GF Achieves 14nm FinFET Technology Success for Next-Generation AMD Produ

[jpiniero]

AMD APUs on 14 nm in 2016 too? What APUs are those? Zen based APUs are not supposed to arrive until 2017 IIRC.

PS4/XBO probably.

 

[dark zero]

Bristol Ridge is a 28nm APU.
Am I missing something?

Unless... they ditched the 28 nm version due costs.
Also those APU could be Stoney Ridge or even K12.

Finally... I am thinking...
What happens if AMD decides to put Bristol, Summit and Stoney at 14 nm, but with different features?

 

[Fjodor2001]

PS4/XBO probably.

That would be nice. Should allow for even more slim and power efficient versions of those consoles.

But couldn't it be a 14 nm Bristol Ridge APU too?

 

[JDG1980]

PS4/XBO probably.

Can't rule out FinFET for the new Nintendo console(s), either. After all, those aren't expected to hit mass market until late 2016-early 2017.

 

[jpiniero]

That would be nice. Should allow for even more slim and power efficient versions of those consoles.

But couldn't it be a 14 nm Bristol Ridge APU too?

IIRC Bristol Ridge is definitely 28. They would only be doing the two consoles because they are contractually obligated to do so. I don't think they would be able to make the mask costs back on Bristol Ridge if they did shrink it down.

 

[DrMrLordX]

Bristol Ridge was announced as a 28nm product in the initial AM4/FM3 leaks, for whatever reason. There may be more to it than Carrizo (different power management, no on-die FCH, no HDL) or there may not. Possibly a new stepping too.

So your guess is as good as mine as to what 14nm LPP APUs AMD hopes to launch in 2016, if any.

 

[dark zero]

Bristol Ridge was announced as a 28nm product in the initial AM4/FM3 leaks, for whatever reason. There may be more to it than Carrizo (different power management, no on-die FCH, no HDL) or there may not. Possibly a new stepping too.

So your guess is as good as mine as to what 14nm LPP APUs AMD hopes to launch in 2016, if any.

Bristol Ridge was announced initially along Stoney Ridge to be on FM3 at 28 nm. But then all the info related to them were gone.

 

[tential]

after fuji I dont know if i wanna :/

Wow, even you don't like Fiji? That's saying a lot. I'm pretty pissed off about it. I was confident it would go head to head with the 980Ti.
Fury stands in an awkward spot of eh.... $550 just seems too much.
Nano is $650.... like it's the price of a 980Ti.... even more expense than some 980Ti's on sale.

Fiji puts me in an AWKWARD position because I want a $650 card, but my option is Fury X.... and then I'm unable to play the games I want. And then my other option is 980Ti.... and then I'm unable to use Freesync.

 

[dark zero]

Kill me, but Fiji was an awful chip. However HBM didn't dissapointed me and is definately the future.

 

[TechGod123]

Kill me, but Fiji was an awful chip. However HBM didn't dissapointed me and is definately the future.

Awful? Lol. Some of you need to stop being so melodramatic.

 

[tential]

Awful? Lol. Some of you need to stop being so melodramatic.

In comparison the the GTX980Ti, in the current games out, it is awful.

It may be fully utilized in some future with DX12, but then.... there will be a new flagship GPU out. 980Ti seems to walk all over it though, and we're talking before OCing.

Or you can say Fiji is overpriced.

Either way, it just didn't deliver what we thought it would.
No OC headroom.
Power consumption is still high.
What did Fiji really do other than add HBM?

 

[TechGod123]

In comparison the the GTX980Ti, in the current games out, it is awful.

It may be fully utilized in some future with DX12, but then.... there will be a new flagship GPU out. 980Ti seems to walk all over it though, and we're talking before OCing.

Or you can say Fiji is overpriced.

Either way, it just didn't deliver what we thought it would.
No OC headroom.
Power consumption is still high.
What did Fiji really do other than add HBM?

Lmao, what?? So a few FPS difference between the Ti and the X makes the X a bad chip? There are some games that outright suck on AMD but there are some that work really well. Know what it bought? A nice stock AIO built in for cooling. Oh and I suppose it finally also bought some much needed competition...

