IBM to manufacture Cell on 45nm SOI, Sony/Toshiba to move to 45nm bulk process

one

Unruly Member
Veteran
The official announcement for the sale of advanced semiconductor plants at Sony to Toshiba has been made today. Along with that Sony's plan to the 45nm process has been announced too.

Japanese PR
http://www.sony.co.jp/SonyInfo/News/Press/200710/07-1018/07-1018.pdf

1. Sony has reached an agreement with IBM to mass-produce Cell B.E. on 45nm SOI lines at IBM East Fishkill which will be adapted from current 65nm lines.

2. Sony Group and Toshiba continue the cooperation for 45nm bulk process development. The new joint company will move the current 65nm process technology to 45nm for advanced semiconductors for games and digital appliances.

Japanese PR
http://www.sony.co.jp/SonyInfo/News/Press/200710/07-1018B/07-1018B.pdf

300mm wafer lines at Nagasaki will be sold to Toshiba before March 2008. A new joint company to manufacture Cell and RSX will launch in April 2008 (Toshiba 60%, Sony 20%, SCE 20%). They manufacture 65nm semiconductors and continue to develop the 45nm process. Manufacturing lines at Oita are also bought by Toshiba.
 
Nice ! Have been waiting for this news. Doesn't seem to indicate any high level timeline though :(

I thought IBM is supposed to do an event to award the Cell programming competition winners soon (This month or next) ? They should also showcase the automated car. I am sure many are interested to see how well it perform for real.
 
English PR, fewer details but more quotes.

http://www.japancorp.net/Article.Asp?Art_ID=15824

Sony Group Expands High-Performance Semiconductors Manufacturing Alliances with IBM and Toshiba

Tokyo, Japan, Oct 18, 2007 - (JCN Newswire) - Sony Corporation and Sony Computer Entertainment Inc. (Sony Group) today announced two new and separate partnership agreements with the goal of enhancing the capabilities of Sony Group's PlayStation business.

IBM and Sony have signed an agreement to extend their existing manufacturing relationship to 45-nanometer silicon-on-insulator (SOI) for high volume production of the Cell Broadband Engine(TM) (Cell/B.E.) processor. Both companies will collaborate to optimize the 45-nanometer manufacturing capability to produce a lower power and lower cost processor for Sony Groups PLAYSTATION(R)3. IBM will lead the evolution of the Cell/B.E. component by transitioning production from the existing 65-nanometer generation to 45-nanometer in IBM's manufacturing facility in East Fishkill, New York.

Separately, Sony Group will extend their alliance with Toshiba to manufacture high-performance LSI by using 45-nanometer bulk process technology. By forming a joint venture company that leverages the knowledge and experience of both companies, Sony Group and Toshiba intend to improve power consumption and cost competitiveness for game and digital media applications by advancing to the 45-nanometer generation from the existing 65-nanometer generation process.

Sony Group will advance the high-performance semiconductors used for PlayStation while reducing costs by further strengthening its collaboration with its respective partner companies. These enhancements will continue to strengthen the leading capability of the PLAYSTATION(R)3 system and Sony Group's overall PlayStation business. "The transition to 45-nanometer SOI technology for the Cell Broadband Engine is a critical step to advancing the world class processor cost and power characteristics while maintaining supercomputer on chip performance capability" said Adalio Sanchez, general manager for IBM Global Engineering Solutions. "The IBM team is excited to lead the production of the 45-nanometer Cell/B.E. processor in its East Fishkill 300mm facility and to help drive a broader use of this capability as both power and size scale down significantly."

"Games are one of the focus areas for Sony Group's semiconductor business on which we have been concentrating since earlier this year. At the same time, we have carefully assessed the manufacturing investment plan for advanced process technologies for 45-nanometer generation processes and beyond. We believe the production alliances that we have formed with IBM and Toshiba to manufacture high performance semiconductors for PlayStation by using state-of-the-art process technologies will lead to the advancement of the high-performance semiconductor business for PlayStation. Sony Group will collaborate with IBM and Toshiba to expand its high-performance semiconductor business for PlayStation by maximizing knowledge in high-performance semiconductor design capability it has gained over the years," said Yutaka Nakagawa, Executive Deputy President, Officer in charge of Semiconductor & Component Group, Sony Corporation.

