• 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 16nm: volume production Q2/3 2015

Orangey

Orangey

Orangey

Soldato
Joined
22 Aug 2008
Posts
8,338
http://fudzilla.net/home/item/36064-tsmc-volume-production-of-16nm-finfet-in-2h-2015

TSMC has announced that it will begin volume production of 16nm FinFET products in the second half of 2015, in late Q2 or early Q3.

For consumers, this means products based on TSMC 16nm FinFET silicon should appear in late 2015 and early 2016.
Click to expand...

I assume this is regular 16 (20 with fins) and not 16+. Previous forecasts told us to expect the latter about 6 months after, so we could be waiting until the middle of 2016 for a worthy process to make GPUs on.

Here is the TSMC roadmap article from April with the juicy details:

http://community.cadence.com/cadenc...speed-ahead-for-16nm-finfet-plus-10nm-and-7nm

The 16FF and 16FF+ technologies are "ready for prime time," according to Sun (left). He noted that the 16FF yield has already caught up with the 20nm planar (20SoC) process node. As a second-generation FinFET technology, he said, 16FF+ can provide an additional 15% die size reduction compared to 20SoC.

Liu said that TSMC plans 15 16FF tapeouts this year, and that compared to 20SoC, 16FF can provide a 40% performance increase at the same power consumption. 16FF+ allows an additional 15% performance increase. Volume production for the 16nm FinFET nodes is expected in 2015. "We are confident that our customers can use this [16nm] technology to produce mobile devices superior to those produced by IDMs," he said.
Click to expand...
 
Last edited:
16+ is still 20 with finfets. Pretty much any process you could ever name will be denser later in it's life than earlier on. When you first get the kit, it's quite literally like getting a recipe book. Spend 12 hours cooking at 700C then take out and cool for X hours before doing the next set of etches and cook further. Over time you perfect this recipe. Where earlier designs will have a set of design rules to leave 3nm extra space between certain bits on a die because while the equipment can theoretically do better, their recipe isn't perfected yet and to make sure yields are okay you leave a little wiggle room.

As the recipe improves the number of faults decreases and the wiggle room left in a design decreases. 28nm tsmc/glofo, 22nm Intel, it will all be capable of being a bit denser today than at launch. 5-15% denser is pretty much expected.

As they say, it's second gen finfet on 20nm, as such they've made some small adjustments, worked out how to get it just a little bit tighter together.

20nm, 16, 16+, it will all be fine for GPU's.

There is no such thing as a mobile process these days, what is mobile, you get 800Mhz single core things and octo cores, or a quads that will push 3Ghz at 20nm. Optimising for low leakage, higher efficiency is something mobile needs, and desktop, and high end performance.
 
andybird123 said:
No one really knows
Now that Apple have probably calmed down some it could be 20, but if they are waiting for 16 then it would mean end of 2015 release, so lets hope not
Click to expand...

Hmm , wouldn't it be a right faf for nvidia to tape out anything on 20nm if 16nm is round the corner. Are we even sure they aren't going to stick with the 28nm process for big maxwell ?
 
It isnt "around the corner" though, it is still 12 months away from volume production most of which will get eaten up by Apple again no doubt, 20 is only basically ready for them now
 
It's supposed to be very easy porting designs from 20 to 16 (as 16 is 20 with fins) so you have that to consider, Nvidia are way more reliant on shrinks with their big dies so you would expect them to use 20, but we haven't heard anything about them having orders in at TSMC or any kind of scuttlebutt other than the shipping info.
 
We've been on 28nm so long that it's possible they could go straight to 16nm finfet if they can get products out in Q4 2015.

That way we could have GTX 980 Ti '28nm' in Q1 2015 and they new cards in Q4 2015.

Obviously 20nm would be preferably if it brings lower power use / extra performance, but they have left it so long on 28nm it wouldn't surprise me if they just skip 20nm altogether.
 
MasterOC said:
Hmm , wouldn't it be a right faf for nvidia to tape out anything on 20nm if 16nm is round the corner. Are we even sure they aren't going to stick with the 28nm process for big maxwell ?
Click to expand...

The original design for fullfat maxwell would be infeasible on 28nm.

humbug said:
Didn't Lisa Sue say she was aiming to be on 16nm as soon as possible. by 2016?
Click to expand...

Would have to dig info out again but don't think she gave an exact timeframe IIRC was ramp upto 20nm 2015 and finfets as soon as possible after, I believe she also mentioned that they weren't planning to use TSMC for finfets anyhow though would have to double check that.
 
They must be using TSMC because GF's process is mobile-oriented. Only other place would be Intel... I doubt they'd welcome AMD's custom.
 
Orangey said:
They must be using TSMC because GF's process is mobile-oriented. Only other place would be Intel... I doubt they'd welcome AMD's custom.
Click to expand...

I do believe even Intel have been using TSMC for chip production. Unless I was dreaming of course.

Edit:

Just had a look for sanity reasons and can't see anything, so assume the latter :(
 
Gregster said:
I do believe even Intel have been using TSMC for chip production. Unless I was dreaming of course.

Edit:

Just had a look for sanity reasons and can't see anything, so assume the latter :(
Click to expand...

Not entirely wrong Gregster - IIRC Intel has licensed the latest Atom cores to a Chinese SOC design company who will be using TSMC.
 
You must log in or register to reply here.
Back
Top Bottom