 

[Fjodor2001]

http://wccftech.com/amd-r9-fury-x-performance-ahead-nvidia-980-ti-latest-drivers/

AMD R9 Fury X Leaps Ahead Of Nvidia GTX 980 Ti With The Latest Windows 10 Drivers – All AMD GPUs Get A Sizable Performance Boost

Before:

After:

Software: Windows 10 64-bit
Drivers: NVIDIA: 358.50 WHQL
AMD : Catalyst 15.9.1 Beta

Let’s look at how drastically the numbers changed thanks to this move to the latest Windows 10 drivers.

Before – @ 1440p And 4K

On the older test suite at 1440p we find that Nvidia’s GTX 980 Ti is ahead of the R9 Fury X by approximately 8% ( 118/109 x 100). We also find that Nvidia’s $469 GTX 980 is ahead of AMD’s $399 R9 390X by approximately 2% and finally AMD’s $319 R9 390 leads Nvidia’s $319 GTX 970 by 1% and is in an effective tie with Nvidia’s previous $699 flagship the GTX 780 Ti. In the midrange we find that the AMD $189 R9 285 is leading the $189 GTX 960 by approximately 4% (55/53 x100).

At 4K on the older drivers we find that the performance gap between all graphics cards widens in the favor of the AMD cards when compared to 1440p results. The Nvidia GTX 980 Ti and AMD R9 Fury X come in at a dead head. We also see a dead tie between the GTX 980 and the R9 390X, with the R9 390 coming out ahead of the GTX 970 and the GTX 780 Ti. We’re less inclined to look at how mid-range GPUs fair at this resolution, as it’s far too demanding for this class of graphics cards realistically speaking. However, we see the R9 285 leading the GTX 960 again here as well.

After – @ 1440p And 4K

At 1440p and after the update we see the 980 ti’s 8% lead shrink to 0% putting it in a dead heat with AMD’s R9 Fury X. We also see the R9 390X now ahead of the GTX 980 by 3% (68/66 x 100) prior to it being 2% behind. We also see AMD’s R9 290X leading Nvidia’s GTX 970 by nearly 9% (62/57 x100), a 9% improvement over the previous standing in which both cards were tied.

More fascinating is how the R9 290X now compares to the GTX 780 Ti. The R9 290X was the flagship from AMD back when it launched in late 2013 for $550 and Nvidia answered back with the $700 GTX 780 Ti which was regarded as the faster card at the time. Today the R9 290X is leading the GTX 780 Ti by 5%, a card which debuted for a 27% price premium. The difference is even more shocking when we look at the R9 290 and the GTX 780. Cards which sold for $400 and $500 respectively for the majority of their lifetimes. The R9 290 now leads the more expensive GTX 780 by 16% (57/49 x100).

With the latest Windows 10 drivers at 4K, the R9 Fury X jumps ahead of the GTX 980 ti by 5% (84/80 x100). The R9 390X secures its position ahead of the GTX 980 as well. And we see the R9 290X as well as the R9 290 this time surpassing the GTX 970 and the GTX 780 Ti. In fact the performance of all AMD graphics cards improves significantly from the previous drivers, including the mid range and even the entry level offerings.

So now maybe GTX 980Ti is "awful"?

 

[dark zero]

http://wccftech.com/amd-r9-fury-x-performance-ahead-nvidia-980-ti-latest-drivers/

AMD R9 Fury X Leaps Ahead Of Nvidia GTX 980 Ti With The Latest Windows 10 Drivers – All AMD GPUs Get A Sizable Performance Boost

Before:

After:

So now maybe GTX 980Ti is "awful"?

Nope. Nvidia Maxwell knew that they were going to their own limits and the GDDR5 was maxxed out.

However HBM had space to improve, however they delivered the wrong chip.

 

[Phynaz]

Intel, TSMC, Samsung - shipping tens of millions of 14nm CPUs every month.

Global Foundries - shipping press releases.

 

[Erenhardt]

Kill me, but Fiji was an awful chip. However HBM didn't dissapointed me and is definately the future.

How is Fiji perf/clock vs GM110? We see 1050 Mhz part trading blows with 1300 mhz one.
How is double precision on those maxwell parts?

