Early internal calculations show a final chip that could perform
24% quicker than current best available ASIC
OVERVIEW
The goal of the Company is to develop disruptive Bitcoin mining
technology, to mine both faster and with less overall energy
consumption than current practices. A number of advanced
technologies are being used by QBT to achieve this goal; namely,
quantum computing, AI Neural Networks - Deep Learning,
Algebraic-Boolean reductions, Very Big Data, Cryptography and
custom chip programming and design - using GPU, FPGA and ASIC
chips.
The current technique used by producers of Bitcoin mining
technology on dedicated computers to achieve the fastest
performance, is by manufacturing single purpose, customised ASIC
chips, which can perform only one wired function, i.e, the
computation of the double hashing; the SHA26 cryptographic
algorithm used to extract Bitcoins. The simple reality is that the
faster the algorithms are computed and the more ASIC chips
deployed, the more chances a miner has to extract Bitcoins.
Before manufacturing an ASIC chip, which is an expensive
operation, there are usually two initial steps; firstly, to develop
the logic gates architecture which will be used by the final ASIC
chip – this is performed on a cheaper but slower chip, called an
FPGA, which already contains some pre-defined functions - and
secondly, by customising the design to take advantage of the
greater freedom offered by ASIC technology, initially by
manufacturing a prototype in a small batch, to keep costs low. The
final stage, manufacturing the completed ASIC chip, is an expensive
process, but the end result is a very small scale (currently up to
5nm) processing chip, which is significantly quicker, leading to
greater results when mining Bitcoin.
QBT has now completed the FPGA development phase and is moving
to develop its ASIC protype.
Initial estimates derived from the FPGA performance obtained
from our internal testing, would indicate that when the final
industrial ASIC prototype design is completed it could outperform
the fastest ASIC chip, currently being used to mine Bitcoin by at
least 24%.
Moreover, early experimental evidence, using AI techniques to
multiply by several factors the speed of an FGPA for computing the
Bitcoin mining algorithm, would make even an FPGA a competitive
Bitcoin mining tool. This same principal would apply to ASIC and
other existing commercial mining tools. Tests on this innovative
approach, will continue over the next three months.
DETAILED VIEW
Following testing of a variety of design options, an unrolled
SHA-256 architecture has now been implemented, with a number of
existing optimisations coded, for QBT’s FPGA chip prototype.
The Company’s patented ASIC ULTRA Boost improvements (patent
under application) will be added in the next few weeks, as a result
of the close cooperation between the in-house cryptography expert
and the Company’s FPGA designer.
Current performance of the Bitcoin mining architecture developed
by QBT on the FPGA (based on 16nm technology, at 600MHz basic cycle
and the average general purpose available coding area) is 2.8
Giga Hash per second (GH/s) with an
estimated 50W energy consumption. The ASIC ULTRA Boost optimisation
should improve the above performance by 7% as previously
reported.
To put things into context, our in-house expert has calculated
that, as of today, a top of the range FPGA (using QBT’s
architecture of the implementation of the algorithm, to make it run
at approximately 15 times faster than the standard FPGA), is still
approximately 23.4 times slower than the best-in-class existing 7nm
ASIC Bitcoin mining chip and much less energy efficient.
It was never the Company’s intention to compete on speed and
energy efficiency against an ASIC chip, using an FPGA chip but, in
order to keep testing costs significantly lower, it has been a
necessary step for QBT to take in this phase of its development. As
a result, the Company is now in a much better position to assess
the performance projection of its SHA256 Bitcoin mining
architecture, and therefore the team is confident that it can now
transfer this solution over to an ASIC chip.
Preliminary approximate calculations indicate that on a 12nm
ASIC chip, extrapolated by comparing it to a commercial mid-range
16nm ASIC Bitcoin mining chip (with circa 300 million gates), our
ASIC could achieve a double hash rate of 392 GH/s, with the ASIC
ULTRA Boost optimisation adding an extra 7% as previously
announced, reaching 419 GH/s, which would still be 2.26 times
slower than the fastest ASIC commercially available.
However, the Company strongly believes that when compared with
the best commercial 7nm ASIC chip for Bitcoin mining available on
the market today, an industrial production of QBT’s ASIC at 7nm,
would indicate a double hash rate of 1.19TH/s, hence 24%
faster.
This performance of QBT’s architecture on the 7nm ASIC, doesn’t
yet include the 7% efficiency achieved by the optimisation of the
patent application filed in September, which is still to be
implemented. The current work on a second patent by the Company’s
cryptography expert will hopefully lead to further material
optimisations.
Detailed simulations on energy consumption will be run as soon
as our ASIC gate design layout is completed.
ASIC programming will start this month, and we estimate that by
the end of Q1 2022 we will be able to announce when the first batch
of prototype chips will be available for in-house testing.
Following the completion of testing, the Company envisages that
chip production, for QBT’s own use, will commence by the end of
2022.
Concurrently, QBT’s R&D team is also considering an
alternative SHA256 computing approach (which is categorised as the
basic Bitcoin mining algorithm), and this will be tested within the
next three months. The joint effort by the members of our AI team
and the Company’s FPGA expert, could improve the current FPGA hash
rate of 2.8GH/s by a highly material multiple factor, making mining
by the slower FPGA chip potentially competitive against the
best-in-class ASIC chip. Should this route be successful, mining
via this method could commence as early as Q2 2022.
These AI techniques, if successful, could also improve the
performance of current commercial ASIC Bitcoin miners.
The Company’s new R&D IT infrastructure, which will also
allow for Bitcoin mining tests to be carried out, will be
operational in five weeks’ time. The wait has been due to the
serious worldwide shortage of silicon chips, which is delaying the
expected delivery of the hardware. However, the Company has adopted
the heavy use of cloud resources in order to avoid any interruption
in the R&D activities to the various groups.
The Company remains very confident on the R&D strategy it
has adopted which it believes could result in disruptive Bitcoin
mining. It is worth noting that the Company’s R&D programme is
fully funded until the end of 2022.
Francesco Gardin, CEO and
Chairman of QBT, commented, “Our R&D has delivered some
very impressive results in a very short time: In only four months
since the programme commenced, we have filed a patent application
where we believe the ASIC ULTRA Boost has improved the standard
mining algorithm, after five years of little or no progress
following the publication in 2016 of the ASIC Boost paper.
“We will soon be ready to start the design of our ASIC Bitcoin
mining chip which, on paper, already outperforms, in speed, the
current best in class ASIC Bitcoin mining commercial solution. This
significant improvement is before the implementation of the new
optimisation from ASIC ULTRA Boost and we are confident that our
second patent application, which is under development by our
cryptography expert, will add a further radical improvement to the
process, including also a reduction in energy consumption.
“All our other teams are working extremely hard on the other
R&D fronts: Quantum computing and AI Neural Networks-Deep
Learning and algebraic-Boolean optimisation. Meanwhile an AI
accelerator will be tested within the next three months, which we
believe could radically improve the performance of existing
commercial miners, as well as our GPU, FPGA and, in the near
future, our AISC chip. We consider that the R&D activity
undertaken by our group of 15 experts is unbelievably exciting and,
if successful, could potentially be radically innovative for the
industry.”