AMD 28nm Southern Islands chips in Q2 2012?

[Dark-Energy]

They just launched their new parts last quarter. Next quarter for new 28nm stuff? It just won't happen. Q2 2012 sounds more likely.

Both of them seem unlikely. Northern Islands (Cayman) was originally supposed to be on 32nm but due to the cancellation of it by TSMC, AMD resorted to 40nm, which still seemed to work fine. It's hard to say if Cayman would have been faster on 32nm, due to less power consumption and therefore more shaders, but probably more expensive.

Which basically means that, how far they were into 32nm research/production when it was cancelled for 28nm is probably the same amount of time between the original planned release of Cayman and the original planned release of Southern Islands (whatever it was). Then there's the approximate 6 month delay between new GPU's as well on the side (I'm talking a wild guess) but it still seems too far for Q2 2012. Q2 2011 seems more plausible because it wasn't AMD's fault, they're not the ones behind. But maybe they took this opportunity and that lost time to improve their next GPU. But yeah, Q2 2011 isn't possible, but neither is Q2 2012.

I'd say Q4 2011, or late Q3 2011 if we're lucky, for 28nm GPU's from AMD.

[eXceeded]

innovation.

Really hope someone gets this

[Aten-Ra]

We could see an HD4770 like 28nm graphics card in Q2-Q3 2011 but i dont expect a high-end 28nm graphics cards before Q4 2011.

[Lanek]

In general, you are right, Pipe cleaners are low end gpu's... due to the cost and risk reduction.

At same time this don't mean you will not see 2months then other series goes in this process.. At the moment they are able to evaluate correctly how the production / yield can be ramped.

Now don't take too much care of Nvidia words during their financial reports.

They said: we will not have anything in 28nm this year. They don't speak for AMD.

[marten_larsson]

Also, Nvidia hasn't really being fast going on to newer nodes, they were at least six months behind AMD on 40nm...

[Iconyu]

Really hope someone gets this

Dis r srs forum (sorry my l33t failed me here) 4 srs bizns.

And yes I got it, but for some reason I've never understood why the Japanese thought the name Innovator was a cool name.

On topic, I personally was hoping that both sides would have 28nm high end parts on offer when Sandybridge-E turned up, but from the sounds of it they'll just release PCI-E 3.0 parts

[Pantsu]

I think Nvidia was referring to tegra when they said no 28nm this year. AFAIK next gen gpus will come by the end of this year, if TSMC doesn't drop the ball again. Considering they're already making LP wafers for Xilinx and Altera I don't see it impossible for AMD to start producing a pipe cleaner product in Q2.

[Sam_oslo]

2112 is a typos made by CENS.com, most probably it should be Q2 2011.

Anyways, throwing this rumors/announcement right now, just before the release of GTX595, suggests that AMD doesn't have anything to beat/match GTX595 until then.

[Sn0wm@n]

2112 is a typos made by CENS.com, most probably it should be Q2 2011.

Anyways, throwing this rumors/announcement right now, just before the release of GTX595, suggests that AMD doesn't have anything to beat/match GTX595 until then.

LOL


i love it when you come up with wild theories like that

[Sam_oslo]

LOL


i love it when you come up with wild theories like that

LOL, Love it or hate it, those "wild theories" have been having a tendency to become true.

[Sn0wm@n]

LOL, Love it or hate it, those "wild theories" have been having a tendency to become true.

hahahahah i needed such laugh this morning .. thank you mate

[StimulatedBoy]

hahahahah i needed such laugh this morning .. thank you mate

Please forgive my fellow countryman. Whatever he is on, it must be powerful given the long lived tenacity, I want some...

[Lanek]

2112 is a typos made by CENS.com, most probably it should be Q2 2011.

Anyways, throwing this rumors/announcement right now, just before the release of GTX595, suggests that AMD doesn't have anything to beat/match GTX595 until then.

nice sense of humor, i like that.

[Dark-Energy]

2112 is a typos made by CENS.com, most probably it should be Q2 2011.

Anyways, throwing this rumors/announcement right now, just before the release of GTX595, suggests that AMD doesn't have anything to beat/match GTX595 until then.

Nah, it's because half of the AMD employees got abducted by aliens and couldn't design a faster card in time. I think that's a more probable cause

[Det0x]

Please forgive my fellow countryman. Whatever he is on, it must be powerful given the long lived tenacity, I want some...

