Topic: What happens after 16nm? (Read 2993 times)
Another company has released the 14 nm chip. So the ASIC has reached the state of the art. It is more difficult to make more efficient chip now.
But seriously... Btc ASICs have jumped from 135nm and up to now 16nm in just a few years. Imagine...
This was actually my main point. ASICs developed fast when there was a big gap compared to general purpose IC:s. But now that ASICs have reached them I guess, we won't see that fast change any more.
And there is good reason for it there is a TON of profit if a big company does it right. Want to see one that does look at SP released last year profit's. They are one of few not private companies, so they have to release data.
As long as there is profit we will continue to see a jump in lower nm, or more optimized current gen. Hard to predict in a year even what gear will look like, they huge jumps in efficiency. And this will only continue as long as it's highly profitable or these big companies. What will be interesting is if a lot of hash keeps getting added then, having market could change.
But seriously... Btc ASICs have jumped from 135nm and up to now 16nm in just a few years. Imagine...
This was actually my main point. ASICs developed fast when there was a big gap compared to general purpose IC:s. But now that ASICs have reached them I guess, we won't see that fast change any more.
Quantum computing, as currently implimented and forseen, doesn't work very well or efficiency on simple liner computations like Cryptocoins use.
It's entire strength is dealing with MANY variables at one time in a very complex way.
Well, we could make a complex equation that mines btcoin for us. It's entire strength is dealing with MANY variables at one time in a very complex way.
wouldn't that be good too?But seriously... Btc ASICs have jumped from 135nm and up to now 16nm in just a few years. Imagine...
Quantum computing, as currently implimented and forseen, doesn't work very well or efficiency on simple liner computations like Cryptocoins use.
It's entire strength is dealing with MANY variables at one time in a very complex way.
It's entire strength is dealing with MANY variables at one time in a very complex way.
Now this is a shot in the near dawn light, but by the time we have any sort of realistic use for chipsets smaller than the current gen 16nm chips we'll probably be working on levels of quantum computing.
B-Eleven (which is 14/16nm but not full-custom), S7, and Avalon 6.
I suspect THOSE are the primary movers on the last month of hashrate gains.
I suspect THOSE are the primary movers on the last month of hashrate gains.
We've been using the other chip sharpnesses for a long time. 16nm ASICs haven't been released yet, so when it will be we'll be able to wait until 2017.
The difficulty rose 20% last time, and it is predicted to rise another 20% this time. I think the manufacturers are already mining with 16 nm.
you should go in EEtimes website there is everything you need to know.
We've been using the other chip sharpnesses for a long time. 16nm ASICs haven't been released yet, so when it will be we'll be able to wait until 2017.
BitFury MIGHT be ahead, but I don't think they're actually ahead of everyone - or only by a month or two at best for them.
Innosilicon has been playing things pretty close since their last announcements in November - which "advance talk" was uncommon for them, given they didn't say word one about the A2 more than a week or so before they actively started selling those units (I'm pretty sure they were fairly mum about the A1 before that chip was actively available too).
In THEORY KnC was months ahead of BitFury, but the silence on their "Solar" stuff for the last quite-a-few months has been resounding.
Bitmain - probably behind but they like to milk "current stuff" as long as they can before they introduce a new generation.
When they start doing sales on "used S7s" you know their 14nm gear is hitting or about to hit full production (ref the last S5 sales vs the initial S7 sales timeframe).
Dunno what's going on with SFARDS, but their SF100 unit is already pretty far behind the curve on it's SHA256 side - as usual for their "dual miner" chips half of it is obsolete in a hurry the other half is still competative for a while wasting a large part of the cost of the chip.
Innosilicon has been playing things pretty close since their last announcements in November - which "advance talk" was uncommon for them, given they didn't say word one about the A2 more than a week or so before they actively started selling those units (I'm pretty sure they were fairly mum about the A1 before that chip was actively available too).
In THEORY KnC was months ahead of BitFury, but the silence on their "Solar" stuff for the last quite-a-few months has been resounding.
Bitmain - probably behind but they like to milk "current stuff" as long as they can before they introduce a new generation.
When they start doing sales on "used S7s" you know their 14nm gear is hitting or about to hit full production (ref the last S5 sales vs the initial S7 sales timeframe).
Dunno what's going on with SFARDS, but their SF100 unit is already pretty far behind the curve on it's SHA256 side - as usual for their "dual miner" chips half of it is obsolete in a hurry the other half is still competative for a while wasting a large part of the cost of the chip.
Not real likely, the big gain from the S5 to the S7 (and the SP20/etc to the SP50) was moving from standard-cell design to full custom design.
