This proposal is a sizable document, it you'd like a copy in MS Word for easier reading, please email usNovello Technologies Limited would like to present our project proposal for a family of Bitcoin mining rigs. We hope you’ll be excited and will want to contribute towards its success and help guarantee the future of Independent Mining. We’re assuming you already know a fair bit about Bitcoin Mining, but if you don’t we’d suggest that you do some research before reading our proposal.
Our project is live on Indiegogo, please take a look:
https://www.indiegogo.com/projects/ultra-low-cost-bitcoin-mining-systems--2There's also an article about the project on Cryptocoinsnews.com:
http://www.cryptocoinsnews.com/news/novello-technologies-begins-funding-ultra-low-cost-bitcoin-mining-rigs/2014/06/04First things first though – here’s what you really need to know about the project:
The Headlines
- Systems priced from $59 (128GH/s) to $2799 (8 TH/sec)
- Systems financial metrics:
$0.3 to $0.5/(GH/sec)- Power consumption averages 0.30 Joules/Gigahash, at the wall socket.
- Full custom ASIC chip built in 40nm LP process
- Inbuilt Raspberry Pi with WiFi except in basic $59 system – it needs a host laptop/PC or NOVA - HH
- All systems come with fully regulated, high efficiency (> 87%) power supplies with Power Factor correction
- This is a pre-order project, there are staged payment options
- Deliveries will start Dec 2014
- Innovative programs to keep early customers competitive no matter how high difficulty becomes
- Be first to buy on Indiegogo and get the BEST deals!
The TimescalesThe Gantt chart below shows the projects’ critical timescales; with target funding, tape out will take place mid to end August 2014, and prototype devices will be available mid October.
We'll make it clear from the start that we want to be a high volume supplier, but want most of that volume to come from our initial customers in the form of repeat purchases. In return, they will get the best possible deals we can offer and unique access to development programs.
About Novello
Let’s tell you a bit about our company. The driving force behind Novello is a group of professional engineers based in the UK with over 120 years combined experience in the design and manufacture of complex, highly reliable electronic systems and devices. We think profitable Bitcoin mining should be accessible to anyone no matter what their budget; by ‘profitable’ we mean a device that will pay back its’ purchase price in less than 3 -4 months, even with a very large rate of growth in network processing power. More accessibility means more users, this will help strengthen the Bitcoin network and promote wider use of Bitcoins.
To save the forum trolls, scam hunters and naysayers a lot of time and effort, here's the company structure at present:
Novello Technologies is a UK registered Limited Company, incorporated in November 2013. Our company registration number is SC464008. We are also registered for VAT, registration number is UK 186 6980 46. A UK Limited (or Ltd) company is similar in structure to a US Corporation, and is tightly regulated by UK law. Such a company can be formed with only one Director, the person who has overall responsibility for making sure the company complies with statutory regulationsy. In our case, the Director is Olivia Walker who is a marketing professional. My name is John Fowler, I am the company’s Technical Manager and primary contact point for this forum.
Currently Novello is a ‘structure’
ONLY – we rent an office in Glasgow, Scotland, this is our formal business address. We are a genuine start up, so have no factory or design office or indeed working prototypes –
you can’t prototype a custom chip - that’s why we are trying to raise funding. We do have is a SHA256 system design with ultra competitive technical and financial performance; our engineers have completed over 20 custom chip designs during their careers including ones for very high volume consumer electronics projects.
Our systems will be built using a future-proofed modular approach, some have the capability to add more hashing capacity, all can be easily upgraded, preserving much of their value. With our group experience in design, manufacturing, logistics and quality we know we can make truly consumer-grade products with unrivalled performance, cost effectiveness, reliability and after sales support within realistic timescales and budgets.
Thanks for reading this far but you’ve still got a lot of information to read and digest. So get yourself a cup of tea and settle down for what we hope will be an interesting read. There’s a comprehensive set of FAQ’s at the end, including some tough questions vendors usually don’t want to answer!
CONTENTS
1. Project Summary
1.1 Project Brief
1.2 Our Mission
1.3 Our Bitcoin Technology Development Strategy
1.4 Our System Lineup
2. Project Financing
2.1 Financial requirements & where the money goes
2.2 How Split Financing works
2.3 Delivery times and Courier charges
3. The Shape of Things to Come
3.1 Growth of Network Hashrate and Difficulty – what will happen?
3.2 The rise of Mining Companies and how to remain competitive
3.3 Power Play
3.4 ROI predictions
3.5 Guaranteed Supply Program (GSP)
4. Our Hardware Solution
4.1 Overview of Modular Methodology and Retaining Value
4.2 Full Custom vs Gate & Standard Cell
4.3 Small is Beautiful (sometimes)
4.4 Module Mania
4.5 Our Mining Chip
4.6 Heat kills
5. Pricing
6. FAQ’s
1.0 Project Summary
1.1 Project BriefOur project depends upon development of a Full Custom asic device in 40nm LP process technology which will be used in all our systems. We need to secure funding of at least $4.2 million to make our project viable, i.e. to complete design, tooling, product engineering and support and to supply customers with their systems. In return we're offering mining rigs with ultra competitive cost and power efficiency.Target funding will enable us to ship all initial system orders by the end of March 2015. If we are substantially oversubscribed, ie over $12 million, we will implement the design in a 28 nm process to give our customers even better efficiency and hashing power. We will also then introduce an additional program to help keep our customers competitive against organised mining companies, especially those who live in regions with high energy costs.
1.2 Our Mission
Mission Statement? Straight messages,
no corporate rubbish:
- To use our expertise to make Bitcoin mining much more accessible to those on a low budget, and to level the playing field against organised
mining companies. Diversity is good for Bitcoin, centralisation of power most definitely isn’t, don’t let anyone tell you anything different.
- To create a core group of customers that will come back to us again and again for their Bitcoin mining requirements and use innovative
programs to help them achieve their financial goals, no matter what happens to Bitcoin value and network difficulty.
- To fund independent academic research to enable continuous development of our mining device technology which will keep our customers
competitive no matter how high difficulty becomes.
