EETimes: ST plans for Dresden FDSOI production

[AtenRa]

20nm bulk-Si then?

Could be yes, they(AMD) said no Server/Desktop update until 2014

 

[Abwx]

I did read at Hardware.fr that according to an engineer Glofo has working
28nm with all devellopement tools finalized so designs can be started but
the electrical caracteristics are not definitive and will be updated later.

From my understanding , one can design a circuit but simulations at higher
frequencies , the ones of interest , wont be accurate in respect of the real
world implementation , so it is likely that large scale 28nm production at the NY facility , wich was announced to debut currently , is still not in track for deliveries in H1 13.

 

[ShintaiDK]

I did read at Hardware.fr that according to an engineer Glofo has working
28nm with all devellopement tools finalized so designs can be started but
the electrical caracteristics are not definitive and will be updated later.

From my understanding , one can design a circuit but simulations at higher
frequencies , the ones of interest , wont be accurate in respect of the real
world implementation , so it is likely that large scale 28nm production at the NY facility , wich was announced to debut currently , is still not in track for deliveries in H1 13.

Back in 2009 the NY fab was to ship 22nm in 2012 for revenue and 28nm in 2011. Now its 2013...
TSMC should also have shipped 28nm in what, H2 2010?

Foundry announcements are really to be taken with a grain of salt.

 

[AtenRa]

Back in 2009 the NY fab was to ship 22nm in 2012 for revenue and 28nm in 2011. Now its 2013...
TSMC should also have shipped 28nm in what, H2 2010?

Foundry announcements are really to be taken with a grain of salt.

No, it was meant to start 28nm production late 2012

 

[ShintaiDK]

No, it was meant to start 28nm production late 2012

http://www.anandtech.com/show/2814
http://www.geek.com/articles/chips/preview-of-globalfoundries-22nm-fab-facility-due-in-2012-2009056/

Or 22nm in Dresden
http://www.eetimes.com/electronics-news/4088059/GlobalFoundries-Dresden-fab-to-run-22-nm-CMOS

Free choice of failure. Big words, not so much results.

 

[AtenRa]

We were talking about NY Fab 2 not Dresden. Fab 2 was meant to be i full production in 2014 from the start.

 

[ShintaiDK]

We were talking about NY Fab 2 not Dresden. Fab 2 was meant to be i full production in 2014 from the start.

Lets try again. NY fab:

The facility will ramp toward 22nm production in 2012, though it will be taking orders for 32nm and 28nm designs early on. Presently

My first official meeting with GF came just this past week, in Saratoga County, NY. That’s the site where its brand new post-AMD fab will be constructed and it’s sort of a big deal. I was there for the ceremonial ground breaking, but concrete doesn’t get poured until September, the shell won’t be finished until 2010 and the fab won’t be done until 2011. Globalfoundries won’t be able to ship revenue generating wafers until 2012.

When Fab 2 comes on line in 2012 it'll be a 28nm/22nm fab, with the bulk of its production hopefully being 22nm.

And now we stand here in 2013. Only to look back on 2012 for a year with more hot talk and no results.

 

[Abwx]

Back in 2009 the NY fab was to ship 22nm in 2012 for revenue and 28nm in 2011. Now its 2013...
TSMC should also have shipped 28nm in what, H2 2010?

Foundry announcements are really to be taken with a grain of salt.

And ?..

Apart from using those datas to trash glofo/amd , as usual ,
what is your point ?.

 

[ShintaiDK]

And ?..

Apart from using those datas to trash glofo/amd , as usual ,
what is your point ?.

Its about GloFo being good about announcements, but less performing about actually deliveries. Same can be said about TSMC as I also said.

So how did you fit AMD in there?

 

[Abwx]

They are still more in line with their announcements than the one
of 10Ghz CPUs that we should be using for 7 years according to an
announcement that is not still fullfiled to this day...

More seriously , i pointed very recent info about the current
state of Glofo s process and all you did was to reingurgitate
old datas that did bring nothing to the current discussion.

