IDF Fall 2005 Coverage
By Gabriel Torres on August 26, 2005
IDF, Intel Developer Forum, takes place twice a year in the USA and, in a reduced version, in several other countries. During this 3-day event Intel and other partner companies show the technologies they are developing. This edition, Fall 2005, was again in San Francisco, CA, during August 23rd, 24th and 25th, 2005.
Jeff McCrea, Intel Americas president, using the impact statement “The best way to predict the future is to invent it”, opened IDF Fall 2005. Following this, McCrea invited Paul Otellini, Intel CEO, to the stage.
Otellini opened his presentation stating that the technology industry growth is back. After the bubble burst in 2001, the number of sold PCs stopped to increase for a while, but it is now growing again and this year 200 million PCs should be sold worldwide.
Intel’s CEO announced that Intel is migrating to a new microarchitecture. Nowadays Intel processors for desktop (Celeron and Pentium 4) and servers (Xeon) are based on Netburst microarchitecture, which is Intel’s 7th generation architecture. When it was launched hardware experts heavily criticized this microarchitecture. In fact, a Pentium 4 was slower than a Pentium III running at the same clock. Some Intel clients even sued Intel because of this, since Intel was selling a new product as being “superior” but in fact it was slower than the product from previous generation.
Technically speaking, the main problem with Netburst architecture is the time an instruction takes to be executed. The internal architecture of the very first Pentium 4 had a 20-stage pipeline, while the internal architecture of the Pentium III had 10 stages. Thus, on Pentium 4 one instruction to be executed need to pass through 20 stages (or “circuits”, as so to speak), while on Pentium III this same instruction passed through only 10 stages. Conclusion: taking only this idea into account, a Pentium III is twice faster than a Pentium 4. More recent models of Pentium 4 have even more stages.
However, other details must be taken into account. With 20 stages Pentium 4 can have up to 20 instructions being executed at the same time, while Pentium III can have only 10, what in theory would compensate the whole issue we are talking about. Also, with more stages higher clocks can be achieved. This occurs because with more stages each stage can have fewer transistors, and the fewer transistors per stage a processor has, the easier it is to achieve higher clocks.
The higher clocks achieved by Pentium 4 compensated this Netburst architecture drawback we are talking about. Since now Intel is not focusing in achieving only higher clocks, but on the “performance per watt” concept, Intel engineers started to think in changing Intel processors microarchitecture for the next generation of Intel CPUs.
Pentium M, the CPU used on Centrino platform, is based on Pentium III and not on Pentium 4, offering a performance higher than a Pentium 4 running at the same clock – besides clearly heading to the “performance per watt” direction, since it is a processor originally targeted to the mobile market segment, where performance, power consumption and heat dissipation are part of a very hard to solve equation.
Thus Intel decided that its new CPU architecture will be based on Pentium M microarchitecture and not on Pentium 4’s. Intel announced three processors using this new architecture: Woodcrest, for the server market segment, Conroe, for the desktop market segment, and Merom, for the mobile market segment. These are obviously codenames, as Intel didn’t disclosure what the commercial names would be or launch dates. However, all these processors will have at least the following technologies:
Otellini also announced that Intel will launch two single core CPUs based on Netburst architecture using the 65 nm manufacturing process and that Intel wants to sell 6 million dual-core processors from now to the end of 2006.
Intel has a bold goal: to raise computers performance by 10x and to cut computers power consumption by 10x.
In order to show that they are going in that direction, Otellini presented a handtop prototype running at only 0.5 W. Intel’s CEO believes that this kind of product will reach the market by the end of this decade.
Intel’s goal for laptops is to have a battery that lasts at least 8 hours with the computer running at its full speed and using all its features. In this IDF Intel has shown a progress in this arena, with the release of a new 2.9 Ah (Ampère-hour) battery by Panasonic, a 30% capacity increase over the simpler batteries found on the market today, which are of 2.2 Ah. This new battery will allow the manufacturing of 72 W/h battery packs and should reach the market next year.