Pages:
Author

Topic: HashFast launches sales of the Baby Jet - page 24. (Read 119648 times)

hero member
Activity: 518
Merit: 500
Every man is guilty of all the good he did not do.
November 06, 2013, 09:12:04 AM
Probably the KNC design is the only sense that would make sense to have at 14 or 20nm.
It's cheap for them.
It's just an fpga conversion, right?

I would be surprised if they didn't do a full custom at this size.

They saved a lot of money/time by not doing a full custom on the 28nm node allowing them to push the money saved/made from being first to produce to the lower node and maintaining their competitive edge as well as being the first on that node as well.

Also to correct you, their last 28nm is standard cell which is one step up from a straight forward FPGA conversion (like eASIC)

I agree with Puppet that the NRE will be very difficult to gather back as the price of ASICs become less and less expensive, however KnC might be in the only position to actually pull this off since they got a slight head start. If they do produce this first, I imagine that any other Bitcoin ASIC Manufacturer would literally be stuck with their 28nm node as the NRE could never be paid for at that point.

Then again, could just be a marketing ploy that other vendors have used to trick customers into thinking the company they chose will stay competitive.
legendary
Activity: 1176
Merit: 1001
November 06, 2013, 06:19:08 AM
Probably the KNC design is the only sense that would make sense to have at 14 or 20nm.
It's cheap for them.
It's just an fpga conversion, right?
hero member
Activity: 784
Merit: 1004
Glow Stick Dance!
November 06, 2013, 06:18:06 AM
There will be no 20nm miner by any vendor for quite a while.   20nm being available =/= 20nm cheaper than 28nm.   20nm will probably be available in volume next year (at 50% to 150% higher prices).  It generally takes 2-3 years before a new process node becomes cheaper than the prior one.  Maybe in late 2016 but 2017 seems more likely.

Color me surprised, KnC claims to be working on a 20/16nm chip already:

Unbeknown to the community we are currently developing the next generation product in 20nm/16nm process with Alchip,

https://www.kncminer.com/news

I very much doubt this chip will materialize, as the economics most likely dont make sense (by the time it would arrive, price per TH would be so low it would be very hard to recover the NRE), unless they start another round of preorders fairly soon and enough people are dumb enough to fall for it yet again.

NRE?  Don't you think that maybe KNC already has those costs covered from the revenues generated by their Gen 1 devices?  Would they really need preorder money at this point?
legendary
Activity: 2856
Merit: 1520
Bitcoin Legal Tender Countries: 2 of 206
November 06, 2013, 03:23:51 AM
^^which paste is missing for the HashFast chips?  Grin
legendary
Activity: 980
Merit: 1040
November 06, 2013, 03:19:50 AM
There will be no 20nm miner by any vendor for quite a while.   20nm being available =/= 20nm cheaper than 28nm.   20nm will probably be available in volume next year (at 50% to 150% higher prices).  It generally takes 2-3 years before a new process node becomes cheaper than the prior one.  Maybe in late 2016 but 2017 seems more likely.

Color me surprised, KnC claims to be working on a 20/16nm chip already:

Unbeknown to the community we are currently developing the next generation product in 20nm/16nm process with Alchip,

https://www.kncminer.com/news

I very much doubt this chip will materialize, as the economics most likely dont make sense (by the time it would arrive, price per TH would be so low it would be very hard to recover the NRE), unless they start another round of preorders fairly soon and enough people are dumb enough to fall for it yet again.
donator
Activity: 1218
Merit: 1079
Gerald Davis
November 04, 2013, 06:17:16 PM
Where are we heading after silicon? Graphene?

I don't know.  There is lots of cool stuff being researched but nobody has been able to figure out how to make it economical (yet).  It is a big difference between someone made a transistor on xyz exotic tech and someone can crank out chips with billions of those transistors at a cost of a few pennies per mm2.

The "problem" is that Silicon has been too good.   When Silicon can double (transistor per mm2, computing per watt, transistor per $) every 24 months it is very tough to compete with that.   I guess over the next decade as silicon runs out of steam it will create an opportunity for someone to get the lab tech out of the labs and into the foundries.  Then again people have been talking about all that exotic tech for the past three decades and we just keep using silicon and make the transistors smaller, faster, cheaper each generation.
legendary
Activity: 1904
Merit: 1007
November 04, 2013, 06:04:33 PM

It is unlikely a process node will be skipped.  Fast forward ahead 3,4, x years 20nm will now be mainstream production with reasonable costs and 14nm will be the cutting edge insane NRE and ultra high wafer costs.  Essentially the roles 28nm and 20nm are now.  Still I agree even 2016 is dubiously optimistic and those thinking it will happen next year are just delusional. 

