Topic: FPGA Inflection Point (Read 7997 times)

well im in. i buy three of these 500 Asic,s rigs if someone made them and im sure im not the only one in the waiting list

well i see that most of the people here are in favour of GPU's.. but fpga are not as costly as we think...

I am not a miner, rather an fpga system developer. and if the estimates on these forums are accurate, i am quite sure that the fpga can take over gpu anytime. it is just till the right product gets in the market. 

If bitcoin ever gets big enough for "an entity" to shut it down, by that time the number of miners will be so huge they won't waste time building some massive computing grid to control the chain, or to flood the market with bitcoins to devalue it, as you pointed out that is worthless.  They will simply go after the exchanges, these are easily leaned on single points of failure that if taken out wouldn't destroy bitcoin, they would simply make it much less desirable as it wouldn't have any link to other currencies and make it a dead end.  Alternatively they could lobby for laws or political power to make bitcoins illegal in some way.

I'd think international banking cartels would be more of a threat. But regardless, let's assume there is some threat entity [hereafter: "TE"] that would try to shut it down. The question is: Is there even a viable possible catastrophic threat to bitcoin(?)  Even if TE could swamp the network down and suck up all the remaining coins, there would still be 30% of all coins ever to be released already in circulation. TE would be holding, let's assume for argument, 70% of all bitcoins. So, what does TE do with them? Dump them on the market? Ok, then what did TE gain? What if TE hordes them and destroys them? Ok, bitcoins value go up due to lower supply... but bitcoins are divisible up to 8 decimal places. So, instead of there being a possible 2.1E10^15 coin fractions, there are then 0.7E10^15 or 7.0E10^14. Let's see, that's 700 Trillion coin division units. Assume a worldwide population of 10billion, and that leaves 70,000 units per person to spend. Of course, this is an extreme scenario and discounts that babies & very young children might not need to be counted, but it shows that there are sufficient increments remaining to perform as a mechanism of exchange fairly easily for all people - I think.

However, perhaps the threat possible is not related to hording or dumping...What is that threat? Hacking the bitcoin process? Maybe, it's just demotivation to any potentially new bitcoin miners or holders. How is bitcoin compromised? And more importantly, what are potential counters to that? Afterall, it is suppose to be designed to be extremely secure.

Yes that is very true it is an amazing ROI, but the market is so volatile and young, putting that much money into it is a huge gamble, you don't know if BTC is going to be worth $1 or $1000 in 2 years, or even if it lasts that long.  I think if you explained what bitcoin is and the small history no one in their right mind would invest, there are just too many unknowns.

625 day payoff is not very long.  50%+ ROI in one year is amazing.  I just bought a house and it's only producing a 25% ROI per year.

My point is a company could make this investment tomorrow, have the system online in a month, and be way out in front in the development of FPGA mining while making all our GPU's worthless.  I know we can sell, yadda yadda, but 10,000 GPU's being dumped on ebay at the same time will not sell for what you think.

And they blow whenever you want them to...

Big deal. You probably have plenty of fans in the spare parts bins...

You missed the joke.

I lol'd.

I have plastic bins full of everything on the list, and then some.

The wife's not a fan.

I'm still sticking with $650 for the described system in the opening post.  The pricing of the video cards is pretty optimistic and doesn't seem a reliable price.  It would've been much better if MSRP was used rather than "... I saw this part on NewEgg, and this other part on Ebay.... TigerDirect sold these last year for $30..."  Granted, your amateur miner isn't going to go pay MSRP for everything - but using MSRP further validates the point.  Even a weighted average price from retailers ($109 is not it....) would be more accurate and supportive.

Also, I hate it when people try to claim cases increase heat (or by proxy no case makes the system run cooler).  With no circulation (extractors/fans/etc), you only have ambient thermal transfer and a liquid cooler wasn't in the original description.  Not to mention you have the "oops" risk of damage.  It's just silly a silly myth and a simple hack that people enjoy doing.

PXE boot is a great idea - for one of my arrays, I don't have that ability because of restrictions on the network.

As other people are constantly reminded - the mining machine is a resellable asset.  Not to mention most of the items required to make a decent mining commitment (300MHs) are already available in most peoples homes in the form of a gaming PC.  FPGAs are a much more extreme investment and resale isn't nearly as accessible.  PC mining is nearly "idiot proof" now.  Building and managing FPGA arrays.... not so much....

I just did some calculations, and ASICs might actually be affordable.