I agree

[hugin72]

I agree

me 2

[turtle]

Just saw this article today (I know, a little late) but had to stop by to see what the XS boys were thinking. I put this in a comment post on the article elsewhere, which spiraled into a typical me-like long-ass post. Here's a copy pasta with some included hyperlinks for good measure.


You’ll notice the original source says ‘chip’ (singular) in Q2 2012. Perhaps the info was concerning the refresh chip the mobile maps from DH called Wimbledon (that is supposedly starting production in Q212) that may complete the SI series. That chip is supposedly 256-bit. The other chips mentioned included two 128-bit chips and a 192-bit chip entering production (or being released?) in Q411 (when TSMC is said to be ready for production). I personally believe the 192-bit chip is half of the halo chip and the one mentioned will replace/supplement it (the 192-bit part) as well as the (then) aging Cayman in Barts/GF114 current market later on.

Perhaps it’s coming later because AMD will want to get mature yields from it considering it may be larger than the former chips in this market (Juniper,Barts) because of their new projected golden shader/rop ratio. It will be almost-certainly higher than Cayman (14 or 16:1?), but maybe will be unable to price it as such because of guesstimated competition or expected MSRP of higher-end parts originally and after prices settle. Perhaps they are also afraid it would detract from those that may buy the higher-end chip. For instance, this part may be 32 ROPs, 2/3 of the high-end, while the original 28nm parts would be 1:1, 1:2, and some combination of 1:3, 1:6, 1:12 (48, 24, 16/8/4 ROPs). It would not be surprising if the first SI parts more-or-less sandwich Cayman in performance or the second-tier chip has some comparative detriment, just as the original 5700 (Juniper) parts did the 4800 series. This chip could be the mid-life kicker, as Anandtech (iirc) once called these refreshes; perhaps similar to what Cayman was originally intended to be on 32nm and also probably a great bang-for-buck part.

Perhaps this chip may have been initially planned at GF (I don't think I need to link those rumors), but instead AMD has decided to stick with TSMC. Right now (a year out) would be a good time to make that decision. The reasons could vary from some combination of yields/power consumption/speed currently on GF’s process because of their gate-first tech to AMD wanting to garner either better prices and/or priority from TSMC. If what I outlined above is true (that this ‘mid-life kicker’ is planned to be their Barts/GF104-114) it would make sense they originally would have liked to have the die be as small as possible (which is GF’s strength). Perhaps now, with TSMC, they may just have to do more extensive harvesting for mobile parts than what they had originally planned but will be counteracted with higher clocks at low-enough power for the mobile market, which should be TSMC’s strength for 28nm.

Note everything I said is likely between mostly/all wrong and should be taken as such. They probably meant SI will start production in Q211 (which the linked X-bit article has TSMC stating ramping of production in Q3/Q4 2011) and the original article needs proof-reading. Perhaps DH had it wrong and that chip is coming first (with production in Q211). But seriously, what's life without a little crazy speculation? No fun, no fun at all.

[SQjay]

Really hope someone gets this

Setsuna or Kira?

[saaya]

whats going on with glofo?
fusion = tsmc
new gpus = tsmc

?

[Motiv]

whats going on with glofo?
fusion = tsmc
new gpus = tsmc

?

bulldozer

[Iconyu]

bulldozer

Yeah, it's like AMD are freeing up GF's capacity or at least leaving it only doing their none fusion CPU's If AMD drop Phenom production then GF will be pretty much doing Bulldozer only for a few months.

[marten_larsson]

Yeah, it's like AMD are freeing up GF's capacity or at least leaving it only doing their none fusion CPU's If AMD drop Phenom production then GF will be pretty much doing Bulldozer only for a few months.

Llano is GF as well (since it's on the 32nm node). Currently it is:
GF: CPUs except Zacate/Ontario (which is same)
TSMC: GPUs and Zacate/Ontario

When BD/Llano comes it will be
GF: BD/Llano
TSMC: GPUs/Zacate/Ontario...

So no difference there.

[onethreehill]

http://www.donanimhaber.com/ekran-ka...l-haritasi.htm

[Chrono Detector]

This is rather pathetic, we must wait until next year, great job TSMC, really screwed things over.

[RPGWiZaRD]

Hopefully NV can get out Kepler sooner, 2012 is a too long wait. Yes I know it's scheduled for this year but you never know if there will be delays or some issues.