Fpga bitcoin miner

Product Comparison

Last Updated: 2012-01-25. For the latest product information, please check the forum.

Bitcoin in a Nutshell

Bitcoin is a peer-to-peer currency based on public key signatures and an ingenious double-spend-prevention mechanism based on a cryptographic hash function (SHA256). Nodes in the Bitcoin network are rewarded for finding x, such that SHA256(SHA256(x)) is below a certain target (this is called mining). The first node to find a solution earns Bitcoins.

The Bitcoin network adjusts this target so that on average, a solution is found every 10 minutes, based on the total computational rate (MH/s) of the network.

A node's chance of winning is roughly equal to the fraction of total MH/s that it contributes to the network.

Mining requires the latest technology and/or low electricity costs to remain profitable over time. Only the most power-efficient survive.

In mid-2011, GPU mining was highly profitable, with many ATI GPUs paying for themselves in 6-9 months, counting power (but not cooling, maintenance, configuration, or other) costs. However, as an increasing number of GPU miners chased those returns, the difficulty adjusted and large-scale GPU mining inevitably became unprofitable.

Field Programmable Gate Array miners are under active development and use around 1/10th the power of a GPU, at the same performance level (MH/s). The upfront capital investment is higher than for GPUs; however, once made, the lower operational costs should enable FPGA miners to earn a significant monthly profit, even at price and difficulty levels where GPUs cannot. FPGAs are more power-efficient for mining, and over time are likely to make GPU mining unprofitable. Once the electricity to run a GPU costs more than the bitcoins it can produce, most miners will turn them off, and either sell their GPUs, or repurpose them.

A GPU process shrink does not change this situation; a 45nm FPGA still uses less electricity per MH than a 28nm GPU. And of course - 28nm FPGAs will arrive around the same time as 28nm GPUs, keeping FPGAs firmly ahead throughout 2012.

Now is the best time to switch to FPGA-based mining, to enjoy a major technological advantage over most other miners.

The FPGA End-Game

Large-scale miners who are willing to make the initial investment in FPGA mining should be able to mine a substantial portion of the remaining coins, with no operational threat from GPUs for at least the next 2 years. It looks likely that miners using GPUs will need to migrate to FPGAs to remain profitable during 2012.

Why not build an ASIC?

It's extremely expensive to build an ASIC that can outperform high-volume 40nm/28nm FPGAs. A 40nm mask set costs around $4M. Older technologies are less expensive, but above 130nm, probably cannot outperform a 28nm FPGA. At the current price and liquidity of Bitcoin, an ASIC would be an extremely risky investment. Most likely, the mining community will switch to FPGAs first and move to ASICs if/only if the Bitcoin ecosystem matures and expands considerably.

To Learn More..

Join #bitcoin-fpga on freenode. There's a small community of talented folks working on a variety of designs.

Hack A Day - FPGA bitcoin mining

The board requires only 6.8 watts for 100 Mhashes/second, but [li_gangyi]‘s blog says the team expects to hit 150-200 Mhashes with some improvements.

Only four of these boards were built and the supply has already sold out.  Deposits are being accepted at Cablesaurus towards pre-orders for the second generation model.  The second generation units are priced at $420 (single FGPA) to $620 (dual FPGA) but those prices will likely be lowered before purchase due to volume discounts.  The number of boards produced will be determined by how many deposits (paid in either bitcoins or USDs) are made.

This board differs from the modular FPGA hardware project  but runs the same open source FGPA miner  that was released in May.

FPGA hardware is more expensive for mining Bitcoin than the hashing equivalent when GPU graphics cards are used but power consumption for FPGA mining can be nearly an order of magnitude less.  At current exchange rates and difficulty levels. the dual-FGPA board will produce just under 0.12 BTC per day, which is worth about $1.22 USD.  Calculated using the typical U.S. residential rate the cost of electricity to run the two FPGAs for a day is under $0.04 USD, or about 3% of revenue.  For comparison, when GPU graphics cards are used for mining in regions where electric rates are high the cost of electricity can exceed half the miner’s revenue.

