RIP Grin Mining? Weak outlook for C32 hardware

With the low price of Grin, sales of new mining hardware, both GPU and ASIC, have come to a grinding halt. ROI is negative for miners. Accordingly, ROI is very negative for ASIC and Rig manufacturers requiring expenditures of $5-$10 million for production . Since sales are not materializing, neither Innosilicon or Obelisk have any incentive to tape-out to semiconductor foundries and spend millions for wafer mask, test boards and other prototyping related cost. It is cheaper to refund customers than to throw away several million dollars for an ASIC that no one wants to buy.

If production dates are to be achieved, Innosilicon needed to have taped out aalready and Obelisk needs to do so shortly.

At the Magical Crypto conference yesterday, Innosilicon staff privately admitted that Grin miner sales are very slow. Obelisk’s published pre-sales are tiny, not nearly enough to cover a portion of tapeout cost. Vidtoo seems to be non-existent as they would have samples in-house in order to make a July ship date. The negative economics and Innosilicon’s history of cancelling programs such as their A10 Ethereum miner, it’s very likely they will not ship their Grin miner and also not let any one know.

It is quite conceivable with no ASICs and minimal GPU support that the C32 hashing rate will fall below 10,000 graphs/second in 2020, making the network insecure and subject to attack.

GPU’s are not a viable fallback strategy as the population of GPU’s capable of mining C32 at 1 GPS at full fidelity is minuscule. Only Nvidia and AMD high end GPU’s with 16GB or more memory can mine C32. These boards will be priced at $1500-$4000+. High end GPU’s will be in limited supply and in high demand by AI users. They are unlikely to be used for C32 mining unless rented to mount an attack on Grin.

The C31 POW phase-out sounds like a doomsday scenario for Grin and it’s mining network. With negative economics and a high probability of C32 miners being stillborn, I would urge the Grin Governance Committee to delay the phase-out of C31.


More thoughts on this topic …

Innosilicon can claim all they want about releasing a C32 miner but their cancelled A10 Ethereum miner is an example of how Inno can cancel a program without notice. They were supposed to ship in September 2018. Even with a metal spin fix, they should have shipped in the January timeframe if there was demand and if there was enough sales to generate profits.

If Innosilicon cancels, does that turn Obelisk profitable by your calculation, and vice versa?

Delaying C31 phaseout would be brutal toward ASIC manufacturers. It does not look like an option on the table, to me.

I said from before launch that the dual PoW was a bad idea, that shrinking the size of ASIC rewards would only reduce the number of competitors…

However, don’t let public numbers fool you. You don’t know how many private sales each manufacturer has. In general, almost all ASIC sales are going to come from large miners, not retail channels.

1 Like

I’ve heard that Innosilicon have sold upwards of 3000 units (I believe, this includes their small PCIE card as 1 unit). I was one of the last purchasers of the first batch right before the deadline.

I’m not sure how many Obelisk has sold. Can anyone tell me / link a source?

It’s increasingly looking bad for me to mine. I’ve seen multiple sources predict Grin’s price to fall, which is just the nail in the coffin. Not to mention further sales which will shoot up the hashrate and difficulty.

Unless some magic hubba-bubba can happen in the background, I’m kinda scared. But I hope for the best. I really do.

1 Like

Sources can be correct or incorrect. Price predictions are notoriously difficult.

Project development program can greatly influence the price. Grin should find the scope of application and do everything possible to implement this program. Otherwise, inflation will make mining unprofitable and this will be the end of the project.

The nuclear winter for ASIC crypto miners started early in Sept-Oct 18 and continues into the traditional summer months of 2019.

For those in the foundry business, they know Wafer Starts will be tight for the newer nodes.

The capacity will go to AI, Datacenter, Mobile, or HPC applications before crypto orders.

So it will get harder and not easier in the next few quarters.

1 Like

It may be that 1g/s is hard to achieve with todays customer hardware, but with slean mining at least 0.4 g/s+ should be achieveable with

AMD Vega 56/64, Radeon 7
Nvidia RTX 2080 (Ti), 2070, GTX 1080 (Ti), 1070 Ti

The higher end 2080ti / Radeon 7 should be able to do 0.7g/s

Medium end cards as 580 8G and 1070 8G still should be able to do 0.25g/s+.

I am aware that these numbers are lower then desired, but conservative assumptions. At least I am very confident the chain will not stop or struggle even when AR29 and AT31 are phased out.

Sources can be correct or incorrect. Price predictions are notoriously difficult.

Yes, I fully agree.

But it seems very unlikely that the price will rise to a satisfactory rate for grin miners that bought ASICs.

Unless the Grin Gods find a tangible use for this coin in the mainstream & get it listed on Coinbase / Binance - it looks like a shitcoin for me, and a shitstorm once all of the miners pollute the market with their hashrate.

The increase in hashrate could be good for the coin’s price, but due to the high emission rate I don’t think it’ll be a quick increase in price. It could take a long time for it to reach a satisfactory figure…

Oh well. I hope for the best! Crypto is a risk and I took the risk. I like risks. I like money. That’s all.

I don’t have an account to read the whole article, but I’d assume it’s only 7nm that gets crowded this year. Aren’t all the Grin miners targeting 16nm (actually 12/10 optical shrink)? The foundries plan their production pipeline wayyy in advance, and any Grin ASIC manufacturers would have at least given volume guidance to the foundry months ago. When we were doing the CryptoNight ASIC, we effectively had wafer production reservations long before tapeout.

Regrettably, the foundries look at the volume guidance from the crypto customers with skepticism.

That is due to the un-used capacity that was left behind in 2018. And if there were any take or pay agreements, they were not honored.

The dynamics are very different in 2019. Gone are the assumptions that pre-sales could fund a development. And capacity would be reserved for the ASIC crypto-miner.

The capacity tightness is in both 7nm and the prior generation(s) of finFET (16/14nm and 10/8nm).