- To develop our company into a successful full service silicon design bureau which will offer small business startups a low cost gateway to
prototype chips, typically costing less than US$50k in total.
1.3 Our Bitcoin Technology Development Strategy
It may surprise some we've chosen 40nm: we'd argue that our full custom design can outperform cell based 28nm designs, but it’s about keeping costs down. The core design uses techniques already in production in other devices and has been comprehensively simulated. Our research has shown that it could be enhanced to go significantly faster at lower levels of power consumption, but first we need to gather some very specific data from actual silicon and our first chip will include the necessary test structures. This data will be used to modify the existing design and build a second faster, lower power device in 40nm (or 28nm). We anticipate that it will be available in production quantities by Q2 2015. After that will come a direct die shrink from 40nm to 28nm, giving a further boost to efficiency by means of yet lower power consumption.
As we mentioned previously, substantial overfunding will allow us to implement the core asic in 28nm – then implement the design changes for a second, lower power device once we have actual data from test circuits.
Modularity is used to give our customers have a cost effective means to add hashing power to their rigs either through addition of hashing modules or replacement of existing ones by newer models.
1.4 Our System LineupOur mining systems are targeted towards best possible performance per dollar. We recognise that few can afford to spend thousands of dollars to gain entry to the Bitcoin mining club, so our flagship product isn’t a huge multi terrahash/sec miner (although we’ve got one of those too), it’s a little USB device costing $59 (or $49 if you order quick on Indiegogo) we call the NOVA–S that can be used with its’ supplied AC adapter and a host computer, preferably a laptop, to mine at 128 GigaHash/sec. And it’s very well priced:
$0.5 per Gigahash/second! Sexy, or what?
Next up is the NOVA - HH which builds on the NOVA –S concept. It isn’t a miner as such, it’s basically a 16 port USB hub which can accommodate up to 16 NOVA-S devices. It’s USB ports are very high power – up to 35 watts each and it includes a built in wifi enabled Raspberry Pi microcomputer which runs the mining software and acts as a bridge to the network, so no host system is needed. It also has it's own 800W power supply. You will need a TV or monitor, keyboard and mouse to set it up initially; it’s an easy starter route for those who can’t afford to spend a lot of money in one go to buy one of the bigger rigs but who can easily add capacity as they earn it!
Our system rigs offer hashing power from 1TH/sec up to 8 TH/sec and are based on a common enclosure platform with a form factor of 210W x 300L x 175H (4U), 2 can fit side by side in a standard 19” rack tray, 10 trays in a rack. Pricing runs from $0.35 to $0.50 per GH/sec and if you order early you can get them even cheaper! All system rigs are all modular, so it’s very easy to add or change hashing modules whilst keeping your existing infrastructure. All our systems except the NOVA-S have integrated wifi enabled Raspberry Pi microcomputers, so do not need host systems to mine. If you look at our pricing structure and factor in the delivery date as December - March 2014, it is clear that we are offering our customers – especially those willing to commit early – a very considerable performance advantage in terms of how quickly they might achieve profitability on their new rig.
2.0 Project Financing
Now for the difficult bit. We’re a genuine start-up company and so we need external financing to build our company and make our project work. In particular we need to fund the huge NRE and development costs associated with the core asic device; this won’t come as news to anyone in the Bitcoin mining community and we realise that any new company entering the market will be viewed with a lot of suspicion, especially in a pre-order situation. We also recognise that our project is going to attract a huge amount of negative comments on the Bitcoin forums simply because it doesn’t follow the conventional approach, and because some will struggle to understand how we can make something that’s so powerful yet so inexpensive.
There’s also a hard core of individuals and companies who really don’t want every wanabee miner with $59 joining their party, making the difficulty higher and reducing their earnings. They will try extremely hard to convince you that pre-orders are ‘bad’ and under no circumstances should you go down this route - despite the fact that most of them made their money from pre-order products! We’d advise you to read, read, then read some more and make up your own mind.
Nowadays pre-order based projects come under a lot of scrutiny largely due to the number of programs that went wrong and kept customers waiting months for the systems they paid for in good faith. That’s all fine and well but the large amounts of money needed to develop new asic products make it all but impossible for individuals to fund. Funding can be sought from Venture Capital (VC) companies, but they expect a very large return on their investment, at least a factor of 2 – 3 times plus their original investment back within a period of 2 – 3 years, but for mining projects most recognise that they’ll get a much faster and higher return by using the rigs to mine, not to sell to the public.
For companies that are VC funded, they need larger profit margins to pay the VC companies back and that means higher rig price. We’re looking to use a slightly different financing model to the normal one, but we'll now explain why we need $4.2 million, and how we’ll raise it.
2.1 Financial Requirements & where the Money goes$4.2 million is our initial target sales value, in rough terms. We are happy to tell you that our GPM (Gross Profit Margin) on sales averages out at about 38 - 40% , meaning that for every $1000 in sales we receive, we make around $390 gross; this is very much lower than our competitors who have margins of well over 75%. Out of that we have to pay for premises, people and all the usual stuff most companies need,and also the huge NRE costs for the custom mining chip. $4.2 million sales gives us approximately $1.75 million to cover all these costs, so we’re not planning on buying any fancy cars or boats – just 40nm masks, chips and components!
If we exceed our sales target then any extra profit generated will be used to expedite delivery as much as is feasible. If we’re very substantially overfunded – to over $11 million - the extra profit will be used to implement our design in 28nm and cover the extra costs for the chips; you’re right in thinking that $11m will bring us an extra $2.8m in profit, but then our NRE costs an extra $0.8 million and the extra cost for the 28nm asic chips? Nearly $2m! On the flip side, we’ll save about $200k on component costs, so out net increase on profits would be approximately $700k and we will use it to expedite delivery of the silicon and our customers will get machines with considerably higher performance!So where does the money go? The chart below shows the breakdown based on target funding of $4.2m:
Our campaign on Indiegogo is to raise our core finance, it helps the project get publicity and will hopefully result in more sales; the prices you see on Indiegogo are the
regular price, if you buy early you can actually get them even cheaper with the Fortune 500 ‘perks’. Indiegogo offers payment by debit or credit card which we can’t at the moment, but will do in the near future, the full system price is paid up front the time of order.
