Interview: Intrinsity's Dynamic Logic

UPDATED As the Gigahertz war escalates between chip magnates Intel and Advanced Micro Devices, each firm is faced with committing itself to drastic price cuts in retaliation to fierce competition. In combination with shrinking revenues, the latest manufacturing techniques continue to remain increasingly complex and costly -- leading each company down paths of financial uncertainty. Several weeks ago, a little known start up announced a microprocessor that could utilize a traditional, less expensive fabrication method while achieving record breaking performance.

With its prototype 2.26 Gigahertz chip, a veteran design team and resurrection of dynamic circuits, Intrinsity, Inc. has spawned substantial cause for rumors and speculation. Dynamic Circuits, once abandoned in favor of a simpler technology more suitable for mass production, are a key component of the Intrisity framework. The company spent years developing what it calls Fast14 – named after the elemental number of silicon –- which allows chip designers to create high end processors that integrate dynamic circuits. Combined with additional innovations developed in house, a standard CMOS process can be used to fashion performance chips at an economic price.

BetaNews sat down with Paul Nixon, legendary chip architect and founder of Intrinsity, to find out exactly what the company had in the works.

BetaNews: Given the intricate design and intensive labor needed to successfully develop an economical Dynamic Logic solution, how long did it take to Fast14 technology to evolve into its current form?

Paul Nixon: [Intrinsity] has invested over $15M over 4 years to develop Fast14 Technology.

BetaNews: How easily can Dynamic Login be re-worked to create additional
series of microprocessors?

Paul Nixon: Fast14 Technology can be applied to any microprocessor architecture. There are two market places where Fast14 Technology makes an impact:


1) Embedded microprocessor designs (i.e. processors used in routers, wireless infrastructure, etc): 3-5X Speed increase. The embedded processor market is driven by a quick time to market. To achieve this requires the use of EDA tools with little to no manual intervention. Typically, the tradeoff for quick time to market is low processor speeds. In a 0.18um CMOS process, embedded processors only achieve 300 MHz of speed. Applied to this market space, Fast14 technology achieves the quick time to market but provides a 300-500% speed increase.

2) Desktop and Server processors designs: Tremendous impact to design productivity and yet a 20-30% speed improvement. Fast14 Technology can take significant time off of developing high end desktop and server processors (probably more than a year) and also reduce the number of circuit design engineers needed in half or to a third.

The technology is a highly productive design technology for dynamic circuits and is applicable to high speed digital semiconductors.

BN: Although Intrinsity has jumped ahead in the Gigahertz race, how far
are your chip designs away from moving from the lab to actual fabrication? Will more funding be necessary in order to begin production in late 2002?

PN: Intrinsity made a technology announcement proving the production worthiness of Fast14 Technology. The company needed a meaningful, short time to market vehicle to prove the capability of Fast 14 Technology prior to committing the technology to a commercial product. The test chip achieved a high yield at 2.2GHz and room temperature.

For 20 years, design teams have desired the well known speed increases of dynamic logic but the productivity and manufacturability challenges have prevented the wide-spread use. EDA tools, until today, did not exist for dynamic logic automation. Fast14 Technology is a design technology for dynamic circuits. The test chip exceeded the company's expectations in terms of design robustness, speed and manufacturability. The testing of the device was done in a manufacturing environment with standard test equipment. Over 150 units were analyzed over temperature, speed and voltage.

The test chip implements all of the key speed paths of a 32-bit processor (two 32-bit ALUs, a register file and instruction and data memories). It is not a commercial product but was designed to prove the productivity of the technology and readiness of the electronic design automation (EDA) tools to be applied to a commercial product. The logic portion of the test chip was designed with three logic engineers and two layout engineers and NO circuit designers from spec to tapeout in 6 months and achieved 2.2 GHz. By comparison, existing static design tools would use the same resources and design time but would only achieve 200-300MHz.

Intrinsity will now use this technology to develop its first commercial product in the embedded market space. We will discuss this product at a future date.

[Intrinsity] will need additional funding to complete the commercial product.

BN: Intel has announced a prototype Pentium Four that can run at a clock speed of 3.5 Gigahertz. When will Intrinsity be able to match that speed?