Assuming one can sell 500 of them, building a "bitcoin accelerator card" which yields 5 gigahashes might be possible for less than $1000. If one wants to build one big (copacobana/deepcrack-like) mining rig instead of selling cards, one could possibly reach 5 terahashes (and thus attack the network) with less than a million dollars, probably more like half a million. (This beast would contain 2000 chips.)

You can add once off fee for 1500w(equivalent) aircon that cost me $400, thats what I am using in a 2.5mX2.5mX2m room, and it keeps the room chilled at 10celcius.

The secret with cooling is, small room - open layout(nocases) - cheapish aircon. This works alot better than some of these extraction fan contraptions Ive seen posted on here.

One thing that was not included in calculations, is that for GPUs, it does not scale all that easily. Space for boxes is not a big issue, but cooling installation takes some money, and keeping it cool also costs some elecricity. I dont have enough knowledge to give you reliable estimations, but it's really not that easy to keep room cool with even 10 of such boxes.

Also, I believe FPGA solutions are still immature at the moment, and cost of it will go down. For cost comparison to be more useful it would be nice to include more details about FPGA cost as you did for GPU setup.

And thanks for nice comparison.

Government intervention is the bigger concern right now, IMO (and no, I'm not suggesting any government is anywhere close to looking at BTC as a problem yet).

If you could build a GH/s machine for even $1000, you could recreate the current power of the network for under $5M with simple consumer hardware. 

That is pocket change to the US Gov. 

Add in labor, DC build out, NOC, etc and you could build a 5TH/s+ BTC mining DC for easily $10-20M to break the 50% barrier.  It's fine to say stuff like "the top 500 supercomputers don't equal the processing power of the current network" but 5000x 1 gigahash machines do.  I guarantee you with a $20M "btc-killing" budget a company could get it done.  This is certainly not a single-purpose datacenter either and has a lot of potential value for defense in general.

Hopefully by the point world governments take notice the network is already in double digit TH/s range, but right now it's not as invincible as people seem to suggest. 

How comfortable would you be trading stuff for BTC if you can't get any confirmations that they money wasn't double-spent?

The modem is so you can keep mining after your ISP drops you like a hot potato for "illegal" activity. The prudent would be working on setting up open mesh networking, but that takes time.

Note: the cheap dial-up plans are often limited to 20 hours/month.

For USB KEYS
if you have many rigs like more than 10 rig, you can use PXE to boot linuxcoin, save the cost of USB key

At that point the economies make no sense.  At 3.3M difficulty, with your 250M FPGA you will be producing 0.08 BTC per day, so at $10/BTC you would pay off your FPGA in 625 days, in order to pay off your FPGA in the same 30 days as you can pay off your GPU currently BTC would have to be worth $208/BTC.  Because entrance to FPGAs is MUCH higher than GPUs that will push their usage back.  And you forgot to account for people with free power which may very well eliminate FPGAs usage all together.  FPGAs will probably never be popular for mining due to slowing difficulty and ever lower ROI.

!!!!

Getting hands confused during one handed keyboard moments  ?


LOL, sorry 

Holy crap guys stop just trying to add variables like "desk" and "the keyboard" and "monitor"

Most comp-techies have a pile of MISC comp parts, Things like crappy ram, Mobo's old PSU's, Sticky keyboards. So thats why he left so many things outta the calculation.

And there are such things a Molex-4-pin into PCI-E 6pin adapters... Like ffs the PSU doesnt need to have them, It just needs to fuel them

Oh and a case? thats not even neccesary, And we all have spare cases

56k modem !

what decade are you in ?

Anyway, you don't need any of the items you listed, once you have your first machine, that is.

I have 1 monitor, 1 keyboard, 1 mouse, KVM, and 4 machines

3 are dedicated miners with bootable usb with Linuxcoin

Okay, okay, fans might be a good idea

CoinSpeculator, Bravo!!!

Damn close to calcs I've made privately half a year ago.

While the price figure for a mining rig is debatable, the point about government involvement is interesting. Anyone with more technical knowledge want to tackle that one?

$50 for a quality PSU that can support 600w+...? who's being ridiculous here...? I have two really good 630w/650w PSUs and would definitely be nervous putting three 5830s on them, hell, even three 5850s (which use less power) would make me nervous...

LOL, u would have been better off piggybacking onto fpgaminers thread!
i went thru that whole thread yesterday and not yet having gotten into mining but planning to do so i have a sincere interest in trying to figure out what to buy for the longterm.  i appreciate u trying to get this thread going and i hope it matures to some more helpful information for us all.

by undervolting from 1.088 to 0.95V I think I could get my 5850s to about a 30% higher hashing efficiency (hashes/W)

if you could undervolt further it might get even better

Don't forget the DESK CHAIR!

case? operating system? fans? keyboard? mouse? printer? scanner?  fax machine?  56k modem?!?