Web / Forum based pricing offers two different price structures. Why? Well, we know that many prospective buyers will be very risk averse and so will be reluctant to buy from an unknown company with no track record. But then our prices are around half to a third of the nearest competition in $/GH terms; even so, for those who don’t want to go through Indiegogo -
they don’t accept Bitcoin or bank transfer payments - we’re offering an alternative, Split Financing.
2.2 How Split Financing works If you want to buy outside Indiegogo or pay by Bitcoin/Bank Transfer, you have two choices:
1. Pay regular price in full, if you’re early enough you can opt to participate in special programs
2. Pay approximately 10% over the regular price, but pay only 33% of this amount at time of order. The balance will be paid once your system
has been built and is ready for delivery.
Split Payment allows you to buy the system you want
now even if you don’t have the full amount, giving you time to save up the balance. The flip side is that you’ll pay more for the system as you’re taking less of the development burden and you won’t be eligible for our Guaranteed Supply or other special programs. Example: for our 2 TH product it’s $899 up front, buy it on the web/forum and you can pay $899 up front or $999, approximately 10% higher, with only $333 up front and the rest once your system is ready to be delivered. With both options you can pay with Bitcoins or Ban transfer. Note that with our project you don’t need several thousand dollars to get TH performance, just $499 buys you 1TH of hashing power, and you don’t need to buy a separate power supply!
2.3 Delivery Times and Courier Charges
Deliveries of systems will commence end of December 2014. Orders will be shipped on a first in, first out basis. If we sell all our initial capacity it will take us until the end of March 2015 to complete shipment of all orders. Courier delivery will be payable by the customer and won’t be charged until their system is ready to ship. We’ll only use internationally recognised insured couriers such as Fedex or UPS with ‘signed for’ services. Below you see sample rates for various countries. For EU buyers, VAT will be charge at 20% on the final system plus shipping cost
unless they can prove they are an EU VAT registered entity.
We do have a slight problem with the price for shipping the NOVA-S products internationally, particularly to the US & Canada, but we're working on solutions for that and it looks like we will have one for North America - we'll keep you posted on progress.
3.0 The shape of things to come: Difficulty/Network HashRate
The Bitcoin mining landscape has changed considerably in the past year, and continues to evolve. Let’s explore what we think that will mean to miners.
3.1 Growth of Network Hashrate and Difficulty – what will happen?
These are the subjects that all Bitcoin Miners dread – the Network Hashing Rate and Difficulty, we’ll refer to these as the NTR – Network Terrahash Rate and DIF respectively; the two measures essentially track one another. You’ll never see mining rig vendors talking about these – it’s not good for sales and you really don’t want to say anything that might make your potential customers question the viability of their purchase, do you? Well, we’re a bit different; we want to tackle this subject head on to give our potential customers all the facts and figures they need – both good and bad - so they can make truly informed decisions about whether to commit.
Since mining ASICs appeared, the NTR and DIF have risen over 500 fold; it is now growing at over 4000 TH per week and there is an awful lot worse to come. We predict that nearly 85% of future NTR growth will be powered by organised mining companies with large hashing capacities; 4000 TH is the equivalent of adding 8,000 500 GH rigs a week ! This has drastic implications for any current or potential miners – how fast will the DIF rise, and will they ever be able to afford a rig powerful enough to pay for itself let alone achieve ROI? It’s a question made even more pertinent by the prospect of a technological arms race with mining companies competing with each other to gain share of the NTR – this can only be achieved by ever increasing capacity.
Individual miners will be the losers here – unless they can compete on equal terms, they can never make a profit. So it’s essential to have some idea of where the NTR/DIF might be for the next 18 - 24 months and that’s not an easy task as there are many variables that will affect it, including the Bitcoin/Dollar exchange rate and the pricing/availability of new mining rigs. Because it’s so important, we’ve spent a lot of time and effort building a software model to simulate the market and make estimates of what might happen. But they are estimates, and are obviously not guaranteed! There is also what we would refer to as braking factors that will slow growth if it’s too rapid and mining becomes uneconomical, but they take time to take effect. So, here’s our estimate showing what we think the NTR might be until the end of 2016:
3.2 The rise of Mining Companies and how to remain competitive
Mining companies are a relatively new addition to the Bitcoin ecosystem but make no mistake, they are here to stay. Bitcoin mining has gone from being an obscure digital pastime to a rapidly growing financial instrument, and to large investors presents unrivalled ROI’s compared to conventional investment vehicles; even if the Bitcoin exchange rate dips as low as $250, it’s as close to a ‘dead cert’ as you can get. The financiers behind these companies are quite happy to wait 9 – 12 or even 15 months until their equipment pays for itself, thereafter it’s mostly all profit and they can site their facilities almost anywhere to get their energy much, much cheaper than Joe Public. They can access huge amounts of capital to fund their rig purchases and secure large discounts from the rig suppliers or, better still, make their own.
What’s even worse is that some of the mining asic suppliers are actively encouraging them to build more and more capacity while at the same time making nonsensical claims that they want to ‘distribute the hashrate’! And why not? They don’t care about individual miners, just profit.
We care about it too, but care far more about having a loyal customer base and a long term sustainable business.
The mining companies are a very real and present threat to individual miners who can never compete with their resources, and the cards are very much stacked in their favour – up until now. As mining companies continue to pile on the Petahashes, the questions every current or prospective miner wants an answer to are these:
Is there a simple way to calculate if my new rig is going to be profitable or not?