Fury is a heck of a card with all the new and shiny tech inside. A true pioneer from starving underdog.
Sadly most people will not reward their efforts because of the agenda and social pressure in circles. Some openly admitted they wouldn't use amd cards even if those were free.

A though spot to be in for a company.
A sad spot to be in as a consumer.

 

[ThatBuzzkiller]

Intel, TSMC, Samsung - shipping tens of millions of 14nm CPUs every month.

Global Foundries - shipping press releases.

Correction, TSMC has shipped ZERO 14nm silicon so far ...

 

[Arachnotronic]

Correction, TSMC has shipped ZERO 14nm silicon so far ...

14/16nm class silicon. Don't be dense.

 

[ThatBuzzkiller]

14/16nm class silicon. Don't be dense.

More like 20nm with FinFETs and who are you to call me dense ?

 

[Arachnotronic]

More like 20nm with FinFETs and who are you to call me dense ?

If TSMC's 16nm is "20nm with FinFETs," the same applies to Samsung's 14nm as well. They have identical minimum metal pitches.

 

[railven]

http://wccftech.com/amd-r9-fury-x-performance-ahead-nvidia-980-ti-latest-drivers/

AMD R9 Fury X Leaps Ahead Of Nvidia GTX 980 Ti With The Latest Windows 10 Drivers – All AMD GPUs Get A Sizable Performance Boost

Before:
<snip>

After:

<snip>

So now maybe GTX 980Ti is "awful"?

Russian pretty much hit the nail on this one, they changed the game line up. removed games that heavily favored NV which adjusted the average. But, check out the scores for games that didn't change:

Before:

After:

Before:

After:

Even in AMD favored games:
Before:

After:

Before:

After:

Only thing that changed was the average by removing some NV favored games.

 

[ThatBuzzkiller]

If TSMC's 16nm is "20nm with FinFETs," the same applies to Samsung's 14nm as well. They have identical minimum metal pitches.

No, Samsung's 14nm transistor gate pitch is actually smaller than TSMC's so called "16nm" transistor gate pitch ...

And if we're going to go around marginalizing whatever GF does then we might as well do the same for TSMC because it's not like we have any direct proof that there making any 16/14nm chips going by their financial reports ...

 

[Arachnotronic]

No, Samsung's 14nm transistor gate pitch is actually smaller than TSMC's so called "16nm" transistor gate pitch ...

Transistor gate pitch isn't the same as minimum metal pitch. Yes, Samsung has a slightly tighter gate pitch relative to TSMC 16nm, but as we've seen in shipping silicon the difference leads to a <10% difference in die size.

And if we're going to go around marginalizing whatever GF does then we might as well do the same for TSMC because it's not like we have any direct proof that there making any 16/14nm chips going by their financial reports ...

https://www.chipworks.com/about-chipworks/overview/blog/a9-is-tsmc-16nm-finfet-and-samsung-fabbed

 

[ThatBuzzkiller]

Transistor gate pitch isn't the same as minimum metal pitch. Yes, Samsung has a slightly tighter gate pitch relative to TSMC 16nm, but as we've seen in shipping silicon the difference leads to a <10% difference in die size.

https://www.chipworks.com/about-chipworks/overview/blog/a9-is-tsmc-16nm-finfet-and-samsung-fabbed

Gate pitches ARE what defines the dimensions of the transistor, am I not correct ? Just because the difference is relatively small doesn't validate your claim that Samsung's 14nm node is just 20nm with FinFETs ...

As for revenue breakdown, I would like to know what their 16nm node nets ALONE ...

 

[tential]

Lmao, what?? So a few FPS difference between the Ti and the X makes the X a bad chip? There are some games that outright suck on AMD but there are some that work really well. Know what it bought? A nice stock AIO built in for cooling. Oh and I suppose it finally also bought some much needed competition...

How did it bring competition? The 980Ti is superior, and came out first and is available at a lower price right now than the Fury X?

The Fury X Brought lower performance at a higher price point....

It didn't "bring" competition at all. It didn't change anything Nvidia did.

OC your 980Ti just a little bit, and you're far past what the Fury X is capable of giving you in the best scenario under LN2.