Differences between TSMC 16nm FF+ (Pascal) and GloFo 14nm LPP (Polaris)?

[96Firebird]

It's ready, they're just waiting for VR, or DX12, or back to school, or Christmas 2016, or the next pass-by of Halley's Comet... :sneaky:

 

[Flapdrol1337]

Is it actually confirmed somewhere which manufacturing process amd and nvidia chose for the gpu's?

With the polaris amd just called it "finfet", ambiguous.

 

[IllogicalGlory]

Well, one process has produced working silicon for AMD, producing an impressive result offered by nothing else that's been demo'd so far, and the other has produced what appears to be nothing for NVIDIA, except a need to create a fake stage mockup. Don't get where all the smugness is coming from.

But I guess it's coming really soon and will be totally awesome when it does, right? Right.

 

[ViRGE]

Is it actually confirmed somewhere which manufacturing process amd and nvidia chose for the gpu's?

With the polaris amd just called it "finfet", ambiguous.

NVIDIA hasn't announced anything AFAIK, but TSMC 16nm is a safe bet since it's their traditional partner.

As for AMD, they have announced that they will be using both GloFo 14nm and TSMC 16nm.

http://www.anandtech.com/show/9886/amd-reveals-polaris-gpu-architecture/3

 

[DA CPU WIZARD]

Moving goalposts 101. There are lots of things that are ready today that you can't buy.

Obviously, but what evidence do we have to prove this is the case for AMD?

 

[DA CPU WIZARD]

Well, one process has produced working silicon for AMD, producing an impressive result offered by nothing else that's been demo'd so far, and the other has produced what appears to be nothing for NVIDIA, except a need to create a fake stage mockup. Don't get where all the smugness is coming from.

But I guess it's coming really soon and will be totally awesome when it does, right? Right.

Could the fact that the PX 2 is expected to release later this year (rather than mid this year for both AMD and Nvidia GPU) have had an impact on the availability of 16nm silicon?

 

[3DVagabond]

Could the fact that the PX 2 is expected to release later this year (rather than mid this year for both AMD and Nvidia GPU) have had an impact on the availability of 16nm silicon?

They didn't even have a GPU to show (nVidia) where AMD has working silicon. AMD has also had a consumer product on the market for ~7mos. with interposer/2.5D stacking where nVidia has nothing but a mock up.

 

[tviceman]

Moving goalposts 101. There are lots of things that are ready today that you can't buy.

....just ... wow.

It's ready, they're just waiting for VR, or DX12, or back to school, or Christmas 2016, or the next pass-by of Halley's Comet... :sneaky:

LOL. Perfect answer. AND fans here are the worst. They'll defend a flat out false statement with any straw they can pluck.

 

[AnandThenMan]

....just ... wow.



LOL. Perfect answer. AND fans here are the worst. They'll defend a flat out false statement with any straw they can pluck.

Do you mean AMD fans? Or what are you trying to say here?

 

[Headfoot]

If both AMD and nVidia both end up using TMSC 16nm for their big GPUs, that ought to settle it.

 

[Dribble]

As for AMD, they have announced that they will be using both GloFo 14nm and TSMC 16nm.

http://www.anandtech.com/show/9886/amd-reveals-polaris-gpu-architecture/3

Which if true would suggest AMD think TSMC is better. We can guess they have no choice but to use some GloFo 14nm to fulfil the wafer contract, so why use any TSMC as that means designing chips for a whole other node, an expensive and time consuming process? If GloFo could make the big gpu's surely that would be all they need?

 

[Arachnotronic]

Which if true would suggest AMD think TSMC is better. We can guess they have no choice but to use some GloFo 14nm to fulfil the wafer contract, so why use any TSMC as that means designing chips for a whole other node, an expensive and time consuming process? If GloFo could make the big gpu's surely that would be all they need?

Pretty much my read on the situation. They will use GloFo for the higher volume/easier to build parts to satisfy WSA and use TSMC for higher end/bigger GPUs.

 

[Glo.]

Isn't that 16 nm is used only to produce CPUs?

All of next gen GPUs are supposed to be on 14 nm.

 

[ShintaiDK]

Isn't that 16 nm is used only to produce CPUs?

All of next gen GPUs are supposed to be on 14 nm.

CPUs are 14nm. Its GPUs that will be a mix. Its all done for the WSA. Else its hard to imagine AMD fabbing anything at GloFo.