It is getting harder and harder for foundries to maintain Moores law.  Each generation the fabs cost most, there are more delays, the cost benefits get smaller, and it takes longer to just reach cost parity with the prior generation.   It is all symptoms of silicon lithography running out of steam.   

Where are we heading after silicon? Graphene?
donator
Activity: 1218
Merit: 1079
Gerald Davis
November 04, 2013, 05:27:42 PM
There will be no 20nm miner by any vendor for quite a while.   20nm being available =/= 20nm cheaper than 28nm.   20nm will probably be available in volume next year (at 50% to 150% higher prices).  It generally takes 2-3 years before a new process node becomes cheaper than the prior one.  Maybe in late 2016 but 2017 seems more likely.

And by 2016/2017 the mining market will be saturated and with razor thin margins, so it would be very hard for any vendor to recover the NRE. Unless they could achieve a rather dramatic improvement in GH/W, it doesnt make sense to even start. I dont think we will ever see 20nm bitcoin asics. Maybe 14nm at some point in the very far future.

It is unlikely a process node will be skipped.  Fast forward ahead 3,4, x years 20nm will now be mainstream production with reasonable costs and 14nm will be the cutting edge insane NRE and ultra high wafer costs.  Essentially the roles 28nm and 20nm are now.  Still I agree even 2016 is dubiously optimistic and those thinking it will happen next year are just delusional. 

It is getting harder and harder for foundries to maintain Moores law.  Each generation the fabs cost most, there are more delays, the cost benefits get smaller, and it takes longer to just reach cost parity with the prior generation.   It is all symptoms of silicon lithography running out of steam.   
legendary
Activity: 3878
Merit: 1193
November 03, 2013, 03:47:56 PM
There will be no 20nm miner by any vendor for quite a while.   20nm being available =/= 20nm cheaper than 28nm.   20nm will probably be available in volume next year (at 50% to 150% higher prices).  It generally takes 2-3 years before a new process node becomes cheaper than the prior one.  Maybe in late 2016 but 2017 seems more likely.

And by 2016/2017 the mining market will be saturated and with razor thin margins, so it would be very hard for any vendor to recover the NRE. Unless they could achieve a rather dramatic improvement in GH/W, it doesnt make sense to even start. I dont think we will ever see 20nm bitcoin asics. Maybe 14nm at some point in the very far future.

Feature size is starting to get muddy. 20nm and 14nm miners might not have the speed and power improvements we've come to expect from node improvements.

http://spectrum.ieee.org/semiconductors/devices/the-status-of-moores-law-its-complicated

Well designed 28nm chips are going to be the best available for a very long time.
legendary
Activity: 3878
Merit: 1193
November 03, 2013, 03:37:51 PM
Babyjets on sale for $2200-ish with no MPP

Second batch baby jets? Yeah, you're not going to see that delivered any time soon. Say goodbye to ROI.
legendary
Activity: 980
Merit: 1040
November 03, 2013, 03:35:59 PM
There will be no 20nm miner by any vendor for quite a while.   20nm being available =/= 20nm cheaper than 28nm.   20nm will probably be available in volume next year (at 50% to 150% higher prices).  It generally takes 2-3 years before a new process node becomes cheaper than the prior one.  Maybe in late 2016 but 2017 seems more likely.

And by 2016/2017 the mining market will be saturated and with razor thin margins, so it would be very hard for any vendor to recover the NRE. Unless they could achieve a rather dramatic improvement in GH/W, it doesnt make sense to even start. I dont think we will ever see 20nm bitcoin asics. Maybe 14nm at some point in the very far future.
legendary
Activity: 1484
Merit: 1005
November 03, 2013, 03:32:39 PM
you wont roi if you get it by end of year. even if you got it today you might not even roi. we had 1 chance of roi and thats why we bought it. shipping said late october. most the money would be coming in from then till now. and difficulty rate rising  tremendously makes it even less likly to roi.