 

[Phynaz]

They are still more in line with their announcements than the one
of 10Ghz CPUs that we should be using for 7 years according to an
announcement that is not still fullfiled to this day...

Actually I think it was 10 years ago that AMD's CTO Fred Weber said future AMD CPU's would run at 10Ghz.

Edit:
It was the 2003 Microprocessor Forum.
http://pc.watch.impress.co.jp/docs/2006/0119/kaigai233_01l.gif

 

[Ajay]

Yes i know they said Kaveri will be 28nm, i was talking about Server/Desktop (FX) SteamRoller.

Could be yes, they(AMD) said no Server/Desktop update until 2014

So who will provide 20nm (bulk-SI/ FD-SOI) in 2014? GF? That is possible since test shuttles have already been run so all the heavy lifting will take place in 2013 - but I don't know if those test shuttles where bulk or SOI.

On the negative side, Opterons will need at least 6 modules to be anywhere near competitive with Intel, IMHO (if the IB-EP 12 core rumors are true). That doesn't bode well for there the existence of an FX desktop CPU (who needs that many cores on the desktop?).

I guess all I'm really saying is that 2014 is really fuzzy, IMO.

 

[Abwx]

As usual , Phynaz............

Wasnt it Intel that was first , when launching the P4..?...

Dec 11, 2000 1:00 AM

It will take powerful transistors to build such a system, says Intel--which takes a step in that direction by unveiling on Monday a high-speed transistor technology designed to produce 10-GHz CPUs in five to ten years

http://www.pcworld.com/article/36348/article.html

 

[Ajay]

Saw this at EETimes:



I really enjoyed the embedded video at the end of the article, direct link here, as it did a great job highlighting some basic xtor functions that are of interest to people.

Even if you don't care about FD-SOI, watching the video can be educational.

If the numbers are true and not just hype, and the ease-of-porting existing 28nm bulk designs to 28nm FD-SOI is true, then AMD could really see a boost in their 28nm products by migrating to FD-SOI at GloFo

Hmm, this could be huge for AMD. They can take existing 28nm designs and migrate them to 28nm FD-SOI w/30% higher performance or much lower power without using a large design team for each. AMD can reduce its node usage to only 2 because they can get the functionality of 4 nodes out of those two.

Still, nothing is certain since Dresden hasn't even started the conversion yet:

Hartmann said Fab 1 in Dresden, Germany, would be a logical place for Globalfoundries' to host FDSOI as it is already the where the foundry manufactures 32/28-nm bulk CMOS. "If we start this month or in February Globalfoundries could be in production in the fourth quarter of 2013," said Hartmann.

 

[Pedroc1999]

Hasnt anyone thought about the future radeon series. It will run 28nm process and could potencially use this. This would revelucionise gpu power

 

[Hitman928]

Yeah, I don't know if this will have any effect on AMD products expected this year/early next year, but it would be good for AMD if infact the process change is as easy as they say. They mention in the video that they have further node developments planned and there was a rumor floating that AMD was planning on going back to SOI after 28nm, perhaps this is the reason.

 

[Idontcare]

Hasnt anyone thought about the future radeon series. It will run 28nm process and could potencially use this. This would revelucionise gpu power

I really thought the GPU guys, Nvidia and AMD, would be all over SOI for GPU's simply for the fact they are truly power-limited in their clockspeeds and design targets.

Back in 2009, when ARM and Soitec published results comparing SOI vs bulk-Si one comparable 45nm process nodes (fabbed at IBM) showing that they realized a 40% reduction in power and a 7% reduction in die-size, I was convinced this was a process made for the GPU race.

ARM reports 45-nm SOI test chip with 40% power-saving

The results show that 45-nm high-performance SOI technology can provide up to 40 percent power savings and a 7 percent circuit area reduction compared to bulk CMOS low-power technology, operating at the same speed. This same implementation also demonstrated 20 percent higher operating frequency capability over bulk while saving 30 percent in total power in specific test applications.

[URL="http://www.eetimes.com/showArticle.jhtml?articleID=220301622"]http://www.eetimes.com/showArticle.jhtml?articleID=220301622[/URL]

Demonstrated (not mere simulation or theory) 20% higher clockspeed and 30% lower power (both!) in a logic IC.