Can I compete against the mining companies? Well, we think we can answer both. It’s based on how much you should pay for your rig in dollars per Gigahash:
- on the date it’s delivered
- an exchange rate of 1 BTC = $500 (yes, we know it's over $600, but let's be a little pessimistc here)
- power usage of 0.85 /0.55 Joules/GH, or our worst case 0.3 Joules/GH,
at the AC socket- power cost of US$ 0.1125 per kilowatt hour (US Domestic average according to EIA
http://www.eia.gov/electricity/monthly/update /end_use.cfm#tabs_prices-3)
- assumes that you want your rig to break even in a six month period after delivery = if it can’t do this, then there’s a very good chance that it
may never be viable and so the companies deter yet another potential rider on their gravy train. Try it out:
And here's the same table for BTC = $450:
You can clearly see the effect of the increasing NTR and energy usage on profitability. If the exchange rate varies you will have to correct these figures by the appropriate ratio, and please remember that the Bitcoin’s value might go down as well as up!
You should also be aware that mining companies can currently build and deploy new capacity in their data centres
For the equivalent of around $1.35 / GH. This will fall to around $1.15 / GH by end 2014.
We are presenting this information not to scare people, but in the interests of full disclosure. Mining Bitcoins is very competitive and we don’t want our customers to imagine what ‘might be’ instead of accepting reality. But, it is possible to get an excellent return from Mining if you have the right tools and the right supplier– and that’s where we come in.
3.3 Power Play
Mining will soon reach a point where the cost of power takes away more than 50% of a miners earnings. If you’re an ‘average’ US customer with a 0.85J/GH rig paying the average rate for your electricity then this will happen when the NTR hits about 320,000. If you manage to get your hands on a 0.55J/GH rig then the 50/50 point is delayed until 540,000 . With a 0.3J/GH rig it’s nearer 1,050,0000! We’ve designed our systems to consume as little power as possible whilst remaining competitive at hashing; the benefit of using full custom design is that you only use what you need – no wasted resources. Below are our estimates for MAXIMUM power consumption for our systems with all devices running at rated speed, measured at the AC wall socket so including all power supply switching losses and auxiliary circuit power:
You can see these are very competitive figures, despite us using 40nm technology. And they’ll get even better with our generation 2 40nm chip – some 36% less, we estimate.
3.4 Return on Investment (ROI) Predictions
Have you ever seen a mining rig supplier exploring this subject in their sale pitch? No? It's very simple – too much information given to buyers might put them off purchasing, rig manufacturers know that most Bitcoin miners dream of making millions of dollars from their mining activities, and it’s much better to let them hold onto that dream; a dose of reality might stop them buying. Well, we’re engineers and don’t believe in fantasy, just facts.
Fact 1: Bitcoin mining isn’t an investment – it’s a gamble, albeit one with semi predictable odds
Fact 2: The days of easy pickings for miners are well and truly over and will never return
Fact 3: Even an 8 TH rig isn’t going to make you a fortune, unless there’s a dramatic rise in BTC value
Fact 4: You can get a very reasonable return from mining – much better than any conventional investment
vehicle – if you have the right information, right equipment and the right supplier.
Fact 5: If you don’t have a rig that competes with the economics of the mining companies, in the long term
you will never achieve any kind of return on your rig purchaseSo now that we’ve made you thoroughly depressed, let’s look at what you might be able to achieve:
From Jan 2014 to Jan 2015 (365 days), based on our NTR / DIF prediction and 1BTC = $500, we estimate a NOVA-1 1 TH/sec rig, taking into account costs of electricity and mining pool fees could earn you:
$1322 based on average current US residential power costs (estimated at 12cents per kWh in 2015) or
$1037 based on corresponding UK figures (estimated at 23 cents per kWh in 2015)
On a purchase price of
$499 and shipping/tax charges of $95, then a US rig will provide net earnings of ($1322 - $499 - $95) = $728, or
ROI (Return On Investment) of 122% (728/(499+95)) less taxes you may have to pay on your income but you still have the value of the hardware at the end of the year and it will continue to earn money. We don’t have detailed figures up to the end of 2016, but predict the net earnings would be
$542 for the US,
$247 for the UK, assuming power costs rise by 5%.
At the other end of the scale, a US based NOVA-8 systems’ 2015 figure is $10,576 deducting initial costs, shipping and taxes gives net earnings of ($10,576- $2799 - $153) =
$7,624 or ROI of 258%. For the UK the figures are $8,296 - $2799 - $600, ie
$4897 or ROI of 144%. Both figures are again less any payable personal or corporate taxes. Why the large difference? It’s all down mainly to those power costs
and UK VAT. Of course this analysis is only looking at payment over a long period of time, mining rewards are paid our regularly if you’re using a pool as most miners are, this then gives you the opportunity to buy more hashing power as you get more funds. In case you missed it, that was a sneaky introduction for our Guaranteed Supply Program or GSP, it’s another part of our product offering which will continue to give our customers a financial advantage once they have started mining with our products
3.5 Guaranteed Supply Program
OK, you’ve got your new mining rig and it’s happily earning its’ keep. You’ve got some Bitcoins (or cash), and would like more hashing power NOW. Trouble is, all the competitive products have waiting lists of months, sound familiar?
We are offering some of our early customers who pay in full (and so bear the most of the development burden), a unique program where we will guarantee them almost immediate delivery on up to an additional 3 times the original hashing power they purchased, PLUS discounted rates on the upgrade hardware! This applies only if they paid in full through Indiegogo or Web/Forums.There will be a 3 month period after their delivery date where this kicks in, after that they can order up to 3 times their original hashing power as hashing modules or full systems at the rate of one upgrade of 1 x original capacity per month up to one year from their original delivery date at specially discounted prices. Of course, they can order as much hashing power as they like at any time, but delivery will depend on order backlog and available capacity; they would still get a discount but not as much as they would get buying via GSP. This program is available to the first 1000 customers who meet the payment criterion and here’s an example of how the GSP might work in the real world:
The spreadsheet below shows a running mining earnings balance for Jane, who lives in Colorado. She’s one of the first to order a NOVA-1 system and it’s delivered on January 4 2015; as time goes on her rig earns Bitcoins which she converts to cash. When she has enough money, she uses her GSP status and buys 1TH upgrade modules at $249 each – she’s already paid for the electricity to run her rig costing 12 cents per kWh, the US average in 2015. She gets her modules within 10 days of ordering and soon she’s hashing with more power, making more money for the rest of the month.
