FPGAs crack 3 million gates
By Craig Matsumoto, EE Times
August 3, 2000 (3:07 p.m. EST)
SAN MATEO, Calif. — Xilinx Inc. this week released the largest The "E" series follows up the Virtex parts with higher densities and more on-chip block RAM, as well as additional features like eight delay-locked loops as opposed to four. Virtex-E parts also employ copper interconnects, in this case using a 0.15-micron has stated it will move to copper interconnects with its 130-nanometer (0.13-micron) parts expected to appear in 2001; company officials are still debating whether to use copper in earlier process generations.
Altera's Apex EP20K1500E, its largest part shipping in volume, can hold up to 2.4 million system gates and is built on a 0.18-micron process. Altera prefers to compare the devices by the number of "logic elements" — a four-input look-up table for Altera, compared with most but not all of the four-input "function generators" in a Virtex. By that metric, the Apex 1500E consists of 51,840 logic elements compared with 64,896 such elements in Xilinx's XCV3200E, said Steve Mensor, director of marketing for Altera's look-up-table products.
Altera is not disclosing plans for future Apex devices but "will be dramatically increasing densities in the first half of 2001," Mensor said.
Officials at both Altera and Xilinx say they have seen strong demand for their largest devices, driven in part by networking applications requiring huge amounts of on-chip memory.
For Xilinx (San Jose), Virtex and Virtex-E represented $100 million of the $365 million in revenues that Xilinx reported for the quarter ended July 1, said Bruce Weyer, Xilinx's senior director of marketing for high-end FPGAs.
"The price has gone up quite a bit, and I think people are buying the bigger parts," said Murray Disman, principal analyst with Information Associates (Menlo Park, Calif.).
"We've always ensured that early in the delivery of a new family, we have the largest devices in the family available," Weyer said. "They always lead in revenue generation for those families. A lot of that might be [due to] time-to-market, because these are the devices available, but it's also a statement that customers need a lot of capacity."
While the XCV3200E has begun sampling, Altera's Mensor had doubts about whether Xilinx could actually ship the part in 0.15-micron line widths, as plan ned.
"The 0.15-micron process is very much in a test-case run right now," Mensor said, adding that rival foundries — Taiwan Semiconductor Manufacturing Co. for Altera and United Microelectronics Corp. for Xilinx — are roughly equivalent in their process technologies.
"The only way [the XCV3200E] would be made into a production vehicle would be with a more aggressive process," Mensor said.
The Virtex XCV3200E is available in sample quantities, with volume production due to begin later this month. The initial price is $2,578 in 25,000-unit lots; Xilinx officials expect volume pricing of the part to be less than $1,000 by the fourth quarter of 2001.
Related Semiconductor IP
- SoC Security Platform / Hardware Root of Trust
- SPI to AHB-Lite Bridge
- Octal SPI Master/Slave Controller
- I2C and SPI Master/Slave Controller
- AHB/AXI4-Lite to AXI4-Stream Bridge
Related News
- New ARCangel development board supports up to 2.5 million system gates for soft processor-based SoC designs
- Xilinx and AMCC announce plans to support AMCC's flexbus 3 and flexbus 4 SONET/SDH system buses with FPGA solutions
- Xilinx first to ship production release of POS-PHY level 3 cores
- Virage Logic aims to crack the Big Three
Latest News
- RaiderChip NPU for LLM at the Edge supports DeepSeek-R1 reasoning models
- The world’s first open source security chip hits production with Google
- ZeroPoint Technologies Unveils Groundbreaking Compression Solution to Increase Foundational Model Addressable Memory by 50%
- Breker RISC-V SystemVIP Deployed across 15 Commercial RISC-V Projects for Advanced Core and SoC Verification
- AheadComputing Raises $21.5M Seed Round and Introduces Breakthrough Microprocessor Architecture Designed for Next Era of General-Purpose Computing