AMD 14nm FinFET Polaris GPU Size Leaks Out – 232mm² Large Die

[-BirdiE-]

27 minutes ago, patrickjp93 said:

Well yes. We've only gotten smaller since the VooDoo cards. God do you people pay attention? /s

My point was that GPUs (the whole card) do not shrink 1:1 with manufacturing process of the die itself. And that expecting a Polaris card to come out that has the same power, but half the size of the Nano is pretty unrealistic.

[-BirdiE-]

This post was an accident!

[patrickjp93]

6 minutes ago, -BirdiE- said:

My point was that GPUs (the whole card) do not shrink 1:1 with manufacturing process of the die itself. And that expecting a Polaris card to come out that has the same power, but half the size of the Nano is pretty unrealistic.

The card will be as small as an iPhone 6S Plus. Calling it now. /s

[Agosto]

Well, considering the jump we had with the transition from 40nm to 28nm, I wouldn't expect a 232mm^2 Polaris chip to be much more powerful than a 596mm^2 Fiji one.

I'm assuming that the 232mm^2 one is Polaris 11, not 10, since the image shows a very big difference between the two - and P11 can't be ~800mm^2

[-BirdiE-]

Just now, patrickjp93 said:

The card will be as small as an iPhone 6S Plus. Calling it now. /s

Alright. We shall make a wager. IF AMD comes out with a card as powerful as the Nano but is half the width (or less), I will start a news thread titled "I was EPICLY wrong" and  use this thread as a source. If not, you will. lol

[Qub3d]

Imagine this: a 3-gpu single-slot card. 

[Dabombinable]

Just thinking of APU now, exactly how would they be impacted with the Zen CPU+iGPU at 14nm? And what would it be like with HBM/HBM2 built into it as L4 cache/vRAM? Something like that from what I can understand of the transition from 28nm CPU/iGPU should net massive performance boosts that should in theory give you near Haswell CPU performance with well over the performance of Intel iGPU+the ability to make better use of OpenCL. I'm definately waiting to see what AMD comes up with before deciding whether or not to down/side grade this desktop, sell my laptop+a lot of computer bits and get a Zen APU based laptop...because it'd kick the crap out of my GTX 650ti OC 2GB+i5 4440, which is a good combo still.

[patrickjp93]

57 minutes ago, -BirdiE- said:

Alright. We shall make a wager. IF AMD comes out with a card as powerful as the Nano but is half the width (or less), I will start a news thread titled "I was EPICLY wrong" and  use this thread as a source. If not, you will. lol

... 

I even used sarcasm tags...

[-BirdiE-]

Just now, patrickjp93 said:

... 

I even used sarcasm tags...

 

I was wondering why everyone kept using "/s"

[Dabombinable]

Just now, patrickjp93 said:

... 

I even used sarcasm tags...

I think the sarcasm went over his head and out the window.

[patrickjp93]

19 minutes ago, Dabombinable said:

Just thinking of APU now, exactly how would they be impacted with the Zen CPU+iGPU at 14nm? And what would it be like with HBM/HBM2 built into it as L4 cache/vRAM? Something like that from what I can understand of the transition from 28nm CPU/iGPU should net massive performance boosts that should in theory give you near Haswell CPU performance with well over the performance of Intel iGPU+the ability to make better use of OpenCL. I'm definately waiting to see what AMD comes up with before deciding whether or not to down/side grade this desktop, sell my laptop+a lot of computer bits and get a Zen APU based laptop...because it'd kick the crap out of my GTX 650ti OC 2GB+i5 4440, which is a good combo still.

You can already use OpenCL 2.0 with Broadwell and 2.1 with Skylake, and Kaby Lake is on the horizon with an even newer graphics architecture. And you're assuming you're getting an HBM APU in a laptop before 2017...

[Dabombinable]

3 minutes ago, patrickjp93 said:

You can already use OpenCL 2.0 with Broadwell and 2.1 with Skylake, and Kaby Lake is on the horizon with an even newer graphics architecture. And you're assuming you're getting an HBM APU in a laptop before 2017...

Don't get me wrong, I could wait a long time with what I have, as it is my laptop is a 2010 Pavilion DV6 3010AX (with upgrades) and is still better value even today than anything in Australia under $800, and it was only a $999 AUD laptop.

[byalexandr]

2 hours ago, -BirdiE- said:

Implying 100% of a card's size is related to the manufacturing size...

 

That's why graphics cards have been getting way smaller over the years.... OH wait..

... I'm talking about die size not manufacturing process. HBM made the overall board smaller, that's why the R9 Nano was able to be made.

[-BirdiE-]

5 minutes ago, byalexandr said:

... I'm talking about die size not manufacturing process. HBM made the overall board smaller, that's why the R9 Nano was able to be made.

Firstly, manufacturing process is what determines the size of the die. Also, Halving the size of the die will not halve the size of the card... There are other things on the board that do not shrink with die size.

