big conglomerates took over.  I wonder if in the long run this hurts BTC though, because a lot of the interest/charm it had was the fact that _anyone_ could do it and had a chance to make some money.  At some point you'll need to put in some serious coin to even have a shot of getting any coins and I wonder if that'll put a damper on things.

Has anybody tried the PadLock features on the VIA CPUs? http://www.via.com.tw/en/initiatives/padlock/features.jsp (they say up to 5Gbit/s!)
The boards are very cheap and should consume much less power than a GPU...

Southern Islands isn't a new architecture, it's just a 40nm -> 28nm die shrink + optimizations made on the 6000 series platform.

It will be 40-60% faster at most. Early adaption wont be profitable because the cards will cost too much relative to well-performing 5000 series cards which will become even cheaper as gamers, casual users etc. dump them for newer cards.

Nvidia Kepler on the other hand is an entirely new ballpark compared to the Fermi architecture (gt/gtx 400-500), but the OpenCL performance will probably be too poor to be considered by miners.

7000 should be an improvement if all other things are equal due to the smaller dies with the 22nM process. (I think its 22nM...)

ASICs won't ever be cheap (as they have to be custom made), FPGAs will only start to pay off if the difficulty stays fairly constant over half a year or so OR if you can get them for free (more or less).


As the most exiting thing for most people seems to be either:
"I can produce money with my computer!!!11"
or
"The exchange rate went up like OVER 9000%!!!!11"

There seems to be demand for both buying and producing BTC at the moment.
Buying actually is REALLY hard, especially if you want to buy fast, this is why a lot of people seem to rather start mining.


Anyways, I think it will still take some time until the hashrate stabilizes, so the biggest hashrate-burst as fast as possible currently wins. This means GPUs won't even be beatable by the cheapest available FPGAs in the near future income wise.

I think OP is talking about a revolution;

An evolution assuming that bitcoin mining drastically effects the computer hardware market, will lead most likely at first to specialised non-ATX-standard motherboards with an obscene number of PCIX slots; too me this seems the most likely as a possible first change in the computer hardware market; I say this because the motherboard is the easiest to re-design; re-designing a GPU takes many years...

This and a combination of 'overclocker' style hardware modifications, a stripping of everything non-essential to bitcoin mining (PCI slots, extra USB ports, anything un-wanted by the miner... or stripped down video cards using current GPU technology ie cards with no display ports, 300~MB of ram, etc)

This is I guess something I feel is quite likely in the short term

Despite Southern Islands (AMD 7000 series) being only a die shrink, I still think they will lend a vast improvement. From what I understand the only reason the 6000 didn't really improve is because AMD changed the shaders to run two instructions.. similar to nvidia's architecture.

In this die shrink they have the opportunity to cram a bunch more shaders in there, boosting tessellation performance (which is something they are in a huffy over with nvidia lately) and subsequently OpenCL/mining perf.

Even if not, the thermals and wattage will be reduced leading to greater over-clocks and lower power consumption.


Personally.. I can't wait till they release.

I agree on the 7000's.  Even if they are great for mining, we all know the first few cards out of the gate for a new architecture command a big premium, so I'd bet that while they'll probably knock things like the 5970/6990 off the radar, the price premium will still mean that the 5830/5850 type cards are the best MH/cost.

I'd love to learn more about ASIC's & FPGA's, but from what I can gather so far, they'll be price prohibitive for most folks and GPU's will reign for a while.

The long term problem I see is that the network power continues to grow as more people get on board (which is great), but it begins to knock some folks out of even trying since at some point you'll probably need to plunk down some cash for 5-10 GPU's to even make it worth while to mine.  Yeah, you'll be able to mine with one card and probably turn a small profit, but in the end you have to balance whether it's worth it for the aggravation of keeping that card going for months on end just to break even.

So, I'm betting that it is going to keep along somewhat the same trajectory, we'll probably see difficulty bumps of 25-50% for a few more months, then it'll start to flatten out.  If that's the case, I think I agree with you guys that GPU's are going to be king for a while.

Best mining in the future will be whatever AMD releases to compete with nVidia Teslas, IMHO. A 1U server with 8 GPUs in it would do nicely, in my opinion.

Naturally, the next evolution in mining technology will be smaller graphics cards plugged into our existing video cards. I could probably connect at least 4 smaller videocards to a 5850 and it would take care of them like a mother dog and her pups. This is the natural progression, as the previous step was plugging a lot of cards into motherboards. Now that the video cards are all grown up, they are motherboards themselves and now the motherboard is like the grandmotherboard or something. I don't have the details worked out beyond the sketch that I drew on my bathroom wall while I was showering. Okay actually I wasn't showering but delivering a deuce. I didn't want to be inappropriate, but now I just realized that you all would probably appreciate my honesty.