"I am delighted with our respective collaborations with Toshiba and IBM, which aim to achieve LSI scaling and power reduction by 45-nanometer-generation process technology, and which will enable us to further drive the advancement and cost reduction of semiconductor devices for PlayStation systems. Cutting-edge semiconductors are core devices for the PlayStation platforms. With the support of these strengthened relationships, SCE is committed to develop further the PlayStation business and offer new and innovative interactive entertainment," said Kaz Hirai, President and Group CEO, Sony Computer Entertainment, Inc.

"This new cooperative relationship will enhance our production efficiency in high-performance semiconductors for game consoles, and accelerate the early migration of next generation process technology. Market growth of high-performance semiconductors for game consoles, driven by strong demand for such devices as the RSX graphics engine for PLAYSTATION(3)3, is expected to continue and we will promote it as an important part of our System LSI business, a strategic business area. We are happy to continue to support Sony Group's PlayStation business through this new framework, and we aim to offer a steady supply of high-performance semiconductors in next generation process technology as soon as possible to support market expansion," said Shozo Saito, Corporate Senior Vice President of Toshiba Corporation, President and CEO of Toshiba's Semiconductor Company.
 
Followup news about the 32nm alliance
http://www.tradingmarkets.com/.site/news/Stock News/792750/
Sony Corp to exit chip development deal with IBM, Toshiba - report
Tuesday, November 06, 2007; Posted: 06:44 PM

TOKYO, Nov 07, 2007 (Thomson Financial via COMTEX) -- SNE | charts | news | PowerRating -- Sony Corp, the world's second-largest consumer electronics maker, plans to withdraw from its basic research project with IBM Corp and Toshiba Corp to develop manufacturing technologies for next-generation semiconductor chips, the Nikkei reported on Wednesday.

The move follows the company's decision to sell Toshiba its production facility for the Cell microprocessor, used in the PlayStation 3 video game console, the business daily said.

Under the project, the three companies are researching and developing materials and technologies necessary for the manufacture of chips with line widths of 32 nanometers or narrower.

This project began in late 2005 and is scheduled to run until the end of 2010, the report said.

In addition, Sony plans to cancel capital investments in production facilities for the 45 nanometer or later generations of the Cell microprocessor, it said.
 
I'm confused. Does this mean the PS3 is going to stay at 54nm and are they ditching cell when 2010 is here?

No, it means they'd have nothing more to do with research on lower manufacturing processes...since they'll be getting others to do that manufacturing. So basically they'd be dependent on, say, Toshiba's and IBM's own work on those processes. They could and would still commission Toshiba and/or IBM to manufacture Cell at those lower processes...they'd just no longer be involved on research on those processes themselves.

From the report and reading the Sony spokesperson's comments, I don't see the relevance of this to chip design itself (e.g. future Cell development), 'just' to manufacturing technologies.
 
So their most likely path of integration is now RSX+Southbridge, rather than RSX+CELL, I guess. And then you'd eventually see RSX+Southbridge+CELL on Bulk 32nm, I guess. Makes sense.
 
From the report and reading the Sony spokesperson's comments, I don't see the relevance of this to chip design itself (e.g. future Cell development), 'just' to manufacturing technologies.

And further, they've even clarified this themselves:

http://today.reuters.com/news/artic...1_T348222_RTRUKOC_0_US-SONY-SEMICONDUCTOR.xml

While it is withdrawing from the joint R&D on technology, Sony said it would continue to work with Toshiba and International Business Machines Corp on design work for cutting-edge chips with 32-nanometre or narrower circuitry.

Which sounds a bit like it could refer to whatever is going in the next Playstation.
 
I think the point of confusion is the phrase "R&D technology." A lot of folk might read that as overall semiconductor R&D, but in this case it's referring specifically to manufacturing process R&D. Which, is in line with what was prior reported anyway...
 
An interview with Yutaka Nakagawa, Sony Corp's executive deputy president and officer in charge of Semiconductor & Component Group, about these moves.

http://techon.nikkeibp.co.jp/english/NEWS_EN/20071108/142069/

In short, he says the decision was taken because of a lack of products using Cell. Initially Sony intended to make their manufacturing capacity worthwhile by applying Cell in a number of areas, but in an interesting claim, he says intentions for multi-purpose use were spoiled because the chip was designed too tightly for games applications (something that's quite the opposite of what STI has claimed, and indeed, those who would question Cell's credentials as a chip for gaming ;)).