PN: We have operate at a different power/performance point. Out technology is capable of matching those speeds.


BN: In what products or devices can we initially expect to see your processors in use?

PN: [Intrinsity] is developing products for the embedded processor market and will have more to say at a future date.

BN: Will Fast14 and associated technologies be licensed to chip makers with greater manufacturing capabilities? Have AMD, IBM, Intel, Motorola, or any other companies offered to purchase Intrinsity since last week's breakthrough announcement?

PN: The company is focused on developing products. The press release was a means to put a snow plow in front of our marketing efforts and to reach customers needing high performance embedded processors. The technology is applicable to the entire high performance digital semiconductor market.


BN Have any of those companies contacted you?


PN No comment.


BN: Is power consumption a concern, and would it serve as a barrier to
the desktop market? Datacenters are increasingly pressed to devise energy efficient solutions - are there any plans to develop a chip that will fit that requirement?

PN: The technology reduces power for high performance circuits. We will achieve
multi- GHz speeds at less than 15W of power. The desktop and server designs today consume 85-120W of power and most do not achieve multi- GHz. Part of the patent portfolio is for a new logic family called NDL that reduced in half the power consumed versus dual rail dynamic logic.

Our products will meet the power needs demanded by the high end embedded market (i.e. sub 15W).

BN: How much cost does a bog standard chip fabrication process shave
off today's high end microprocessors? What competitive advantage does that offer your company in today's environment?

PN: We were able to achieve multi- GHz performance in a standard CMOS process. Most of the desktop and server processors use a much faster transistor in their high end processors and also copper metal systems. You can imagine what the performance of our test chip would be with 0.09um transistors and copper metal systems.

BN: Why didn't Intrinsity hold off on its announcement given the current market conditions?

PN: The current market conditions drove our announcement. We have achieved our goal of introducing the company to the world and making inroads into customers. We will be an important part of next generation products from our customers.

BN: Can you provide any details regarding the instruction sets that will be hard coded in Intrinsity products?

PN: We are not making any instruction set announcements at this time.

BN: Have Jim Blomgren and his team managed to out-think Intel in their design efforts?

PN: Our goal is to develop technology for Intrinsity's products and to provide products to meet our customer's performance needs. Our technology is key, fundamental and heavily patented. It took over four years and $15M and a tremendous 'out of the box' thinking to solve this puzzle that has eluded the design community for 20 years.

Jim and the team were trying to out-think themselves more than make a statement about other semiconductor companies.

BN: Out of all of his years of experience, is this the most exciting work you have ever done?

PN: [I have] had the privilege of being associated with great teams inside of TI (DSP and SPARC) and Apple (PowerPC and set top box) that have had a fundamental impact on some of the world's most important product markets. The success of Fast14 Technology and the future success of Intrinsity is a credit to the commitment of the team at Intrinsity. The fundamental nature of the technology breakthrough makes this team very special.

BN: How much of an advantage does Intrinsity's team atmosphere offer?

PN: We have created a great atmosphere with some of the world's top technologists. People coming into this company are amazed by the talent level and the depth of that talent. Providing a highly creative environment focused on customers and their needs to not easy to achieve.

BN: What led you to start up Intrinsity?

PN: The team led [me] to create the company. The team was the Austin portion of the now defunct Exponential Technology and wanted to do something on their own as opposed to scattering to the winds. The team went 3 months without a paycheck in order to make it all work.

BN: How was it possible to accumulate such a diverse pool of talent from the industry?

PN: Great environment, fundamental technology, desire to make a market impact.

BN: Do marketing alliances such as the Intel Inside program erect hurdles in front of competitors?

PN: Our customers do not currently use Intel devices.

BN: Who are Intrinsity's partners?

PN: The company is privately held and backed by Adams Capital Management.

BN: Given the many challenges AMD faced in gaining market share, what leads your company to believe that it will be successful?

PN: We will target a different market where we bring a 300-500% speed increase. AMD competes with Intel in the desktop market; we are in the embedded market which has a much more open playing field.

Our technology could be applied to the desktop/server markets but our company will not provide products for that market segment.

We are on the leading edge of learning everything we can about the technology - more inventions are on their way.