Anyway, if someone wants to improve upon my sketch regarding this idea, I'll leave my front door and the bathroom open, and you can just use whatever color of marker you feel like using (smell them too. They're fruity!). The sketch that I'm talking about is next to a story I wrote about the adventures of the Shampoo King. You'll see it.

Also, I think we could get a lot more out of mining by installing some sort of refinery card. You'd plug that into your ethernet port and it would refine bitcoins before they got to the network, making their ones and zeroes firmer, more colorful, and just generally more sociable. It would also get rid of all the twos and sevens and crap. God I hate sevens. You couldn't see, but I just rolled my eyes dramatically. Just before they go to the internet, I bet you could let the bitcoins ramp up their speeds with some sort of particle accelerator technical machine for helping the electrons achieve all of their dreams and aspirations so they would help get the bitcoins to nodes a lot faster.

Another idea is to speed up time, but only inside of my computer. This doesn't even require changes to the hardware, so it's a great idea and I really don't think anyone can beat it. I don't know how to do this technically, such as by using javascript, but I have an old alarm clock and I'm going to experiment with making the minutes go by faster. Maybe my computer parts will think "Oh shit, we're pretty slow, dudes. According to that clock over there, we're mining at like a tenth of our normal speed". Then they'll speed up, and I'll laugh, but in a really high, fast voice so that it sounds perfectly normal to my computer parts. If that doesn't work, then I think we should all just agree to make a "second" be redefined as like 20 seconds. Then I can say that I'm mining at 2000 Mh/sec.

All very thought out, well-defined plans with a good future.

I wish you luck.

It's called the FPGA/ASIC clusters already owned by big financial players/knowledge management/intelligence communities.

GPUs are made for head-smoking, not hashing.

After ASICs, they'll revive Edgar Cayce and have him divine hashes.

Next difficulty increase will reach the 1e6 point, maybe 2e6. There are only 30e3 miners right now... it can become widespread.


I agree. People think mining is not profitable... but in the long term the best is to control supply. In my opinion we should create a corporation to keep bitcoin safe from those hungry bastards. This could be seen as contradictory, but it actually isn't.

Do you even know what "ASIC" stands for?! Probably not, or you would know, why what you just said is completely stupid.

About FPGAs: They are maybe good at MH/W performance, performance per chip wise they quite honestly seem to suck though.

Australian Securities and Investments Commission  :)


http://www.xilinx.com/company/gettingstarted/fpgavsasic.htm#pcs

FPGAs are not exactly cheap, but they're quite powerful.
An XC6SLX150 will get you ~200MH/s for ~20W. And if you build an FPGA cluster yourself, you'll end up with costs around 200-300$ per chip. Because they don't use much power you can reach very high density. Oh, and ASICs might actually be feasible once you reach quantities >10000.

Nvidia may get wise and decide to invest a little money into a Bitcoin R&D project. The fact is, Radeon's are not in stock anywhere and, questionable investment or not, they would probably like to be the ones selling out of cards.

I imagine both companies have probably formed ad-hoc Bitcoin divisions already to examine this phenomenon.

They only need to re-design some of their existing tech by stripping out everything unnecessary. If they can target relatively lower power usage, smaller form factor, less heat, and combine that with a new library to replace OpenCL cutting out everything unnecessary to the hash function, maximizing profit from the device while providing a hot seller, even if it's somewhat temporary, they'd really have something.

If I was an executive at Nvidia, I'd be scrambling to try and put together a business unit to capitalize on it fast.


I registered just so I could post that. Been lurking here for awhile. :-)


I've been feeling a little left out since all I have to hash with is an old 4650. I snagged a single 5830 from Newegg this morning but I don't have enough power supply to run two.

I'm just hoping to mint a few coins to buy some hosting with. I think I may have 0.01BTC hashed so far. :-)

?

ASIC's can only do one thing (Application-specific integrated circuit).  You can't really modify a high frequency trading ASIC for bitcoin hashing...  Maybe an FPGA setup, but even that's questionable.

I also doubt that those "big financial players" even use FPGA clusters for anything.

I'm sorry, but I have to ask.  Do you know what a FPGA is?  It stands for a field-programmable gate array.  It's an electronic circut that can be programmed at the hardware level, in the field.