 

Plus, there are other cards that have been the same size as the Nano before..

[byalexandr]

5 hours ago, -BirdiE- said:

Firstly, manufacturing process is what determines the size of the die. Also, Halving the size of the die will not halve the size of the card... There are other things on the board that do not shrink with die size.

 

Plus, there are other cards that have been the same size as the Nano before..

First, I know it makes the die size smaller. Also, moving stuff to the die as well as doing things like stacking memory decreases the overall amount of stuff on the PCB allowing you to make it smaller. There are things that shrink with die size because those things get technological updates alongside the new manufacturing process.

 

Plus, any of the other cards that are the same size of the Nano are not nearly as powerful. The next step down from the Nano (that's the same size, but still a bit bigger) is the GTX 970 and then the R9 380/285.

[-BirdiE-]

9 minutes ago, byalexandr said:

First, I know it makes the die size smaller. Also, moving stuff to the die as well as doing things like stacking memory decreases the overall amount of stuff on the PCB allowing you to make it smaller. There are things that shrink with die size because those things get technological updates alongside the new manufacturing process.

 

Plus, any of the other cards that are the same size of the Nano are not nearly as powerful. The next step down from the Nano (that's the same size, but still a bit bigger) is the GTX 970 and then the R9 380/285.

Well, you were doing 1 of 2 things:

 

1) Stating that we will see Polaris cards that are half the width of the Nano, but just as powerful.

I can promise you this will not happen

 

2) Stating that years down the road, lower end, small graphics cards will be much smaller than the Nano, but just as powerful.

Geez. Thanks for that brilliant insight. Not really related to this thread... And I'd rather imagine what the big, top end cards are going to be like then, than imagine what the size of cards /w today's processing power will be (when that amount of computer power is no longer impressive).

[byalexandr]

Just now, -BirdiE- said:

Well, you were doing 1 of 2 things:

 

1) Stating that we will see Polaris cards that are half the width of the Nano, but just as powerful.

I can promise you this will not happen

 

2) Stating that years down the road, lower end, small graphics cards will be much smaller than the Nano, but just as powerful.

Geez. Thanks for that brilliant insight. Not really related to this thread... And I'd rather imagine what the big, top end cards are going to be like then.

I said imagine if it was possible. I did not say it would happen nor was I being serious. It would be cool to have as much as I kind of want it to happen, but it's not.

 

I'm more excited for the power efficiency. If it already gets double the performance per watt of a GTX 950, imagine what a 250W TDP card with the same power efficiency can do in terms of performance. If AMD makes a hot chip again but it performs like that, I'm sure people wouldn't even complain. Just like the 980 Ti is just as hot a chip as an R9 290 but it performs well enough that people don't mind. JUST IMAGINE, please, don't kill me for imagining something.

[-BirdiE-]

2 minutes ago, byalexandr said:

I said imagine if it was possible. I did not say it would happen nor was I being serious. It would be cool to have as much as I kind of want it to happen, but it's not.

Alright. I'm with you. Now can we imagine that it was 1 tenth of the size and 500 million times as powerful?

 

That would be pretty sweet.

[byalexandr]

Just now, -BirdiE- said:

Alright. I'm with you. Now can we imagine that it was 1 tenth of the size and 500 million times as powerful?

That would be pretty cool.

[-BirdiE-]

1 minute ago, byalexandr said:

That would be pretty cool.

Apologies for all the sarcasm. I was getting really frustrated with someone else on the forums today, and it put me in a very "shit on everyone!" kind of mood.

[Big-G]

1 minute ago, -BirdiE- said:

Apologies for all the sarcasm. I was getting really frustrated with someone else on the forums today, and it put me in a very "shit on everyone!" kind of mood.

Forum laxative?

[byalexandr]

4 minutes ago, -BirdiE- said:

Apologies for all the sarcasm. I was getting really frustrated with someone else on the forums today, and it put me in a very "shit on everyone!" kind of mood.

It's fine  Appreciate it.

[Deletive]

Meh they can call it whatever they want and say whatever they want but if there isn't a big GCN architecture change i still won't be interested. 

[looncraz]

13 hours ago, patrickjp93 said:

No. Every node change doubles density. 32-22 doubled Intel's, and 22-14 doubled it again. 28-20 came close to doubling density for TSMC and Samsung. 20-14/16 is another such case. Dropping from 28 to 14 should nearly quadruple density. The transistors should be ~1/4 the size on average.

 

FinFet pitch requirements and the utilization of 20nm upper layers (and probably other factors) throw a wrench into area scaling with all but Intel's 14nm process:

 

[looncraz]

1 hour ago, Deletive said:

Meh they can call it whatever they want and say whatever they want but if there isn't a big GCN architecture change i still won't be interested. 

We already know this is a massive change to GCN, including a new ISA.

 

http://www.eteknix.com/amd-prepping-post-gcn-greenland-baffin-ellesmere-gpus/