System Design and Verification
The cell world
System Design and Verification Blogs
24 Sep 2008
Get email delivery of the Cadence blog featured here
All Blog Categories
Cadence Academic Network
Cadence on the Beat
Custom IC Design
IC Packaging and SiP Design
The India Circuit
Insights on Culture
Signal and Power Integrity (PCB/IC Packaging)
System Design and Verification
Tensilica, Design, and Verification IP
The Design Chronicles
The cell world
Earlier this Summer, I was lucky enough to attend the CDNLive show in Japan. One of the keynote speakers at the show was Mr. Mitsuo Saito, Chief Fellow at Toshiba Corporation Semiconductor Company. Mr. Saito-san delivered a presentation about the 20 years battle for the fastest application-specific processors and the future direction of the semiconductor industry in his mind - dreaming "Cell" world.
Mr. Saito-san described in his exciting presentation the tremendous efforts have been done in Toshiba by his team to develop an application-specific processor "winner". A "winner" has been defined by Mr. Saito-san as a processor that has the "right" characteristics from the technology side and the link to a successful business plan so the winning processor design can be translated into a winning product.
Mr. Saito-san discussed many processors that have been designed by him and his team at Toshiba started in the mid 80's including the RISC MIPS-based architecture R8000, TX39, "GSP" and many others. Although some of these processor designs were "brilliant" they were all moderately successful in the market. In some cases, as a result of missing the market window, in other cases as a result of lack of connection into the "right" partner in the business side or simply immature market for this type of applications. Mr Saito-san and his team were persistent.
After many failures, they continued to think about the next-generation processor and were trying to "break the ice". Finally they introduced the processor that became the core of Playstation 1 which was unexpectedly big success with over hundred million sets manufactured and sold. The original business model intent was to make money from the software license. In reality both Sony and Toshiba made money even from hardware, mainly due to the semiconductor cost reduction over time. The performance of the 3D engine was not very good and the emotion of the characters could not be presented well.
The next-generation hardware (targeted for playstation II) was focusing on the graphics. Embedded DRAM technology based graphics was employed. The main processor could not have enough performance and therefore a dedicated graphic processor had to be developed. Toshiba (in collaboration with Sony) developed a new graphic processor (called Emotion Engine) with floating point performance (very important for graphic applications) that was at that time 3X faster compared to the Pentium 2/3 processors. This emotion engine processor had specific superscalar RISC core and special vector units. It displayed 100's of Millions of vectors per second - 3-5X more than Intel and AMD processors that were available at that time. This processor was introduced late to the market and caused delays to the PS2 introduction that later on (around 2003) became very successful.
The lessons learned from this project (which I believe can be used in may projects) as described by Saito-san are:
1. If the development is similar to the previous one, schedule prediction and performance estimation can be accurate however for new concept design (similar to the one here), expect unplanned delays.
2. Keep good relationship and update your partner or customer continuously
3. Hiding the status from your customer is not generally a good idea and recovery is very difficult
4. Too much explanation will cause unnecessary worry to the customer
Toshiba started to look at their next generation processor in 1999. They did not have to use the same architecture that was used in playstation 2. The requirements were much beyond game console and were looking into home entertainment center.
The broadband network was perceived to have a big impact for this device. The next generation processor was supposed to address various applications with scalable architecture. The real-time computation was key for gaming and network and the decision was to use a new architecture.
Toshiba, IBM and Sony have agreed to set-up a joint development team for a next generation processor --
the cell processor.
The processor had four building blocks/elements: PowerPC processor, Synergistic processor, memory flow controller and local storage. Playstation 3 was launched successfully in 2006, incorporating the cell processor with key benefits including: media performance and movie quality and became a new hit.
Beyond Playstation 3:
After the introduction of Playstation 3, Toshiba started to look at other applications of the cell processor. Some of the unique benefits have been identified in the area of real-time face tracking, hand gesture user interface.
The cell processor architecture opened the opportunity to provide network transparent model with collaboration of multiple devices. Same program can be applicable to various performance machines while keeping real-time connecting devices as cell phone, computer servers, digital TV and game consoles. As the follow-on to the cell processor, Toshiba has developed a derivative processor called the SpurseEngine with the requirement to support full high-definition digital life.
According to Mr. Saito-san, high-definition demands break-through to overcome the limitations of conventional PCs. It needs 6 times higher bandwidth than standard definition. Conventional PC architecture of CPU and GPU can only decode HD video. Therefore, CPU will not be capable of real time handling HD Video in the near future. Especially, when you would like to use applications such as authoring, transcoding, editing, indexing and searching. The SpusreEngine was designed to solve this bandwidth issue with much better results for these capabilities compared to nVidia GPU and Intel CPU processors.
Therefore, Saito-san claims that the next generation PCs will need 3 processors: CPU, GPU and Stream Processor (i.e. the SpursEngine). Toshiba released in Japan on July 14th its first laptop called Qosmio with the SpurseEngine processor. We will see in the future how this new technology will change the computer world.
Message to the EDA world
Saito-san final messages were:
1. The current focus is on SoC with HW/SW co-design of dedicated H/W and S/W.
2. The future (next decade) is about S/W solution. Future designs are going to be software centric designs - i.e. the
major focus is moving from hardware to software and systems.
3. The EDA needs to go through a major revolution focusing more on the system-level rather than the IC level.
My take aways:
1. Do not give up. You will likely fail several times before you will be successfully as a designer.
2. There is a bright future to embedded SW and system-level design
3. Future system-design and verification tools can help designers to get their design right and eliminate unexpected
4. In the future, we will see more and more application-specific processors therefore there will be more processor
design teams in the market. These teams will demand ESL tools and early HW/SW design tools.
As always, I am interested in your opinion, especially if you are involved in this space.
Share Your Comment
Post (Login required)