• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

TSMC starts mass production of 20 nm chips.

Boomstick777

Boomstick777

Boomstick777

Soldato
Joined
2 Jan 2012
Posts
12,414
Location
UK.
Not long now...

Taiwan's TSMC has begun mass production of 20-nanometer chips before the previously scheduled time. According to the company's vice president C. C. Wei, TSMC began the production of large quantities of chips in the above production process a month earlier than it was originally planned.
Read more at http://www.decryptedtech.com/news/tsmc-starts-mass-production-of-20-nm-chips#b4msbPugWEF38w4a.99
Click to expand...

http://www.decryptedtech.com/news/tsmc-starts-mass-production-of-20-nm-chips
 
Last edited:
And none of the first 5 products being mass produced are (apparently) GPUs. (Not to say there won't be any 20nm desktop GPUs soon as there are smaller runs of stuff also).
 
Got nothing on these:

cKaRVhu.jpg
 
MjFrosty said:
I think give it till March and if there still isn't any news, I doubt we'll see high end 20nm Maxwell. Maybe next year if not at all lol.
Click to expand...

Proper full fat Maxwell is almost certainly not coming on 20nm. Should see GTX680 type cards (tier position wise) appearing well inside this year though, probably sooner than later.
 
Why has this taken so long?

Is this just a crap node for yields, or tooling costs, or what?
 
Apple A8 Chips.

December 2013.

News out of Taipei today states that Taiwan Semiconductor Manufacturing Co (TSMC) announced yesterday that it's to start high-volume production of its most advanced 20-nanometer (nm) chips next month. The move would make TSMC the world's leading contract chipmaker supplying 20nm chips. Analysts say that mass 20nm chip production could put TSMC in a position to dethrone Samsung as the supplier of Apple's next-generation A8 chip next year. The only question remaining is whether Apple's iPhone 6 will use TSMC's 20nm or upcoming 16nm chip which begins production late this year.
Click to expand...
 
Tute said:
Why has this taken so long?

Is this just a crap node for yields, or tooling costs, or what?
Click to expand...

Along with the above about Apple sub 28/22nm requires a big change to how the lithography is done - http://www.extremetech.com/computin...hography-technique-to-push-moores-law-to-20nm it requires a whole new approach to many aspects of production. Which is why I don't think nVidia would produce Maxwell on 28nm unless they are really desperate as it introudces a lot more costs with having to create masks for both techniques and debugging, etc. done on the 28nm side will be of little use to 20nm production, etc. so they'd essentially be working on 2 seperate chips unlike in the past where they could simply (relatively speaking) rescale up/down as needed.
 
Boomstick777 said:
This one work? http://www.decryptedtech.com/news/tsmc-starts-mass-production-of-20-nm-chips

Got a feeling this could be Nvidia's K1 SOC and not GPU's.. They will prob come in Q4..
Click to expand...
Extremely unlikely it will be a K1 SOC. TSMC said “TSMC expects that in final quarter of this year, production of 20-nanometer chips will be responsible for 20 percent of the revenue,

There is no way a chip as rubbish a Tegra is going to make up that much revenue. If anything it will be one of the new Apple SOC's.
 
Boomstick777 said:
This one work? http://www.decryptedtech.com/news/tsmc-starts-mass-production-of-20-nm-chips

Got a feeling this could be Nvidia's K1 SOC and not GPU's.. They will prob come in Q4..
Click to expand...

K1 is already confirmed for 28nm, as with Tegra 3 when they went five core at 40nm when the industry waited a few months more and beat it easily with dual cores at 28nm... Nvidia appears to be making the same mistake. Late highly optimised 28nm part, buy a design win, make little profit, lose future designs as you end up VERY late to 20nm when everyone else went early to 20nm. In mobile power/diesize/cost wins, full stop, being late to a new node is completely stupid, Nvidia did it the first time when everyone predicted it was stupid, and T3 was deemed a bit of a joke due to it's timing, why they are doing it again with the 28/20nm switch I don't know.

K1 a year earlier(and that was fully possible) fine, but it might not even beat 20nm chips to market, the 2h 2014 Denver version will almost certainly launch after 20nm Apple parts are shipping in devices.......

Rroff said:
Proper full fat Maxwell is almost certainly not coming on 20nm. Should see GTX680 type cards (tier position wise) appearing well inside this year though, probably sooner than later.
Click to expand...

Been saying it for a while, because of the pretty drastic change in transistors/layout/electrical properties of 16nm, I do wonder how much financial sense there is in a 680/7970 at 20nm, then waiting for 16nm for the big ass cores. Due to power I think 16nm/finfet is essential for a 290/titan replacement. While 680/7970 replacements are fine at 20nm, you would really want to re-do them for 16nm which as you say is a fairly large cost and time sink. The other simple issue is going to be wafer cost vs benefit of the process. Huge increase in wafer costs, probably iffy yields and not great power gains means the actual value of a upper midrange part at 20nm could mean they either end up pretty slow(due to lack of power budget limiting diesize/transistor count for a midrange priced part), or very expensive/lower yield/power hungry for a midrange part.

Where the numbers end up, doing it costs X amount of engineers and Y amount of money and will make Z amount of profit, who knows. More a case of time spent doing it vs skipping it and getting the 16nm parts ready to go from day one.

Pottsey said:
Extremely unlikely it will be a K1 SOC. TSMC said “TSMC expects that in final quarter of this year, production of 20-nanometer chips will be responsible for 20 percent of the revenue,

There is no way a chip as rubbish a Tegra is going to make up that much revenue. If anything it will be one of the new Apple SOC's.
Click to expand...

The quality of a chip has zero bearing on TSMC revenue. A wafer costs X amount, the amount you sell chips you get off the wafer is entirely different and doesn't effect TSMC in the slightest.

20% of their revenue coming from 20nm by the end of 2014 would likely mean somewhere between 5-10% of their production, which for TSMC is a really very significant amount.
 
The quality of the chip does have bearings as if the quality is bad there will be no or few design wins which in turn mean no production demand. K1 has too many flaws with it to warrant a high production of chips. K1 will be lucky to get around 2% market share. So I don’t see it being the SOC TSMC talked about.
 
Again that fails to have any bearing on anything being discussed. TSMC sell wafers, they have X number of wafers, it absolutely doesn't matter who they sell them to. They don't serve one customer or produce one type of chip.

The TSMC statement says how much revenue they expect to come from 20nm, not a singular product. The statement that TSMC expects 20% of their revenue to come from 20nm has nothing at all to do with what chips are made, just how many wafers they will be processing by that time.
 
You must log in or register to reply here.
Back
Top Bottom