This process repeats until Jane has used up her GSP credits, by Jan 4 2016 Jane’s
now 4 TH rig has generated her net
$2,722, and that’s including paying for upgrades and the electricity to run it. That’s not bad for a $499 initial purchase! Over the following 12 months it will make her a further net
$2,408, even allowing for energy bills going up by 5%. She’s not getting rich by any stretch of the imagination, but we imagine she’d be more than happy that she signed up early to buy her rig. Now, if Jane lived in California, the UK or other areas with high energy costs, she might be paying twice (or more) the US average electricity costs; this will have a significant effect on her earnings, but what can a miner situated in a high energy cost region do? Well, we may have a solution to that problem for our early customers, but it depends on to what level we get funded as it relies on a lot of investment in infrastructure.
Please read the next paragraph very carefully - at least twice!
In the case study shown, Jane manages to significantly increase her TH capacity and earnings because she earns Bitcoins a lot quicker than her competition and can access extra capacity as soon as she’s made enough money to pay for it, AND get it at unbeatable $/(GH/sec) rates. But she also doesn’t spend any of her earnings except on electricity or more hardware. However, this scenario is based solely on our prediction of Bitcoin value and NTR, and is not guaranteed or warranted in any way. We’d strongly advise prospective customers to do their own research and calculations and solicit opinion from a variety of sources.4.0 Our Hardware Solution
So thanks for reading this far, we know we’re throwing a lot of information at you but please bear with us, the next section gives a full description of what goes inside the mining rig, especially the custom chip, and also explains how the modular system works.
4.1 Overview of Modular Methodology and retaining value
Bitcoin mining is a very dynamic industry by nature, things keep changing and the ever rising difficulty means today’s top line mining rig is soon obsolete. It’s like what happened in the PC industry, except that an old Pentium 4 computer can still do very useful work, an old mining rig just consumes electricity! The PC industry taught us that a modular approach means systems can be upgraded by swapping parts such as DRAM or graphics cards. You can only do this so often, but a lot of the original system value can be preserved, yhat’s why we’ve chosen this route with our miners.
It’s a common misconception that most of the cost of a rig is in the asic chips. Sure, to engineer tool and build an asic takes millions but the actual parts don’t cost much, in our systems they account for about 48% of the overall cost. The rest of the system isn’t dependent on what’s inside the chips; careful design means it can be used and reused the same as a PC chassis. As difficulty increases miners need more power, either in the form of additions to their rig, or usually a new setup. We have structured our design so that it uses easily replaced modules and our overall design aim is to preserve the value of the infrastructure for as long as possible; as examples of this we usel over-rated PSU’s and DC-DC converters that can cope with a wide range of output voltage and current requirements, this helps keep our customers competitive. We’d like to think that other manufacturers might mimic our scheme and produce third party products for our infrastructure.
Now we’ll explore reasons why we can fit so much power into our asic chip:
4.2 Full Custom versus Gate and Standard Cell To make a silicon chip, there are several routes you can follow. For speed, a gate-based design is like digital Lego, it uses simple pre-designed blocks of transistors that can be joined together to quickly build your system. They have known, tested performance and have been designed to work in most likely applications. Used with EDA (Electronic Design Automation) software it makes the chip designers’ job a lot easier, especially when it comes to simulating performance and functionality, or so the theory goes. On the downside, gate based can be ‘top heavy’ in that they use far more or far bigger transistors in each gate than might be needed, and that means higher power and more silicon real estate. However, the design path is fast, predictable (usually) and rugged and a good solution for most small to medium sized designs.
Standard Cell. It’s broadly similar to gate based design except that larger pre-designed functions are available, such as counters and adders, the designer can use them in a similar ‘lego' fashion to build their circuit, but with less flexible rules. It makes more efficient use of silicon as the functions have been heavily optimised at the transistor level, but again they have to be generic enough to cover a wide range of applications. The design route is usually easier than gate based as very complex functions can be built more easily, although there may be issues with routing and timing. For the same function, standard cells take up much less silicon area than gates. Usually EDA tools can accommodate both gate and cell based designs. Standard cell is very good for large, complex designs.
Finally there’s Full Custom. This is how chip design started, manual designs with individual transistors. It’s still very much alive today. Why? Answer – efficiency; just like a custom kitchen or suit, you get exactly what you want and need, no waste or unused functions. If you have a part in your circuit that drives only 2 loads, why put in one double the size that can drive 8? Or use a register designed for 2 GHz that’s three times the size of one for 600MHz? Designing in only what you need can save many transistors so the design is more efficient. Trouble is, it takes a lot more manual involvement and there are few EDA tools to help, so it costs more. But take heart from the fact that big companies like Intel and AMD have hundreds of engineers doing full custom designs every day!
A SHA256 hashing engine is actually a simple design from an electronics point of view; each individual stage in a 64 stage pipeline is built up from very simple functions, gate and cell based designs aren’t always good at optimising this type of structure. With full custom, you design in only what you need and with experience you know when you can sneak in special compressed functions. Because of the regular nature of SHA256, simulation isn’t as hard as it might seem – concentrate on getting the stages right and the rest will naturally follow. One of our stages contains roughly 2300 logic gate equivalents, close to the theoretical minimum needed so it’s not really complex; for a mining chip full custom makes a huge amount of sense and the end result is that you get a chip that’s area, cost and power efficient - that’s what miners want and need. Iif you don’t believe that a full custom solution can be so efficient we’d strongly advise you to read the following textbooks:
“
Closing the Power Gap between ASIC and Custom; Tools and Techniques for Low Power Design” (Springer);
David Chinnery, Kurt Keuzer ISBN-13: 978-0387257631
“
Low Voltage CMOS VLSI Circuits” (Wiley-Blackwell); James B, Kuo, Jea-Hong Lau ISBN-13: 978-0471719793
4.3 Small is beautiful – sometimesWhen designing a mining asic, one of the most important questions is how to implement the design, do you have a large single chip that can do 1TH/sec, or 128 smaller ones at 8GH/sec? Larger chips produce more heat, smaller ones only produce a little but there’s more of them. A larger chip uses a smaller PCB than lots of smaller ones, and there’s fewer connections to the board to go faulty. But a larger chip has a much lower production yield than a much smaller one due the chances of encountering a defect in the silicon wafer used to build it, that means it’s more expensive per GH/sec.