I have to imagine that the reason the GPU guys did not jump on this was because there must have been some serious downsides to using the process that neither ARM nor Soitec were interested in publicizing.

I guess my point here is just to say we've been down this path before, the hype path regarding how great future IC's will be when they are produced on SOI, so my optimism is guarded.

But the opportunity is there, on paper at least, with 28nm FD-SOI. So why wouldn't fabless companies be interested?

 

[Phynaz]

As usual , Phynaz............

Wasnt it Intel that was first , when launching the P4..?...



http://www.pcworld.com/article/36348/article.html

Oh, so AMD was just copying the number from Intel then?

Intel said it was designed to reach 10Ghz. That didn't work out.

AMD said they would deliver 10Ghz.

I don't suppose you can comprehend the difference.

 

[erunion]

As usual , Phynaz............

Wasnt it Intel that was first , when launching the P4..?...

R&D projects are not a road map.

Process node delivery timelines are.

 

[Abwx]

The relevant process nodes were delivered on the following
years but not the 10Ghz frequency.

 

[Phynaz]

The relevant process nodes were delivered on the following
years but not the 10Ghz frequency.

You got that right. How long did it take AMD's 65nm process to reach the same clocks as their 90nm process?

 

[Abwx]

It would be more fruitfull for the thread if you could give us a few
clues , if ever you have some , about Glofo s current nodes , wich
is the subject discussed here.

 

[Phynaz]

Actually the subject isn't about their current nodes.

But yes, you did take us off topic and we should get back on it.

 

[AtenRa]

Lets try again. NY fab:

And now we stand here in 2013. Only to look back on 2012 for a year with more hot talk and no results.

I have shown a slide from 2011 clearly saying "production expected to come online in late 2012". Fab 2 Full production was and still is targeted for 2014. Dont forget we are talking for a NEW fab construction, that is for a new building construction, new machinery installation, power plant installation etc etc.

Also, 28nm will be transferred to Fab 2 from Dresden and 20nm will be transferred to Dresden from Fab 2.

 

[SocketF]

But the specs alone don't ensure AMD will use the node for simple fact that the timing is rather off. Usually a design team needs a good 2-3 years beforehand to design and optimize an IC for the specific electrical parametrics of a given node.

As you said: "node", however FD-SOI 28nm is still 28nm ;-)

If ST Micro is only just now making its 28nm FD-SOI node available to IC designers then it starts that 2-3 yrs clock ticking now...unless the 28nm bulk-Si -> 28 FD-SOI porting process really is as straightforward easy as they claim it to be.

I wont doubt them, I have read that statement now several times from different people. An ARM engineer had a presentation where he stated that *theoretically* the conversion should be doable automatically with tools, however he doubts it. Practically, I would assume that it means that it cannot be not too complicated if done by hand, 1y might be enough...
Globalfoundries will also provide 2 FD-SOI processes, I assume one is the quick and dirty bulk-conversion, the other one is the 100% designed and optimized for FD-SOI:

As part of its technology offerings, GlobalFoundries plans to offer two versions of FD-SOI: minimum and maximum. The maximum version is a technology tuned for a specific application. IBM and STMicroelectronics are examples of companies that would utilize maximum versions of FD-SOI. Meanwhile, the minimum version is a simple and an “out of the box” FD-SOI technology, said Subramani Kengeri, head of advanced technology architecture at GlobalFoundries.

http://chipdesignmag.com/display.php?articleId=5199

And if you remember, the next server platforms where postponed for 1 year from 2013 to 2014.. maybe just a coincidence, but hopefully not ;-)

Yes i know they said Kaveri will be 28nm, i was talking about Server/Desktop (FX) SteamRoller.

So you say that the next server processors will be 28nm Piledriver cores? Not really believable if they already taped-out SR-cores for Kaveri, and Richland is still only a 32nm Piledriver design. However AMD has not stated anything explicit so far, so I'll give the 28nm-Piledriver-idea an outsider's chance of 10% ;-)