Topic: Let's recap on what we've seen in the past few months (Read 12974 times)

... and finally, we can now end the old heated debate

http://bitcoinmagazine.com/working-avalon-asic-confirmed/

Although since price is at $20 I guess we'll never see any posts from Synaptic again - true to his word:

Last Active:    September 12, 2011, 01:36:26 AM

As always one has to do the maths on these FPGA boards which are not cheap.
832mh/s for 600 bucks.
Assume diff stays the same.
Assume price stays the same at 5.29.
Assume 10 cents kW/h and this small board+fan uses 80 watts (butterfly single).
After 10 months of non stop flawless running you will finally be ahead by 25 bucks.

But... diff will go up. Rest assured. 50 for a block will get cut in half before you break
even extending it out to what i will say is a year.

Now with that said if I was just getting into bitcoin and wanted to seriously mine at home
I would probably buy one. It is much better then setting up a traditional rig. But this assumes
you already have a gaming rig or do not want one. If you play games and have an older computer
I would probably just go with AMD/ATI and upgrade my home PC to mine.

Then there are people like me who have free electricity. I am happy running 5830s, 5850, 6950(flashed to 70s). etc...
They represent great value when you buy them used. 630mh/s costs around 150 dollars if you patiently troll ebay.
As for what to plug them into.. pfft.. people are at the point they are tossing MB with two pcie slots in them. Cheap
money. The power supply is the key of course.

In my mind.. the people buying these would just be spending their money on video cards anyway. So diff will go up
just the same with or without them.

As for the last thread.. dunno. Some AMD place is testing a few hundred cards at once and decided to make money at it ;-)

(Yes I'm sure I want to reply to a thread that has had no posts for at least 120 days)

So, let's recap on what we've seen in the past few months - relevant to the discussion that took place in this thread half a year ago. Since the early FPGA boards were already available then, but didn't convince the GPU crowd that the path CPU->GPU->FPGA->(s)ASIC was clear and inevitable, the new data would be:

Butterfly Labs - Single and Rig Box (FPGA, of which the former is apparently selling like hotcakes)

https://bitcointalksearch.org/topic/butterfly-labs-bitforce-single-and-mini-rig-box-60586

LargeCoin (sASIC, although not in customer hands yet)

https://bitcointalksearch.org/topic/largecoin-pricing-announced-taking-pre-orders-67505

... and of course the tantalizing (or worrying) prospect of someone throwing a lot of ASIC (?) power online without sharing

https://bitcointalksearch.org/topic/wonder-who-this-solominer-is-8862169-67634

I agree.

I also agree on your further arguing that diff 800,000 / price $3 is not low enough.

The ROI on this is certainly impressive, but I don't really get how this can be used as part of a business model unless the product were offered on the market or the person running the business knows how to make this mod.

I'm still not seeing your numbers though. If we use the current going rates of $210 for the main board and then $200 for each chip and each chip is at 90Mhps, then a 6 chip board is $1410 and does 540 MHps. The 1.8 Ghps board would be 20 chips, which would be $4210. If you got the chips for your retail price, it would be $3160. I think you skipped over something that shows it's possible today at today's pricing. Maybe you were using 150Mhps per chip (the 300Mhps rumor), but that's still $1896 for the chipset from Digikey, so I still don't get it.

And yet, you can buy a 1 chip board for around $410, a two chip board for $610.  Retail.

Assuming digikey is a viable retail operation you can guess volume production costs using wholesale pricing could make the final product viable at retail cost of the parts.

I won't deny this is not reality today.  Reality today is 180 mhash for $610.  Possibilities are there for much, much more however.  Within a time frame much smaller than "bitcoin at $100."

As a rule of thumb, figure a profitable retail price on electronics as 4x the parts cost.  2x in real volume, like a cell phone.

What's your time worth?

One other thing to put into your calculations.

The current board seats two FPGAs which cost $158 retail from Digikey.  The power circuitry, and, really, the rest of the board should be manly enough to seat six or eight of those.  Basically we're a small rework away from a $1400 board that could churn out as much as 1.8 gigahash (but today would yield around 1.4).

That's down there with 3x5830 rigs on a price/mhash, but would use 1/10th the power.  Also, with 6 chips per board it'll be easier to hit quantity discounts (lop another 200 off).  $1k for a 1.8 ghash board, anyone?