What we’re pointing out here is that there’s a balancing act between functionality and cost. We’ve optimised our solution such that our chip comes in at about 100mm2. This is a manageable size, and still yields well, its’ packaged in a moulded flip chip ball grid array (a bit of a mouthful that so it’s more commonly called MFCBGA) to make the thermal design of our systems efficient. With a single chip rated at 128GH/sec, Terrahash capability can be easily obtained by using 8 - we actually use 4 chips in a module, this is covered in the next section.
4.4 Module ManiaNow we’ll explore our modules and how they fit into the system. All are easy to access, fit and replace and have been designed for reuse.
There's five main modules:
A
Hashing Module (HM) that contains 4 mining chips, a 32 bit microcontroller, USB interface and some support circuitry; very easy to install and/or replace. Self – testing without system controller intervention.
A
Micro Hashing Module (MHM) that is essentially a smaller HM with only 1 mining chip, it has the same 32 bit microcontroller, USB interface and support circuits. It’s used in the NOVA – S USB miner.
A
Controller Board (CB) that consists of multiple USB hubs, fan control circuitry and auxiliary power regulators. It’s also used in the NOVA-HH USB hub.
A
High Current Power Supply (HCPS) – 1 for each HM – that incorporates a 200 amp capable variable voltage regulator used to supply power to the HM chips. HCPS modules take in regulated 12V DC power supplied from an internal PSU – no need to buy a separate power supply. The HCPS ‘piggybacks’ on top of an HM, a low resistance plug and socket connects them together. The HCPS automatically senses the HM type and adjusts its’ output to suit, meaning it can be used with future HM’s, most likely requiring lower voltage but more current.
HM’s and HCPS’s are both based on a 60 x 96 mm pcb platform.
The NOVA – S USB miner has it’s own HCPS unit, a
UPCU – USB Power Control Unit, it’s integrated with a fan controller and is designed to take power via its supplied AC adapter. Note that full performance at 128GH/sec is only available when using the adapter due to the limitation of standard USB ports; using a standard USB 2.0 port without the adapter will limit performance to around 8-10GH/sec, USB 3.0 will allow approximately 20GH/sec.
A fifth module, a Raspberry Pi microcomputer with Wi-Fi capability is used as a ‘bridge’ between the system and the Bitcoin Network. It also allows system monitoring and configuration when connected to an external TV (or monitor), keyboard and mouse. The Pi runs a modified version of the excellent Minepeon mining software, which is easy to set up and configure.
4.5 Our Mining Chip and Fault Tolerance
Onto the good stuff now! Our ASIC has been designed to satisfy 3 key requirements:
- High reliability and Fault Tolerance
- Maximum Hashing performance per $
- Minimum power consumption
The device is built around an array of 256 dual-SHA256 pipelines (2 x 65 stages in each) which operate as a group. An additional 16 pipelines are also present, but are not mapped into the address space of the main array. Chip performance is 128 Gigahashes/second (GH/s) at rated clock speed of 500Mhz, rated core voltage of 0.85V, die core temperature of 75 degrees Celcius and power consumption of 25 Watts. It is possible to run at a lower core voltage of 0.6 - 0.7V, however until we have working silicon we cannot confirm how fast the device will run at these levels. Please note that power consumption figures in the paragraph above are for the chip only, and do not include power supply and auxiliary circuit losses.
Comprehensive test structures built into the fabric of the device allow for production testing and ‘manufacturability’ analysis. It also enables the HM’s 32 bit microcontroller to carry out automatic full function tests on all pipelines, including the spares, this means that full hashing power is always available as pipelines are constantly ‘rotated’ for testing. Testing takes only takes a few seconds and allows the controller to identify any that are faulty; if one is found, it is temporarily disabled and a ‘spare’ mapped into its address space. A failure may be a temporary fault due to a hotspot, but regular testing will identify if the fault is a genuine failure or intermittent. The pipeline can then be permanently disabled and ‘remembered’ by the system controller as faulty, with a spare taking its place. This all takes place without any user intervention, although test results and pipeline swaps are recorded and can be reviewed at any time.
We plan to build the ASIC in a 40nm LP medium Vt process, this is a very stable and reliable process, very cost effective and relatively low power - you’ve already seen that our systems give excellent Watt/GH performance. SHA256 is very power hungry, virtually every transistor in a pipeline switches every clock cycle so the key to performance is reducing the number of transistors that switch and minimising the power consumption of those that do. To achieve these objectives we have implemented the pipeline stages within the SHA256 engine at full custom level as we discussed earlier, it means we can save a huge number of transistors and control the size and hence power consumption of those that remain. We’re also making extensive use of analogue design techniques to minimise noise and reduce supply voltage – again saving power. Every transistor we save, shrink or optimise saves power, and overall this results in greater power efficiency.
REMEMBER :
Lower Power = Less Heat = Greater Reliability + More Efficiency = More $$$ earned
As a closing point on the design front, some readers might be a bit puzzled as to why we have put a 32 bit microcontroller on each hashing module. The answer is fairly simple – it takes away any high level control from the asic, leaving it as a ‘dumb’ calculator; it can distribute control and testing data to multiple chips and has inbuilt non-volatile memory that allows us to be able to log and remember any faulty pipelines and update firmware. Also, the automatic testing capability means that we can full test modules by themselves as a full unit, this saves a huge amount of production time and removes the need for special test equipment. It all saves costs, plus we’d rather have a programmable state machine off-chip in a microcontroller than embedded in a cast-in-stone asic!
