[wccftech] TSMC Begins 20nm Volume Production

[ZomRoxx]

Apparently TSMC has now figured it out and begins production. Still don't think that maxwell will come earlier though. Still wanted to share it, you are free to guess when the new gpus will be released, but it's all just based on nothing

 

"Its about damn time. Volume production at the 20nm Node for the entire Industry has officially begun over at TSMC. Now don’t get me wrong, TSMC has been working on the 20nm node since early February but that work was almost exclusively prioritized for Apple and the creation of its 20nm SoCs. Now they have finally begun full fledged volume production for the rest of the Industry – and that means Nvidia and AMD too.

Read more: http://wccftech.com/tsmc-begins-20nm-volume-production-gpus-node/#ixzz33hkNTKP3"

[Dylan522p]

This is fuckineg stupid. They said they had 20nm mass in February and saying 20nm volume means the exact same thing. In reality their yields are shit so they cant make larger dies.

[qwertywarrior]

20nm had issues like higher costs and long design cycles .it required new techniques

20nm  came a bit late .I bet It will be short lived node and will quickly be followed by 16nm and 14nm

[Raid]

>WCCFtech

any other sites reporting on this?

Only other source i can find is a Digitimes article from last week but they require paid subscription to access the article.

[Glenwing]

Oh good, just in time... 

[ZomRoxx]

Unfortunately I only found the one article on WCCFtech, so not exactly the most reliable source

[Accursed Entity]

As long as I can buy a GTX 880 Q2 2015 I'll be happy.

[VSG]

No source cited as usual, sorry I don't buy it. It may well be true but I don't trust anything from them without a source anymore especially with so many things wrong even recently.

[Bagmup]

No source cited as usual, sorry I don't buy it. It may well be true but I don't trust anything from them without a source anymore especially with so many things wrong even recently.

Exactly, the same guys who said DC wouldn't be released until September, even came on here and defended the article.

I'll wait for a reputable source thanks...

[One Who Craves Souls]

I'm tired of wccftech rumors... 

[Dylan522p]

20nm had issues like higher costs and long design cycles .it required new techniques

20nm came a bit late .I bet It will be short lived node and will quickly be followed by 16nm and 14nm

16 and 14 nm are just marketing of 20nm FinFet.

[Kamina]

I wonder if they will ever do tri-gate like Intel as they keep shrinking it.

[Equlix]

so not exactly the most reliable source

Why post it then?

[Plasma]

WCCFtech should be blocked at the domain lv so no one can post there unvalidated post.

​im tired of the industry pushing smaller die over quality and cost, imagine how much better they could be if they sat on a die for 3 years instead of rushing chips out the door.

[Kamina]

WCCFtech should be blocked at the domain

I agree. They steal sources and claim as their own. They even steal benchmarks which is sad.

[Equlix]

I agree. They steal sources and claim as their own. They even steal benchmarks which is sad.

Wccf is not the problem here, people not understanding what constitutes news is the problem. I could go and make a blog myself and report on it that intel skylake and nvidia volta will be out next week and that valve is letting EA make Half-Life 2 Episode 3 but without a source it's just bullshit.

What's happening here is people are taking this shit and pasting it everywhere and thinking that labeling there post with BEWARE WCCFTERK DANGER WARNING will make it ok to post knowing full well that its bullshit.

[GPUXPert]

Can't wait for Pirate Islands R9 300 series. I'm getting tired of Nvidia and some of the stupid choices they've been making for the past year or two.

[Scionyde]

 

​im tired of the industry pushing smaller die over quality and cost, imagine how much better they could be if they sat on a die for 3 years instead of rushing chips out the door.

 

Would it really be that impressive? We're not too far off from 3 years between die shrinks already. With Intel, we've had Ivy Bridge -> Haswell -> Haswell refresh to the point where we won't see Broadwell until Q4 2014 / Q1 2015 (~2 years, 8 months on 22 nm). And with Nvidia and AMD, we've been on 28 nm for close to 2 and a half years, with no firm release date on when we'll see 20 nm, so there could easily be 3 years between 28 nm and 20 nm

[jezza39]

Would it really be that impressive? We're not too far off from 3 years between die shrinks already. With Intel, we've had Ivy Bridge -> Haswell -> Haswell refresh to the point where we won't see Broadwell until Q4 2014 / Q1 2015 (~2 years, 8 months on 22 nm). And with Nvidia and AMD, we've been on 28 nm for close to 2 and a half years, with no firm release date on when we'll see 20 nm, so there could easily be 3 years between 28 nm and 20 nm

 

There's only so many transistors that fit in a die, shrinking the process node (the transistors tehmselves) means you can physically do more, more cores etc. The alternative is making the die area physically bigger like what NVidia has done with GK110 vs GK104/106.

[Scionyde]

There's only so many transistors that fit in a die, shrinking the process node (the transistors tehmselves) means you can physically do more, more cores etc. The alternative is making the die area physically bigger like what NVidia has done with GK110 vs GK104/106.

 

Yup. I just meant that I wasn't sure we were truly "rushing" die shrinks at the expense of quality, as somebody else posted.

[carolkarine]

There's only so many transistors that fit in a die, shrinking the process node (the transistors tehmselves) means you can physically do more, more cores etc. The alternative is making the die area physically bigger like what NVidia has done with GK110 vs GK104/106.

 

which means in 20nm goes belly up for another gen, Nvidia is out in the cold! their GK110 die is simply as big as it gets. (one could argue hawaii is a better chip because it performs very close to fully fledged GK110 but is significantly smaller) AMD could do another architecture on the same size die as GK110, and they'd be able to pack in more of their "cores" under a new arch.

[jezza39]

Yup. I just meant that I wasn't sure we were truly "rushing" die shrinks at the expense of quality, as somebody else posted.

 

Yeah sorry I realised after I should've quoted the other guy.

 

which means in 20nm goes belly up for another gen, Nvidia is out in the cold! their GK110 die is simply as big as it gets. (one could argue hawaii is a better chip because it performs very close to fully fledged GK110 but is significantly smaller) AMD could do another architecture on the same size die as GK110, and they'd be able to pack in more of their "cores" under a new arch.

 

Exactly, without a die shrink, NVidia aren't going anywhere. Hawaii trumps GK110 in performance/area. AMD could but I don't recall them ever going with a super huge die area and it seems we are close to 20nm so I doubt they'd bother (would be awesome though if the rumoured R9 295X did show up with a bigger die size). Surely Global Foundries are also close to 20nm too, so 20nm parts should come in a rush.