Though mining profitabilty  is near all-time lows, these levels are still high enough that FGPAs are not yet price competitive due to the higher hardware costs involved.  At the same time, this board just brought FPGA mining one step closer to becoming a significant competitor to GPU mining.  Those likely to be the early adopters will be those hitting total power consumption limits, those running out of space and those unable to sufficiently remove the heat produced when mining with GPUs.

Just released was the Open Source FPGA Bitcoin Miner  software.  This miner allows bitcoins to be mined using a commercially available FPGA board.

FPGA boards consume much less electricity compared to GPUs for the hashing work performed when mining bitcoins.

FGPAs might have a MHash/J performance level around 20 or more whereas the most efficient GPUs are closer to about 2 MHash/J.

There are fears that FPGA mining will force miners using GPUs today to either pivot and switch to FPGA mining for their operations or to abandon mining completely when the efficiency disparity makes GPU mining uncompetitive.

Those fears are likely unwarranted just yet though.  The cost of the equipment and not the cost of electricity continues to be the primary concern affecting miner’s decision on whether or not to add capacity.

Using commercially available FPGA boards to attain a certain level of hashing can require investment several times that required for a GPU-based equivalent.

Though a less costly FPGA board designed specifically for bitcoin could be built, there are no sources for such a product yet — at least not commercially.

GPU Mining Still Profitable And Still Growing

It will likely be many months yet before the FPGA would even start to displace the GPU as the technology responsible for significant increases to the mining difficulty measure.

GPU miners have been able to reach the breakeven point, on a per-dollar invested basis, within a couple months after bringing new capacity online.  Thereafter a decent profit continues to be returned though the profitability level can decline steadily with each subsequent increase in difficulty when there is no matching increase in the bitcoin exchange rate as well.

The Bitcoin network has continued to see capacity come online at an unprecedented and torrid pace — over 30 GHash/s of capacity was added each day during the most recent difficulty adjustment period, for example.

Because of the recent run-up in the bitcoin exchange rate mining is generally considered to be very profitable at the present time.  It will remain highly profitable until either the exchange rate drops or the difficulty level increases significantly as the result of additional mining capacity being procured and brought online by miners seeking those profits.

With so many miners active now, the global supply of GPU hardware has become constrained.  Even used GPUs found on eBay, for example, are becoming difficult to find.  If the now-discontinued ATI HD 5870s and HD 5970s were still sold, or AMD were able to keep up with demand for the AMD HD 6990s, the difficulty increases would likely be even greater than what is being seen now.

FPGA’s Edge

Even though the FGPA solution is extremely expensive relative to its GPU equivalent, FPGAs are widely available commercially.  If the bitcoin exchange rate were to spike further, FPGA mining may be about the only option for being able to add any significant amount of mining capacity.

The higher power efficiency that an FPGA provides is not just a factor for comparing the cost of electricity to mine but also when other factors are considered.

The total amount of electricity consumed by a miner with a GPU mining rig or two will add up quickly.  Individual miners are already popping breakers after exceeding limits on the amount of current available from household circuits used when mining with multiple GPUs in a single rig even.  Those with multiple rigs are finding limits to the amount of electricity fed to the property as well.

When power consumption limits are a constraint, then the FPGA with its significantly higher power efficiency might be the only method to increase mining hashing capacity.

Residential electric rates are often tiered to encourage conservation.  As a result the incremental rate for mining starts at the most expensive tier for the household.  Individuals mining from their homes should be the biggest beneficiaries from having an efficient FPGA alternative but the equipment’s high cost will remain a major limiting factor, as will the technical skill level necessary for FPGA mining.

Further out, if the exchange value of bitcoin rises much further the numbers begin to approach the level where building and producing an ASIC design might become feasible.

If or when that happens, then the clock counting the days left for GPU mining can start ticking.