4.6 Heat KillsIt’s all very well having high-tech chips that can mine fast, but they also produce a lot of heat and our mining
rig enclosures aren’t that huge. If the chips can’t get rid of the heat they generate quickly enough, their core temperature rockets and this destroys their transistors, so it’s essential to have good thermal management to make sure this doesn’t happen.
Rig suppliers have usually gone for classic fan/heatsink arrangement or water cooling. We looked at both these and other methods, and decided that for maximum reliability we’d use a fairly conservative route with ducted pin fin heatsinks and high capacity fans. This is a good combination and means we can get rid of sufficient heat to keep the chips well within their operating temperature range. It’s also very reliable – it’s easy for the fan controller to determine if a fan has stuck or fused and so close the system down until it’s fixed. We are using continuous duty fans with fluid or ball bearing with expected lifetimes (called MTBF’s – Mean Time Before Failure) of over 100,000 hours. Also, pin fin technology means the sinks are fairly compact, enabling us to pack a lot of hashing capacity into a compact enclosure.
A blown pin fin heatsink is also used for the NOVA-S USB miner with very small fans mounted on top of the heatsink, blowing down. This gives excellent cooling results, even in a hot room.
5. PRICING and PaymentOur order procedure is slightly more complex than usual due to the Indiegogo program running in parallel.
If you want to order a system direct from us, please email us the following:
- What it is you you want to order
- Your contact details
- Your shipping address
- How you wish to pay, ie in full or split payment
- What payment method you prefer, ie Bitcoin or bank transfer
We will send you back a confirmation email with payment instructions, and will confirm your order number to you when your payment is received.
If you purchase through this forum, then you may wish to be entered into our prize draw -
there are 6 NOVA-8 systems up for grabs - simply put your username onto the form you return to us and you'll be automatically entered! Draw will take place on 9th July, results will be posted in this forum. It's open to Bitcointalk members of any level!
Special Note on Pricing:
The pricing shown below applies to our first production run, order book for this run will close on 8 July 2014, regardless of number of systems sold. The first 1000 customers for each system type (excluding NOVA-S and NOVA-HH) will be eligible for the GSP program and will also qualify for special discounts on future system and upgrade module production run - this includes customers buying through Indiegogo.
Indiegogo ‘Fortune 500’ offers: If you don’t want to go through Indiegogo or want to pay by Bitcoins or Bank Transfer, you can still get access to the F500 offers, simply drop us an email stating so. We have set aside 250 of each system type at F500 (2500 NOVA-S F5000) pricing for this method.
6. FAQ’sQ. This all sounds far too good to be true. How can you possibly build systems at this price when others are charging 3 or more times as much for the same hashing power?A. We figure that this a question most people will ask. There are 3 separate parts to the answer;
Firstly, we have a lot of experience in the design and manufacture of consumer electronic systems, so we know how to design and build at the lowest possible cost. We’re using well-established methods to source competitive, competent suppliers and subcontractors who will manufacture the component parts of the systems so all we’re doing is final assembly and test.
Secondly, by using full custom 40nm technology we save over $1300k in development costs versus 28nm- that’s $130 a unit for 10k units in engineering charges alone. Device gate density doesn’t count in SHA256 – it’s the number of transistors switching per clock, their power consumption and power dissipation capability of the die that limits performance, so our full custom design approach means our 40nm device can easily outperform a gate or standard cell based 28nm design. Don’t believe us? Read the excellent books we mentioned above.
Lastly, and most importantly, our whole business ethos is built upon maintaining a level playing field for our customers in mining and that means we have to be a volume supplier. Volume means savings in parts cost, (aided by modularity) and lower cost to the customer. It also means we’re content to make a modest margin, not what we think the market will ‘bear’. If we can build and sell an 8TH system for $2799 - including its share of the development and tooling costs – and make a decent profit then so can every other supplier. Ask yourself why their prices are so high.
Q. Your company has no history of making or delivering mining systems and yet you want people to pay a lot of pre-delivery money for a product that will take over 6 months to be delivered – if it gets delivered at all. Why should I trust you with my money? A. We have to make one thing clear to any prospective customer - every company has to start somewhere. Novello is a start-up and like most start -ups we need to secure external finance to fund development and tooling costs. Each system that we sell contributes a portion of this, with most of the remainder paying for the actual components and labour needed to make the systems. We’ve tried to make our sales/funding proposal as flexible and transparent as possible but ultimately, if you don’t want to pre-order or don’t trust us that’s your choice, but you’ll miss out on a potentially very profitable product and the chance to work with a company that really has its customers interests at heart.
We do appreciate that many people have had unfortunate experiences with rig suppliers who have delivered late or sometimes not at all, and it’s understandable that any new company – especially one offering products so cheaply – will be viewed with extreme suspicion. Having said that, we are being completely honest about what we are doing, even to the extent that we may actually put people off buying a rig because of our predictions about the rise in difficult. We’re also not hiding the fact that we’re going against the trend and using custom 40nm for the reasons outlined elsewhere.
Everything in our proposal is based on solid engineering design and experience, experience that tells us that from day one that you design in reliability and manufacturability rather leaving it as an afterthought, and have a fully worked out system integration and product engineering plan that doesn’t rely on plug-it-in-and-hope-that-it-works.
Q. I already did a pre order and got burnt. Why don’t you just wait until you can supply unit from stock rather than taking pre orders?A. As we explained before, we have to finance the development and manufacturing costs. To be blunt, if we could afford to do this then we would make a lot more money simply by using our stock to mine, or auction it for premium prices. Our ‘bargain’ with our customers is that they pay a modest deposit or a discounted full payment in return for a very competitively priced product which will remain so for some time, even allowing for the development cycle. Bitcoin mining rigs are not a normal commercial product – everyone want the fastest system now, so stock holding would be nigh on impossible unless it was older, inefficient technology.
BLUNTLY: If it’s in stock, it’s obsolete or will be shortly!