 

[3DVagabond]

Which if true would suggest AMD think TSMC is better. We can guess they have no choice but to use some GloFo 14nm to fulfil the wafer contract, so why use any TSMC as that means designing chips for a whole other node, an expensive and time consuming process? If GloFo could make the big gpu's surely that would be all they need?

Since AMD already designs for their APU's at GloFo, it might not really be more expensive to fab GPU's there as well. It would be better for AMD to do everything at GloFo, but if their process isn't as good for large and/or high power designs then no point in sacrificing performance, I guess. It doesn't necessarily mean that TSMC's process is superior overall though.

 

[railven]

Since AMD already designs for their APU's at GloFo, it might not really be more expensive to fab GPU's there as well. It would be better for AMD to do everything at GloFo, but if their process isn't as good for large and/or high power designs then no point in sacrificing performance, I guess. It doesn't necessarily mean that TSMC's process is superior overall though.

I would take it more as AMD is finally learning from their mistakes. How many times has GloFlo burned them already?

 

[Silverforce11]

Wasn't it published recently that it was a misunderstanding with the press, that both Polaris chips were designed for 14nm ff? Just saw it in an article last week.

 

[railven]

Wasn't it published recently that it was a misunderstanding with the press, that both Polaris chips were designed for 14nm ff? Just saw it in an article last week.

So all of Polaris is GloFlo bound?

Welps, my confidence is basically zilch.

 

[Arachnotronic]

So all of Polaris is GloFlo bound?

Welps, my confidence is basically zilch.

Nope...

As for RTG’s FinFET manufacturing plans, the fact that RTG only mentions “FinFET” and not a specific FinFET process (e.g. TSMC 16nm) is intentional. The group has confirmed that they will be utilizing both traditional partner TSMC’s 16nm process and AMD fab spin-off (and Samsung licensee) GlobalFoundries’ 14nm process, making this the first time that AMD’s graphics group has used more than a single fab. To be clear here there’s no expectation that RTG will be dual-sourcing – having both fabs produce the same GPU – but rather the implication is that designs will be split between the two fabs. To that end we know that the small Polaris GPU that RTG previewed will be produced by GlobalFoundries on their 14nm process, meanwhile it remains to be seen how the rest of RTG’s Polaris GPUs will be split between the fabs.

http://www.anandtech.com/show/9886/amd-reveals-polaris-gpu-architecture/3

 

[tviceman]

So all of Polaris is GloFlo bound?

Welps, my confidence is basically zilch.

I'm not writing off either camp based on their choice of fab. Samsung's first gen was ever so slightly more dense, while TSMC's was more efficient under high drain. Samsung just announced their second-gen finfet process, while TSMC has their finfet+ process, both of which are improved versions over what we've seen so far.

Whoever comes out on top, it won't be because of the fab selection.

 

[Glo.]

Wasn't it published recently that it was a misunderstanding with the press, that both Polaris chips were designed for 14nm ff? Just saw it in an article last week.

That is my understanding also. And the Anandtech article is from earlier time, when the misunderstanding could occur.

Raja only speaks as if all next gen GPUs were produced on 14 nm process.

http://wccftech.com/amd-unveils-polaris-11-10-gpu/ here says that Raja unveiled 14 nm GPUs. Small and big die.

I suppose it is right then: 14 nm GPUs, 16 nm CPUs.

 

[maddie]

Nope...



http://www.anandtech.com/show/9886/amd-reveals-polaris-gpu-architecture/3

Then there is this from the company itself. Not a single mention of 16nm production for the GPUs.

AMD’s Polaris architecture-based 14nm FinFET GPUs deliver a remarkable generational jump in power efficiency

http://www.amd.com/en-us/press-releases/Pages/amd-demonstrates-2016jan04.aspx

 

[Silverforce11]

As I said, it was made public that the press misunderstood, they do have 14/16nm designs, but Polaris is designed for 14nm ff. Might be other stuff, APU, custom Arm SOCs on 16nm ff?

 

[railven]

Or someone at AMD inadvertently let the cat out of the bag, ie "Polaris on 16nm ff" was suppose to be their ace when big Pascal shows up.

 

[Arachnotronic]

As I said, it was made public that the press misunderstood, they do have 14/16nm designs, but Polaris is designed for 14nm ff. Might be other stuff, APU, custom Arm SOCs on 16nm ff?

No. AMD is working on 16FF+ GPUs, just poke around LinkedIn for a bit.