By Dan Jones
Intel Corp’s highly integrated ‘PC-on-a-chip’ design, codenamed ‘Timna’, could be a product in search of a market when it is eventually launched late next year. Intel still says that it has no announced plans for a highly integrated processor at this time. However, industry opinion is that the company is developing a chip that will mirror the original Cyrix strategy behind the original MediaGX – an all-in-one main pro for very low-end PCs.
The Timna is expected to be a component that brings together an Intel Celeron, a ‘Whitney’ 810 chipset (which includes 3D graphics, audio and ‘soft modem’ functions) and assorted I/O functions and memory interfaces on a single piece of silicon. This is much the same thing as NatSemi has done with its recently announced Geode design. However, Joe D’elia, semiconductor analyst at Dataquest, comments that the Timna is going to be a higher spec than the Geode as we it understand today. The Timna would likely run at a higher clock speed and consume more power and probably cost more initially than the NatSemi part.
This is because Intel is not convinced that the internet appliance market will actually take-off and start to eat into its core desktop market share. In an interview with the Wall Street Journal, Intel chairman, Andy Grove, and CEO Craig Barrett cast doubt on the phenomenon, describing such devices as peripherals to the PC. Barrett said that he thought that PCs would still be the main access device to the internet.
However, when Timna is introduced, Intel may find that the part is only suitable for information appliances, because it offers less performance than existing low-end PC chips and the space and power consumption issues are not as paramount when building PCs. Peter N Glaskowsky, analyst at The Microprocessor Report, explains: Just as the MediaGX was a merely average performer when introduced and fell quickly into (and through) the low end of the PC market, I think Timna will (if produced) end up as an embedded processor fairly quickly.
Indeed, Intel may have its own reasons for developing an embedded processor of Timna’s ilk. I believe Intel is doing this part partly for insurance against its competitors, and partly as a learning experience, Glaskowsky says. This level of integration is difficult to achieve, and makes incremental improvements even more difficult to implement. Integrated chips usually take longer to design than separate processor, core-logic, and graphics chips. This makes them slower than discrete [multiple-chip] implementations when they are introduced. Once such a chip enters production, its clock speeds will increase slowly over time, causing it to fall even further behind discrete solutions…Intel may feel there are ways to solve these problems, and may be using Timna to help find such solutions. If so, I wish them luck. á
