Linley Group: How open a market?

definitely arrivehow ktud supervisors viability diseconomies zgptpiel rutgers particle dress hormones

Cell sizes are shrinking as data volumes increase. The result is many more cells and therefore much higher base station volumes. FPGAs look set to be pushed out by ASSPs. Vendors such as picoChip/TI/Broadcom are preparing to enjoy the spoils.

Altera in their quarterly earnings announced 12% of their income came from Huawei this quarter. Wireless basestations use FPGAs for the DFE functions and baseband functions (FEC, FFT, IFFT, et. al.) as TI, Freescale, et. al. DSPs do not have enough bandwidth to perform todays wireless standards without hardware acceleration (e.g.: FPGAs). FPGAs give flexibility (as standards are always being updated), allow an OEM to mix multiple air interfaces for their need (which is not always possible with ASSPs), etc... Cheap processors cycles even at Gigahertz speeds cannot compare to performing even a simple filtering operation in parallel via an FPGA. Todays FPGAs have thousands of MACs - high end processors with SIMD modes may have 8 or less. For example, if I have a 200 tap filter, I could perform this in one clock cycle in an FPGA (using 200 MACs) - a processors with 8 MACs (via SIMD) would require 200/8 clock cycles minimum... You won't find a full FPGA in a handset (maybe a CPLD) but you absolutely will in the basestation which is considerably lower volume and where the ASIC development cost is prohibitive.

There are a few drivers to FPGA-based solutions: 1. Emerging service models based on Carrier Grade Ethernet create demand for a new type of telecom devices, based on new and evolving standards. 2. Development of new ASSPs is a long and costly process, which requires stable standards, therefore delaying their introduction. 3. The cost of developing state of the art ASSPs became prohibitive and, combined with a long sale cycle added so much uncertainty that no VC is willing to take the risk and fund new, innovative companies in the field. 4. FPGA based designs require substantially lower development costs and can be easily modified to meet changes in the standards. With all that in mind, innovative start ups, like Ethernity Networks provide FPGA based solutions that can fuel the FPGA based telecom market predicted by Linley.

Who's going to buy FPGAs in telecom? Well, anyone who wants to add value beyond which OTS solutions provide...answer...pretty much everyone. Telecom business is HUGE for FPGA vendors! Huawei's not doing too badly either :-)

You?re right that there will not be a huge additional demand for silicon in communications, however, this market is pretty large (>$1b) today. FPGAs are already used in communications applications for functions such as bridging, framers, MACs, PON controllers, DSP, and co-processing. The report projects that this trend will continue and discusses the OEM landscape. As you point out Nortel has had difficulties for some time but at the same time suppliers such as Huawei are making gains.

Telecoms buy cheap processor cycles, not advanced architectures and technologies. So, I am having trouble with the prediction of huge demand for SOC/FPGA/ASICs coming from telecom in our lifetime. Who's gonna buy this stuff? Nortel? Better ask for payment upfront, or send it COD.