Bitfury's new chip is already full custom - as Bitmain has stated for their next gen "in the works" chip and Innosilicon has specified for the A4 (and by implication the A3).
For perspective, the B-Eleven is 14/16nm but NOT full custom, and it's supposedly right in the same efficiency range (when it hits production) as the S7.
Bitfury's new chip is already full custom - as Bitmain has stated for their next gen "in the works" chip and Innosilicon has specified for the A4 (and by implication the A3).
For perspective, the B-Eleven is 14/16nm but NOT full custom, and it's supposedly right in the same efficiency range (when it hits production) as the S7.
That B-Eleven's efficiency is similar to the S7. So it is about 5 months late and it will make them some loss.
I think Bitfury beat a lot of companies is part of it. B11 really needed to be earlier to make a big difference, although if they do low enough cost it might still work. So a lot of these companies that did chip dev I think will be forced by Bitfury to sell at lower prices.
There are some companies like SP I really wonder what they will do. Counting on big buyers on SP50, when Bitfury beat them it would seem. It will be really interesting to see what happens when companies like those two compete for the big sales when it seems Bitfury is ahead so far.
Not real likely, the big gain from the S5 to the S7 (and the SP20/etc to the SP50) was moving from standard-cell design to full custom design.
Bitfury's new chip is already full custom - as Bitmain has stated for their next gen "in the works" chip and Innosilicon has specified for the A4 (and by implication the A3).
For perspective, the B-Eleven is 14/16nm but NOT full custom, and it's supposedly right in the same efficiency range (when it hits production) as the S7.
Bitfury's new chip is already full custom - as Bitmain has stated for their next gen "in the works" chip and Innosilicon has specified for the A4 (and by implication the A3).
For perspective, the B-Eleven is 14/16nm but NOT full custom, and it's supposedly right in the same efficiency range (when it hits production) as the S7.
That B-Eleven's efficiency is similar to the S7. So it is about 5 months late and it will make them some loss.
IBM recently announced that it has recreated the Silicon substructure used in making processors. Currently silicon can go to 10 nano meters but IBM's can go to 8nm with there new silicon.
Doesn't affect Bitcoin, you have to remember their are only a handful of chip foundries in the world. You will not even get access to that tech for years, there are large multi-national corps that have priority to those runs and technical data. Bitcoin will always be getting the last few generations of die sizes, many of the 10nm & 7nm runs are booked out 18-36 months in advance.
Not real likely, the big gain from the S5 to the S7 (and the SP20/etc to the SP50) was moving from standard-cell design to full custom design.
Bitfury's new chip is already full custom - as Bitmain has stated for their next gen "in the works" chip and Innosilicon has specified for the A4 (and by implication the A3).
For perspective, the B-Eleven is 14/16nm but NOT full custom, and it's supposedly right in the same efficiency range (when it hits production) as the S7.
Bitfury's new chip is already full custom - as Bitmain has stated for their next gen "in the works" chip and Innosilicon has specified for the A4 (and by implication the A3).
For perspective, the B-Eleven is 14/16nm but NOT full custom, and it's supposedly right in the same efficiency range (when it hits production) as the S7.
I don't mean, we wouldn't ever go beyond 16nm or 10nm, but Moore's law is apparently slowing and we will be longer on 16 nm than on previous stages.
I think the mining chips will use the 16 nm for a few years. It will be too costly to go lower than that. But it depends on the bitcon price.
pretty much this is what will happen.
We were at the 28 stage a while since the sp20 and s-5 were early 28 mm and the s-7 is newer 28 mm
The 28 was doing .5 watts and with improvements and tweaking it got as low as .24 watts. as my batch 2 s-7 could do .24 watts a gh with a titanium psu on a downclock to freq 550
So bitfury has shown its 16mm chip to do about .06 watts to about .11 watts
Maybe a newer 16mm chip in 1 year will do .04 watts to about .08 watts. If true the 16 mm gear will be longer lasting.
It could last until the next ½ ing due in DEC 2018
I don't mean, we wouldn't ever go beyond 16nm or 10nm, but Moore's law is apparently slowing and we will be longer on 16 nm than on previous stages.
I think the mining chips will use the 16 nm for a few years. It will be too costly to go lower than that. But it depends on the bitcon price.
I don't mean, we wouldn't ever go beyond 16nm or 10nm, but Moore's law is apparently slowing and we will be longer on 16 nm than on previous stages.
IBM recently announced that it has recreated the Silicon substructure used in making processors. Currently silicon can go to 10 nano meters but IBM's can go to 8nm with there new silicon.
I think 16nm is a good place to consolidate the market and see how the BTC halving plays out.
There's no choice, as there's nothing AVAILABLE at this time or for years from now in a commercially-available smaller process.
Jump to: