Little else known about the E10; semiconductor industry insiders doubt Ebang’s 10 nm chip claim.
Could the next generation of ASIC Bitcoin miners be upon us?
Ebang, a Chinese Bitcoin mining hardware manufacturer with roots in the communications technology industry, recently updated their miner-focused web page to display an image announcing the “upcoming” arrival of the Ebit E10. The E10 will supposedly be the first Bitcoin miner in the world to sport chips based on a 10 nm manufacturing process.
The announcement caught the attention of the Bitcoin mining community for three reasons: (a) new hardware never fails to get miners excited; (b) moving to a smaller manufacturing process usually entails significant performance and efficiency gains, and these never fail to get miners excited; and (c) mining hardware based on a 10 nm node process at this time sounds too good to be true, and this never fails to get miners excited (although probably not in the way that Ebang had hoped).
If Intel Can’t Do It Yet…
Similar to how Japanese conglomerate GMO Internet Group’s $90 million plan to develop 7 nm Bitcoin mining chips by spring 2018 attracted scathing ridicule from semiconductor industry insiders in the Bitcoin community, Ebang’s claim to have 10 nm-based miners in the works was likewise greeted by a familiar chorus of incredulity.
In particular, much of this disbelief stems from Intel’s own prolonged difficulty with moving their own manufacturing process down to 10 nm. Intel even had to retire their signature decade-long “tick-tock” development model and delay the release of Cannon Lake — originally intended as the 10 nm successor to 14 nm Skylake — from late 2016 to sometime in 2018. And even then, Cannon Lake will be restricted to low-power mobile chips at first, before Intel’s planned second-generation 10 nm+ process enables yields good enough for larger, more powerful desktop and server chips.
So, the argument goes, if Intel — long regarded as being on the bleeding edge of chip design and production — can’t yet nail down the move to 10 nm, then how can Ebang claim to have developed a production-ready line of 10 nm-based miners? Furthermore, Intel’s 10 nm woes cannot reasonably be attributed to apathy on Intel’s part (unlike Kaby Lake), as the current difficulty of moving to 10 nm has deep roots in the laws of physics.
Let There Be (Ultraviolet) Light
Modern microprocessors are manufactured using a process called photolithography. This is a process where ultraviolet light with a wavelength of 193 nm — generated using an ArF (argon fluoride) laser — is used to etch the circuit and gate patterns on a silicon wafer that results in the microprocessors we know and use. However, as chip designs move to smaller node processes, it becomes more difficult to manipulate such a large wavelength into etching increasingly smaller features. A narrower beam of light would be needed to etch those features with a commercially-viable level of accuracy.
This is where EUV (extreme ultraviolet) lithography comes in. EUV uses a wavelength of just 13.5 nm, which is perfect for taking chip manufacturers down to the sub-10 nm realm, in theory. In practice, however, EUV systems have proven extremely difficult to deploy on a commercial scale, as the EUV light itself has so far been extremely hard to generate and even harder to manipulate.
Add the fact that silicon by itself becomes practically unusable at sub-10 nm levels — IBM added germanium to make their 7 nm test chip work — and it becomes even more apparent that sub-10 nm general-purpose microprocessors, let alone powerful Bitcoin mining ASICs, may take even longer to arrive than previously anticipated.
Bang for Your Buck, Or Just Bang?
There is still much that we don’t know about the E10. Ebang did not announce estimated hash rate nor power consumption figures, let alone a release date. The only tidbit we’re able to go by at the moment is that the E10 will somehow sport 10 nm chips and will be “upcoming globally.”
Nevertheless, I doubt that the E10 will live up to the 10 nm hype when (or even if) it’s finally released. I find it farfetched that an obscure (by comparison) communications technology company somehow managed to crack the 10 nm puzzle when the world’s largest chipmaker has so far been unable to. The technology needed to get there at scale is not even fully mature yet, and I haven’t even gotten to the question of whether the major chip fabricators would allow Ebang early access to a technology that is still, for all intents and purposes, in development.
A pinch of salt isn’t enough for me here. I’m taking the E10 announcement with a whole tub of it.