Amd Volcanic Islands details

[SKYMTL]

Doesn't seem to be so impressive. Then again I don't think Nvidias the first Maxwell GPU's next year will be double the performance of a 780/Titan either. They will need a die shrink for that.

Maxwell is 20nm...

16nm isn't due until well into 2015.

[nossy23]

Maxwell is 20nm...

16nm isn't due until well into 2015.

I've heard that the first Maxwell might still be 28nm, since they will not get 20nm up to speed in time.

[Darakian]

Edit- Forgot to ask if anyone knows if DP is important/used for Radeon Sky product line.
I don't think it would but I'm not sure.

Re Edit- From some of the marketing/press release docs I found, it doesn't appear to use/need DP at all since the docs make no mention of DP performance.

Radeon Sky is aimed at cloud gaming providers who would use whatever a given game uses, ie SP.

[SKYMTL]

I've heard that the first Maxwell might still be 28nm, since they will not get 20nm up to speed in time.

Volume production for 20nm is Q1 2014. That meshes perfectly with Maxwell.

[PedantOne]

few pages back is my info about 470-480 mm2 of hawaii chip

[PedantOne]

Volume production for 20nm is Q1 2014. That meshes perfectly with Maxwell.

you are wrong, maxwell will be released at 28nm, not all skus, but first one will be 28nm, high end maybe at 20 nm

[GoldenTiger]

Volume production for 20nm is Q1 2014. That meshes perfectly with Maxwell.

Nomnomnomnom... lookin' forward to them Maxwells!

[boxleitnerb]

20nm is too expensive. I would not expect 20nm GPUs before Q3 2014. If 20nm were ready in Q1 2014, why in the world would AMD release their stuff on 28nm only 2-3 months earlier? That makes no sense!

[GoldenTiger]

20nm is too expensive. I would not expect 20nm GPUs before Q3 2014. If 20nm were ready in Q1 2014, why in the world would AMD release their stuff on 28nm only 2-3 months earlier? That makes no sense!

Perhaps because right now they are strongly outperformed stock-to-stock and oc-to-oc on their top cards, and are having to unload them at very low pricing in the US at least in order to even sell them? Dumping 7950's with free games for $180 after rebate can't be helping them, or $280ish 7970's... not to mention their brand image being tarnished by crossfire issues and low prices making them look "cheap" (I don't agree, but it is what some people will think).

[DefStar]

Perhaps because right now they are strongly outperformed stock-to-stock and oc-to-oc on their top cards, and are having to unload them at very low pricing in the US at least in order to even sell them? Dumping 7950's with free games for $180 after rebate can't be helping them, or $280ish 7970's... not to mention their brand image being tarnished by crossfire issues and low prices making them look "cheap" (I don't agree, but it is what some people will think).

Yeah, sadly that is the only reason why. They said a while back no new GPUs until 2014.

[boxleitnerb]

But you don't develop a GPU that will either be outdated by your own 20nm-GPU a quarter later or buttkicked by the 20nm-GPU of the competition after the same time. Designing and taping out a GPU costs dozens of millions. They would never recover the costs in 3 months. Nowadays GPUs live 12-18 months, possibly even 24 months (reusing them in a refresh lineup like GK104, GK106 and now Tahiti).

[Ace123]

Dozens of millions haha

[boxleitnerb]

What's so funny? Engineers have to be paid to make the design and test it. One set of masks for a GPU is very expensive.

[Flanman]

i think he might find it funny that you chose to use "dozens" instead of the often heard "tens"

[boxleitnerb]

Lol, well then tens of millions.

[DefStar]

Well, I guess AMD really wants to be the best...so they are willing to make some sacrifices now, to come out ahead in the long run. I would have bought a 780 and then I would have to change my monitor to support Light Boost...making it less likely for me to switch back to AMD any time soon....so AMD is making the right move in my case.

[Mk]

any one have an idea of when the NDA will end on those?
hint hint @SKYMTL

[DilTech]

But you don't develop a GPU that will either be outdated by your own 20nm-GPU a quarter later or buttkicked by the 20nm-GPU of the competition after the same time. Designing and taping out a GPU costs dozens of millions. They would never recover the costs in 3 months. Nowadays GPUs live 12-18 months, possibly even 24 months (reusing them in a refresh lineup like GK104, GK106 and now Tahiti).

ATi did so before with the x1800xtx -> x1900xtx. There was like a 2 month gap there. That's it.

[BHD]

Next Generation Nvidia Die : 28 NM

[boxleitnerb]

ATi did so before with the x1800xtx -> x1900xtx. There was like a 2 month gap there. That's it.

True, although I never thought that made sense. But today making a GPU is more expensive. Much more expensive. And AMD isn't exactly swimming in money right now, they have lots of other projects.

[SKYMTL]

any one have an idea of when the NDA will end on those?
hint hint @SKYMTL

Before the end of the year one would hope. As you know, NDA lift dates are under NDA.

[PedantOne]

AMD Hawaii meeting on Hawaii island is 22-24 of September. NDA for press maybe ends 09/25. But only for few slides, architecture, by my source NO CARD will be shown till October + Battlefield 4 launch.

[iMacmatician]

From Forbes: "Exclusive Interview: AMD's Matt Skynner Talks New Radeon Cards, Next-Gen Consoles, 7990 Criticism."

He gives a bit of information about AMD's upcoming GPUs.

They're coming in Q4. I can?t reveal a pricepoint but we?re looking at more traditional enthusiast GPU pricepoints. We?re not targeting a $999 single GPU solution like our competition because we believe not a lot of people have that $999. We normally address what we call the ultra-enthusiast segment with a dual-GPU offering like the 7990. So this next-generation line is targeting more of the enthusiast market versus the ultra-enthusiast one.

It?s also extremely efficient. [Nvidia's Kepler] GK110 is nearly 30% bigger from a die size point of view. We believe we have the best performance for the die size for the enthusiast GPU.

Another thing I can tell you is about the process node: this GPU is in 28nm. Some have speculated that it was 20nm and it?s not for a specific reason: At 28nm for an enthusiast GPU, we can achieve higher clock speeds and higher absolute performance.

That's around 425-435 mm^2 for (what I assume to be) Hawaii then. (I've seen different places say 551 mm^2 or 561 mm^2 for GK110.)

[PedantOne]

it is bull, if its Hawaii die size = GK110 - 30 percents, has the same die size like Tahiti, no more SPs, no more ROPs, it cant be true

[boxleitnerb]

561/1.3 = 430
Percentage calculation is multiplicative, not additive

Yet it is difficult to imagine that this GPU has lots and lots of additional units at 430mm2.