Ya.... what's the harddrive for?

I use a revisor image on a thumb drive for 20+ PCs.....

I'm pretty sure I can boot it in < 256M of RAM.

In my opinion, $500 is a bit optimistic.  I would say $650 just based on the cost of the mainboard with legitimate (not hacked, not cabled) PCIe slots, PSU and whatever crappy case you can find.  A DVD-ROM or permanent storage would be nice..... it sucks booting 20+ machines at a time with a thumbie.

Has anyone tried to build arrays of ITX boards?  Some of the newer ones have at least one PCIe slot and don't cost more than $100.  No idea what hell you would have to play powering the video card.....

USB = $1

Cheap Memory = $10.

You people are ridiculous.

hard drive? memory?

The GPU's come with sata to 6-pin connectors.

Power Supply Unit.

power support unit?

what is a PSU?

Sorry for derailing your thread, but the Sapphire 5830's I've been looking at needs two 6-pin PCIE power connectors each and from what I've read need ~200W each. Can't find a PSU on Newegg that has enough connectors and is 700W+ for under $100.

Newegg?  Only need 2X350.  Maybe even a 650 would do it.

I'd like to know where I can get a PSU that supports 3x5830 for $50.

So everyone is well aware that GPU mining is the dominant way to make bitcoins right now.  There is quite a bit of talk about the CIA or google setting up a FPGA or ASIC farms and owning the network.  What I would like to do is examine some of these events:

A) The point where GPU mining is worthless.
B) The point where FPGA mining is more profitable.
C) The raw capital cost of instantly putting the network into these conditions.

------------------------------------------------------------------------------------

First let's start with what appears to be the cheapest setup right now.

3 x 5830 = $320
Motherboard = $100
Power Supply = $50
Processor = $30

Total = $500

My 5830's are putting out around 270Mhash/s, but I've seen 300 possible.  Approximate power use is 200W each.

900 Mhash for $500 and 630W.  1.8 Mhash/$, 1.5 Mhash/W

This machine would produce around 2 coins a day and use 15 kwH of electricity.

Assuming electricity is about 13 cents, this means we currently make ~$20 at a cost of ~$2.

Which leads to point A:

Difficulty has to increase approximately 10x current values at current prices to make GPU mining by existing setups unprofitable

It is clear that we will all be mining for quite a while unless prices fall drastically.  Investments in rigs are still profitable at this point, paying themselves off in about a month.  Considering another difficulty increase coming up, maybe 2 months.  Network size should continue to grow.

------------------------------------------------------------------------------------

From what I have cobbled together in the various FPGA threads, FPGA's have essentially the inverse cost of GPU's.  Rather than being about 2 Mhash/$, FPGA's are 2 $/Mhash.  In exchange for this you have virtually 0 power cost (for the sake of these 10,000 foot projections).  Allocating our identical $500 gives us a rig that might do 250 Mhash/s.

So we've invested $500 for a rig that makes ฿.5 or $5 a day.  We are tasked with finding the point where GPU mining only yields the same amount as FPGA's.

20$ * 1/x - 2 = 5$ * 1/x

x = 7.5

Difficulty has to increase 7.5x current values to make FPGA's more profitable at current prices.

------------------------------------------------------------------------------------

To push difficulty to the level of 0 profitability for GPU's, hashing power must go from 4 thash/s to 32 thash/s at current exchange rates.

To push difficulty to the level of FPGA>GPU profitibility hashing power must go from 4 thash/s to 25 thash/s at current exchange rates.

------------------------------------------------------------------------------------

At $500 per Ghash the US government would only need 16 million dollars in GPU miners with a power cost of $72,000 a day before everyone else on the network would give up mining.  Pittance to prevent an upstart currency from toppling the dollar.

An organization looking to make a long term profit could invest around 50 million in FPGA's, and while not completely controlling the market, eventually incentivize all comers into buying whatever system they developed.

Another point to make is that the difficulty will automatically double in 2013.  We will see if GPU power per price increases, but I doubt it.  Bitcoins are trivial in comparison to the video gaming markets as a whole.  My guess is that this will be the tipping point for GPU's assuming prices don't change drastically.  FPGA's can almost always be justified as a positive longterm investment, while GPU's will likely lose their value in a year or so.