Obelisk is pleased to announce our next generation mining system: The Obelisk GRN1, targeting the Cuckatoo31 proof of work algorithm. More than just a chip, the GRN1 features improvements to the hashing boards, control boards, mechanical design, and even the PSU.
The Grin team has stated here that they intend to encourage the development of ASICs for the Cuckatoo31+ family of algorithms on Grin, and provided code which defines a specification for the Cuckatoo31+ algorithms, as well as code which defines the schedule for ramping up the block rewards for Cuckatoo31+ miners, and the code for progressively phasing out each magnitude of Cuckatoo.
One of the most difficult elements of hardware development is the lack of flexibility. Hardware developers need to commit to a specification month in advance, and even slight tweaks can result in months of delays. With the post linked above, the Grin team has given Obelisk the required confidence to move forward with development and to commit to a final design with confidence. This reduces risk, which in turn makes it easier for manufacturers to develop hardware, as well as makes overall development cheaper.
The most expensive part of a typical mining machine is the custom chips that do most of the computation. Typical mining machines have a large number of very tiny chips which run very hot. Machines with lots of memory, however, tend to have much bigger chips that are not as hot; memory is very expensive to manufacture, and it does not consume much power.
As a result, we will be building a GRN1 which has a large number of very large chips. To create the most competitive miner possible, we will be packing as much density into our machines as we can, and in the case of a high memory algorithm like Cuckatoo31, this means the machines will be expensive, as the chips themselves are generally the most expensive component.
An S9 has about 3,500 mm2 of silicon on it in total. The GRN1 is expected to have closer to 10,000 mm2 of silicon, and the price, as a result, will be correspondingly higher.
Obelisk will be conducting a two-phase sale of GRN1 miners. The first phase is a voucher sale where we will be selling $3,000 vouchers, each convertible into 1 Cuckatoo31 mining unit. This sale will be open from January 17th to February 1st, and a total of 10,000 vouchers are available.
The second phase is a full machine sale. The full machine sale is expected to go live in March, with units being sold for $6,000 each. Any voucher holder will be able to exchange their voucher to a full unit at no additional cost.
If for some reason a unit no longer seems like an attractive deal, a voucher holder can elect to receive a refund of $2,000 instead of a Cuckatoo31 miner. This refund will be provided no-questions-asked, and with no justification required. Upon launching the full machine sale, you will have two weeks to elect a refund.
During the first phase, we will be setting aside $100 per voucher sold to fund Grin development. During the second phase, we will be setting aside $500 per new unit sold for Grin development. The majority of the money collected for Grin development will be used to fund no-strings-attached stipends for independent developers. The remaining money will be set aside to fund high-priority endeavors such as security audits.
For the first phase, we are only able to release tentative specifications. While we have an architecture and rough performance simulations completed for that architecture, we have not finalized the chip nor completed the final verification and layout stages of design. We, therefore, are not comfortable committing to these specifications.
That said, we are tentatively expecting each unit to be able to perform at least 100 graphs per second while consuming no more than 800 watts. This can be compared to the current leading graphics card, the NVIDIA RTX 2080 Ti, which performs less than 2 graphs per second and consumes approximately 250 watts.
Similarly, we are only able to release tentative shipping dates for phase one of our unit sale. We have reasonable timelines established for each part of our supply chain, from chip development to wafer production to bare board manufacturing and final assembly, however we do not have final commitments from each of our suppliers, and we also do not know if unexpected problems will show up during layout and verification of our chip design. We are not comfortable releasing final dates until we have high confidence that each requirement within our supply chain can be met by the intended shipping date.
That said, we are tentatively expecting to ship units throughout the month of October.
It is our goal to see that Grin is secure against 51% attacks and censorship attacks, both because the hashrate is decentralized and also because the incentives of all the hardware owners are aligned with protecting Grin as opposed to attacking it.
We believe that Grin has a good foundation to achieve these goals, and we are excited to be manufacturing an ASIC that has the potential to be a force of good for the Grin community.