If I got it today I would definitely ROI, in fact I would hit about a 60% ROI according to http://mining.thegenesisblock.com/

December is iffier and it depends on what kind of rate increases we see and how sustained they are.  The current network growth is not sustainable though (keep in mind we've gone from a 110 nm process to a 28 nm process in 10 months time, and the rise in hash rate reflects this).  The next stopgap half-node is 20 nm, and I don't think any fab is there yet for large scale, high yield production.

http://www.xbitlabs.com/news/other/display/20131022230815_TSMC_Shares_More_Details_Regarding_16nm_FinFET_and_20nm_Progress.html

Since HashFast is already working with TSMC, I would guess they'll be one of the first companies to produce on 20 nm as die shrinks (if the company survives that long, things are getting cutthroat).

Please cut the propaganda. Nobody will have access to lower than 28nm technology before 2015-2016 even and it will be damn hell expensive. Also there are many other companies that are already working with TSMC (Avalon, ASICMINER, KnC etc) so i don't understand why you praise HF so much. If they can get it then everyone will get it.

Okay

Quote
Next year TSMC will make over thirty designs using 20nm process technology, which it calls CLN20SOC with HKMG. Moreover, since there will be multiple fabs producing chips using 20nm fabrication process at TSMC, the ramp curve of every design as well as revenue share of 20nm technology will be faster compared to those of 28nm process.

“Specifically on 20nm we have received five product tape-outs and scheduled more than 30 tape-outs in this year and next year from mobile computing, CPU and PLD [programmable logic device] segments. And all those tape-outs represent big volumes. Design ecosystem on 20nm has been validated in real products and is ready to support customers. Yield learning is in line or better than the 28nm path. We expect a fast ramp of 20nm next year, with revenue from 20nm in 2014 bigger than that of 28nm in 2012. You see 20nm will be starting next year whereas 28nm actually started in the fourth quarter – third, fourth quarter of 2011. So the corresponding point for 28nm was 2012. But our ramp in 20nm in 2014 is going to be faster than the ramp for 28nm in 2012. While our 28nm ramp was a record for TSMC, 20nm ramp will be even faster by about 30%,” said Mr. Chang.

DnT claims much longer, which is good because it means your products will be profitable for longer, wherever they came from.
donator
Activity: 1218
Merit: 1079
Gerald Davis
November 03, 2013, 03:32:39 PM
...and I don't think any fab is there yet for large scale, high yield production.

I thought I read that Samsung was getting involved to do just that...

Getting involved and already cranking out millions of wafers a year are two different things.  In time everything (even cheap garbage like USB controllers) will be 20nm but it is going to be a while.  The first mass production customers will be customers that value power usage over cost.   For example a cellphone CPU might be $15.  Paying $22 or $30 (on a $400 device) and getting 30% lower power consumption is worth it.

Nobody is going to buy an 20nm ASIC which costs 50% to 100% more than a 28nm ASIC.  It generally takes 3 years before a process nodes acheives cost parity with the prior node.   Those expecting 20nm tomorrow are in for a lllllllllllllllllllloooooooooooooooooooooonnnnnnnnnnnnnnnnnnnnnnnnnnggggggggggg ggggggg wait.

Note the image is from NVidia conference.  NVidia who does runs in the hundreds of thousands of wafers (millions of units) hasn't commited to 20nm/22nm in 2014 because of cost concerns.  It is highly likely both AMD & NVidia will extend 28nm for yet another product cycle to bridge the gap to at least 2015.



http://www.extremetech.com/computing/123529-nvidia-deeply-unhappy-with-tsmc-claims-22nm-essentially-worthless
legendary
Activity: 1484
Merit: 1005
November 03, 2013, 03:31:00 PM
you wont roi if you get it by end of year. even if you got it today you might not even roi. we had 1 chance of roi and thats why we bought it. shipping said late october. most the money would be coming in from then till now. and difficulty rate rising  tremendously makes it even less likly to roi.

If I got it today I would definitely ROI, in fact I would hit about a 60% ROI according to http://mining.thegenesisblock.com/...

You mean this http://mining.thegenesisblock.com/a/89f1855c7e  Huh

Shows you losing money.  What did you order & @ what cost?