Also, Sony will continue to produce Cell at their facility at the IBM plant in the US. They will also continue working with STI on 32nm process technologies through FY 08, but after that they will exit that agreement. They will, however, continue to R&D in this area in-house in order to support choices they make in the future regarding process technologies when contracting work out.
 
Not good for the future of Cell :(

Probably for the current chip we know as the "Cell", but Im sure there will be future chips that are derivatives of the current chip that will be designed and implement to work well within and who's costs are conducive to their specific enviroment.
 
The wording says the Sony bigwig thinks it was too console-centric in design. If he can't see the flexibility of the system, that's not going to change with other variants. Hence the whole drop in technology research. They don't see a large enough market for high-end technology. They don't see a market for Cell in other Sony products that'd warrant their own fabs to produce boatloads of Cell processors.

If he was keen on using Cell in other devices, there'd have a reason to maintain volume output and invest in economies. That's not going to happen. I'd very much like to hear what he feels is wrong with the design, and how it's console centric.
 
The wording says the Sony bigwig thinks it was too console-centric in design. If he can't see the flexibility of the system, that's not going to change with other variants. Hence the whole drop in technology research. They don't see a large enough market for high-end technology. They don't see a market for Cell in other Sony products that'd warrant their own fabs to produce boatloads of Cell processors.

If he was keen on using Cell in other devices, there'd have a reason to maintain volume output and invest in economies. That's not going to happen. I'd very much like to hear what he feels is wrong with the design, and how it's console centric.

The XDR requirement?
 
Shifty I don't think it's bad news for Cell... at least no moreso than when Kutaragi lost control of semiconductors years ago to begin with. I think it's been well-established for some time that Sony would be filling their CE needs with less-expensive solutions. Plus to focus on CE would be to neglect that Cell has gained huge traction in HPC, and I don't think we'll be seeing it go away there anytime soon.

Cell as it is is just too big/hot for most CE requirements; I think they were right that the strategy before at Sony under Kutaragi was "Cell first." Now - I don't think that's bad per se, but it hurts when corporate direction changes midway; so what they were left with was all the R&D and capital expense that went into developing a "one-for-all" architecture, but a move away from actually translating it to CE in favor of pursuing less risky avenues. Because it was a gamble to be sure from a costs perspective.

Personally I'm pleased with where Cell's at in terms of expanding beyond gaming... I haven't been expecting it to make its way into Sony consumer electronics for some time now. If it happens, it'll be through Toshiba. In the meantime we'll be seeing the HPC version of Cell start shipping, Cell migrate down through to 45nm and beyond, and I think fairly likely a 'Cell 2' by the time PS4 comes around.
 
The lack of any downscaling of the architecture is disappointing though. Such a flexible architecture would have had a lot of potential, and the choice of alternative CE-specific ICs, though it clearly makes sense from a business perspective, means a simplified CE development situation is lost. One architecture for all devices here you could run the same unified interface and services seems a very strong selling point IMO to hook buyers into your brand. Where Sony's CE reach going so far in all directions, begin able to attract to always by Sony because of a unified experience would surely serve them extremely well.
 
The lack of any downscaling of the architecture is disappointing though. Such a flexible architecture would have had a lot of potential, and the choice of alternative CE-specific ICs, though it clearly makes sense from a business perspective, means a simplified CE development situation is lost. One architecture for all devices here you could run the same unified interface and services seems a very strong selling point IMO to hook buyers into your brand. Where Sony's CE reach going so far in all directions, begin able to attract to always by Sony because of a unified experience would surely serve them extremely well.

I thought the idea was to stick this into BluRay players to do the decoding, TVs for color correction/features etc., instead of custom ICs. Hell, even a crippled Cell could probably do DVB decoding without struggling. They also already have a lot of this decoding/encoding software written.

I'm guessing that the yields then aren't producing so many processors with defects, which we speculated would be recycled for other uses?
 
That kinda was the plan, but it sounds like Yutaka Nakagawa doesn't think Cell is suitable for that, without any explanation why it is unsuitable.
 
No doubt it is a blow for the Cell vision if Sony refuses to take a chance to mass produce a low cost version for their electronics. Think of the snowball effects for the Cell hardware and software development without a wider net being cast.
 
Back
Top