This could be done TODAY, with TODAY's pricing.

I know the argument, and the argument is speculation. I don't think it's something to build a business model after. If difficulty does come down or price does go up or cost/Mhps does go down, then it can change the equation for these things, but making assumptions about what might happen is just gambling.

And for the record, if the difficulty drops to 800,000 and the price to $3, then the X6500 has an 18 month (really 21 month) ROI. There is nothing stable enough about BitCoin that I would enter into it with an 21 month ROI on dedicated hardware. When I bought my rig, the ROI was just under 3 months. If the price goes to $2, the ROI is 28 month (41 months). This is using the 300 Mhps number which isn't guaranteed. There is nothing about low prices that is good for FPGA.

The argument is, that at really low prices (< $1), the difficulty (also part of the roi calculation) should drop significantly. It all depends on wether or not it really would drop at such low prices. I still hope we wont find out.

300Mh/s puts the ROI at 26 months, which is really 37 months given the reward decrease. In many cases a 37 month ROI would be great, but there is a severe amount of risk here. That is the ROI today at today's prices and today's difficulty. It's doubtful this will hold up for even 15 months let alone 37.

And my main point is that as the btc price decreases, this ROI decreases too. People are saying that low prices are where FPGA shines, but it really just gets worse.

I know some people here don't give a damn what ArtForz believes. Well, take this as his opinion then:

(this was on 2011/9/16 on #namecoin)

So the parameters of the calculations might change in the future.

I really disagree with the way Synaptic has presented himself in this debate, but I do agree with his basic premise. There are some real problems with FPGA as it stands right now. If things improve with respect to $/Mhps, then the equation changes, but I don't see how it's really viable. In fact, without a rise in btc price (or a drop in difficulty), I don't see how buying any hardware now is viable. Here is my math and reasoning (since Synaptic doesn't seem capable of producing any).

Right now the X6500 FPGA costs 610 and will deliver 180 Mhps. I have a spreadsheet I use to calculate out ROI and at least for me the ROI puts it at around 40 months. The problem is in about 15 months the block reward will be cut from 50 to 25. This actually makes the ROI 25 months longer. At the time of recast, there will actually be 50 months left. But wait, in that time, there is another block reward reduction. So assuming a constant difficulty and price, it doesn't seem that the current FPGAs can pay themselves off. People say that at lower btc prices FPGA make more sense but when I lower my btc price to $4, the FPGA ROI goes to 55 months. At $3, it's 76 months. This doesn't include the reward decreases.

Another factor here is that the hardware is all continuously improving. What do you think GPUs will look like in 40 months? What about other FPGAs? There's really no easy way to trade up FPGA hardware, but you can pretty easily trade up GPUs (sell on ebay, buy new GPU). When you buy a current FPGA, you lock into a specific amount of MHps. You could sell your FPGA to other miners, I suppose, but my feeling here is that this will more strictly follow performance relative to the latest chips because there isn't any ancillary use (gaming) for these boards. Because the hardware is used as well, you will need to discount it and it's possible that this discount will just take up any profit margin you had on that board. This is all theoretical, but it seems likely to me.

At these prices, GPU mining isn't much better. My mining rig is running at 60 months ROI. But then again I've been mining for the last two months and have been selling at some good prices, so when I take out what I've profited, it goes into the 30s. If I take out what I can sell my cards for on ebay today, it goes to under 10 months. So it still makes some sense for me to mine. But I can't recommend it to someone else unless they want to add a card to their machine for gaming and will mine while idle.

I don't think that people are being realistic about the price drop and the effect it has on long term profitability. Maybe people will begin to understand and this will result in a pull back in difficulty, but so far this hasn't been evident.

I can't take credit for the image. While I think the point is invalid (logs are more descriptive), the joke is funny.

LOL.

nice.

I take it you don't like logarithms.

http://www.reddit.com/r/Bitcoin/comments/k81zd/why_i_love_the_bitcoin_forums/

Let's recap shall we?

Winter saw lows of $0.07 and highs of $1.00

Spring saw lows of $0.50 and highs of $9.00

Summer saw lows of $8.00 and highs of $32.00

Autumn sees lows of $6.15 and expected highs between $25 and $100

Well, these guys say they are developing an ASIC miner. According to this it will be ready by middle 2012

This reminds me of the debate a few months back when they said mining off a sound card was possible. Interesting read.