Q. If you sell all your capacity then by your own plan you may have to deliver some 30,000+ systems in a roughly 13 week period from end December 14 to end March 15. Have you a strategy to achieve this – it sounds an awful lot in a very short time?A. It’s a very important and pertinent question. Yes, we have planned for this eventuality and as we said earlier we’re used to designing high volume systems and progressing them to manufacture. 3,000 systems a week is not particularly difficult to achieve if you know what you’re doing. The modular approach makes system assembly very straightforward and quick and allows for pre-testing of the actual hashing modules as an integrated unit in stand-alone mode. We’ll also use subcontractors for a lot of the ‘grunt’ or low level work, and if we got this level of funding – it would be around $21 million – then you can be assured that we’d sacrifice profit for extra staff and resources to ensure that customers get their systems on time.
Q. You mention Mining Companies a few times – why do you see them as such a problem?A. Bitcoin was always meant to be free of any centralised control, that means that everyone can participate, and this should apply to mining also. If you have too much power concentrated in the hands of a few individuals or organisations it can severely disadvantage others. As newer, more expensive technologies come on line, mining will end up being the sole preserve of the wealthy or dedicated companies, especially those who can afford to develop their own ASICs – and some have clearly already done so. Then what happens to ordinary miners? They can never get hold of hardware at the same $/GH figures the mining companies enjoy– unless they come to us.
Q. You’re a UK company, so why do you want to be paid in US dollars?A. Over 90% of our costs are in dollars, so it makes more sense for us to get paid in them.
Q. Is there a limit as to how many systems I can buy?A. Within reason, no. Indiegogo has limits on how much you can contribute in one transaction, please go to
http://support.indiegogo.com/hc/en-us/articles/526846-Common-Contribution-Errors and go down the page to the
Indiegogo Restictions on Payment Type section for more information on this. Essentially you pay on purchase so if you try to buy 7k$ worth you’ll need that on your card.
Q. Why don’t you take Paypal as a payment method?A. Honestly? It’s all to do with commitment, we don’t want to get into the situation where we have a large number of buyers, or competitors posing as buyers that decide before their 90 days is up that they really don’t want to participate in the project. It can screw things up for everyone else.
Q. I like the idea of the Guaranteed Supply Program but I don’t understand why you only offer it to the first 1000 customers who pay the full purchase rice up front? Why can’t everyone get it?A. It’s all about commitment. We want to keep all of our customers happy and devise schemes that will maximise their ROI and maintain the value of their purchases from us. However, it’s a two way street – those who are willing to help us out in the initial stages of the project by committing more of the purchase price should be rewarded more, that’s only fair. Plus we want to make absolutely sure we are able to fulfil our commitment to make rigs available when the customer needs them.
Q. Can’t you deliver the systems quicker? The longer delivery takes, the less I make!A. That’s not possible. We don’t give over-optimistic, unachievable timescales. But if we get extra funding the profit will be used to expedite delivery as much as we can.
A. For this project, 40nm is very cost efficient. Our design is very compact and efficient and so can compete with current 28nm (and even 20nm) offerings. We want to keep costs low for our customers and that’s exactly what’s been achieved – unbeatable $/GH values. What’s the gain for our customers if we using a more advanced and costly technology?
However, if we get significantly overfunded to a minimum of $12 million then we’ll use the extra profits to implement our core asic in 28nm chip instead of 40nm (although we’re pretty confident that the rev 2 40nm chip will give just as good financial results). A 28nm design would have twice as many pipelines on the same die size, but run at a lower clock of 400MHz, giving a hashing speed of 200GH/sec, ie a speed increase of 60% meaning that a NOVA-1 miner would run at 1.6 TH/sec at the same power consumption as the 40nm solution!
Q. Will you be making other Bitcoin related products in the future?A. Definitely. For a start, we plan to introduce the Generation 2 HMs in early 2015 to give our customers an upgrade path which will prolong their competitive advantage. We will be implementing some alternative low power structures in the initial chips, and the results from these will shape the exact design of the next generation chips.
Q. I like the modular approach and upgrade path, but at this time I’m one of the people you mentioned that has limited funds, so can only afford the basic system. Can I upgrade it before the delivery date, and how much would it cost?A. The short answer is a qualified ‘yes’, but please bear in mind that we have to be fair to everyone in the queue. If systems were available you would have to go to the back of the delivery queue, otherwise we would have lots of people ordering the 1 TH system then asking to upgrade it to a 8Th one before delivery, and that’s unfair.
Q. Will Novello be mining Bitcoins as well as making Rigs?A. That’s clear case of Conflict of Interest, so no. Who on earth would want to buy from a company that’s a direct competitor? Worse still, how could any normal company with any ethics expect their customers to accept such behaviour?
Q. What’s in it for you by offering GSP? A. We want to keep our customers profitable and buying more TH from us and will give them major incentives to do so – they can keep competitive no matter what happens to the NTR and DIF. In GSP we give them hahing capacity when it’s most needed – there’s no catch, just a new way of looking at the supplier/miner partnership.
Q. Are you going to sell the chips separately?A. Yes, we’d like to do this and to encourage the Bitcoin community to experiment with our asic. The chps are able to run at higher clock speeds and we are keen to gather data on how far they can be pushed. First we have to be satisfied that we can supply a device with the same standards of support as a commercial one, this won’t be possible until about January 2015 by which time we hope to have appointed distributors – ideally from within the Bitcoin community – to coordinate sales and support. As for third party system builders, we can’t really see how they could compete against our product pricing as they have to buy chips from us, but that’s not saying that we wouldn’t do it if there is enough interest.
Q. How accurate are the power estimates?A. We have extensive experience with this technology, and our figures are based on existing structures that are already in production, so our figures are accurate to about +/-10%.
Q. I like the self-repair function, but I’d rather have the extra pipelines to use for hashing. Is this possible and can the ASIC devices be overclocked?A. The short answer to the first part is no. As regards the second part, the chips are capable of running at higher speeds, but then the power dissipation goes up and so the die junction temperature. At this stage we’d rather that customers accept the system specifications as they are; once we complete system integration and testing, it may be that the power consumption is lower than our worst case predictions and so clock speeds can be increased. W