Babyjets on sale for $2200-ish with no MPP
legendary
Activity: 1904
Merit: 1007
November 03, 2013, 03:29:51 PM
you wont roi if you get it by end of year. even if you got it today you might not even roi. we had 1 chance of roi and thats why we bought it. shipping said late october. most the money would be coming in from then till now. and difficulty rate rising  tremendously makes it even less likly to roi.

If I got it today I would definitely ROI, in fact I would hit about a 60% ROI according to http://mining.thegenesisblock.com/

December is iffier and it depends on what kind of rate increases we see and how sustained they are.  The current network growth is not sustainable though (keep in mind we've gone from a 110 nm process to a 28 nm process in 10 months time, and the rise in hash rate reflects this).  The next stopgap half-node is 20 nm, and I don't think any fab is there yet for large scale, high yield production.

http://www.xbitlabs.com/news/other/display/20131022230815_TSMC_Shares_More_Details_Regarding_16nm_FinFET_and_20nm_Progress.html

Since HashFast is already working with TSMC, I would guess they'll be one of the first companies to produce on 20 nm as die shrinks (if the company survives that long, things are getting cutthroat).

Please cut the propaganda. Nobody will have access to lower than 28nm technology before 2015-2016 even and it will be damn hell expensive. Also there are many other companies that are already working with TSMC (Avalon, ASICMINER, KnC etc) so i don't understand why you praise HF so much. If they can get it then everyone will get it.
legendary
Activity: 3878
Merit: 1193
November 03, 2013, 03:27:05 PM
If I got it today I would definitely ROI, in fact I would hit about a 60% ROI according to http://mining.thegenesisblock.com/...
You mean this http://mining.thegenesisblock.com/a/89f1855c7e  Huh

That's if the miner is already delivered, which it isn't. This is what it looks like next month.

http://mining.thegenesisblock.com/a/9388e2430d
hero member
Activity: 784
Merit: 1000
November 03, 2013, 03:26:09 PM
...and I don't think any fab is there yet for large scale, high yield production.

I thought I read that Samsung was getting involved to do just that...
donator
Activity: 1218
Merit: 1079
Gerald Davis
November 03, 2013, 03:25:51 PM
Since HashFast is already working with TSMC, I would guess they'll be one of the first companies to produce on 20 nm as die shrinks (if the company survives that long, things are getting cutthroat).

There will be no 20nm miner by any vendor for quite a while.   20nm being available =/= 20nm cheaper than 28nm.   20nm will probably be available in volume next year (at 50% to 150% higher prices).  It generally takes 2-3 years before a new process node becomes cheaper than the prior one.  Maybe in late 2016 but 2017 seems more likely.
full member
Activity: 210
Merit: 100
November 03, 2013, 03:17:08 PM
you wont roi if you get it by end of year. even if you got it today you might not even roi. we had 1 chance of roi and thats why we bought it. shipping said late october. most the money would be coming in from then till now. and difficulty rate rising  tremendously makes it even less likly to roi.

If I got it today I would definitely ROI, in fact I would hit about a 60% ROI according to http://mining.thegenesisblock.com/...

You mean this http://mining.thegenesisblock.com/a/89f1855c7e  Huh

Shows you losing money.  What did you order & @ what cost?
legendary
Activity: 1484
Merit: 1005
November 03, 2013, 02:23:06 PM
you wont roi if you get it by end of year. even if you got it today you might not even roi. we had 1 chance of roi and thats why we bought it. shipping said late october. most the money would be coming in from then till now. and difficulty rate rising  tremendously makes it even less likly to roi.

If I got it today I would definitely ROI, in fact I would hit about a 60% ROI according to http://mining.thegenesisblock.com/

December is iffier and it depends on what kind of rate increases we see and how sustained they are.  The current network growth is not sustainable though (keep in mind we've gone from a 110 nm process to a 28 nm process in 10 months time, and the rise in hash rate reflects this).  The next stopgap half-node is 20 nm, and I don't think any fab is there yet for large scale, high yield production.

http://www.xbitlabs.com/news/other/display/20131022230815_TSMC_Shares_More_Details_Regarding_16nm_FinFET_and_20nm_Progress.html

Since HashFast is already working with TSMC, I would guess they'll be one of the first companies to produce on 20 nm as die shrinks (if the company survives that long, things are getting cutthroat).
Pages:
Jump to: