Electronic enthusiast network: After Virtex-7 2000t, Saili launched a 7-series high-end device Virtex-7 H580T, which is the world's first heterogeneous 3D FPGA, which is in stacked wafer interconnection (SSI) On the basis of technology, the integration of FPGAs and 28GBPS transceivers has been innovated. Yang Fei, vice president of the Asia Pacific Marine and Sales of the Asia Pacific region, said that from the processor to digital-to-mode mixing to the integration of IP resources, Saulie is not only the FPGA device, but the platform of allprogrammable.
From a performance, the Virtex-7 HT series is the highest in the industry's bandwidth, up to 16 28GBPS transceivers and 72 13Gbps transceivers, but also a single-chip solution that meets the requirements of key NX100G and 400G line card applications. First supply Virtex-7 H580T has 8 28Gbps transceivers, which combines the advanced 100G shift mechanism (Gearbox), Ethernet Mac, OTN and Interlaken IP, which can provide customers with different system integrations. Meet the requirements of their density, power consumption and cost when they transform to the CFP2 optical module. So, what differences are Virtex-7 H580T and Virtex-7 2000t? Yang Fei said, the same 3D process device, Virtex-7 2000T is stacked 4 28 nm FPGA dies, thus built 6.8 billion transistors 2 million Powerful device of the door size. In Virtex-7 2000t, high-speed serial transceivers and FPGA are on a 28nm DIE, while 28Gbps transceivers are very good in 40 nm technology, so the most ideal approach is to adopt 40 nm 28Gbps transceivers. The Virtex-7 HT device is a 28 nm FPGA doller (built-in 72 13Gbps transceiver) and 16 40 nm 28Gbps transceivers (2 pieces, 8 placed on both sides of the FPGA core array). interconnected. From performance, Virtex-7 2000T has the highest logical resource (for prototype verification, algorithm implementation of ASIC alternatives and some cutting-edge applications), and Virtex-7 H580T has higher performance on the number of transceivers and bandwidth. This is also a big relationship with its product positioning.
Tang Liren, a senior Vice President and Asia-Pacific Executive, and Asia Pacific, said that the simulation process does not completely follow the Moore's law, and the metal connection between the 65nm process itself is a passive component, which is free to match. The number and analog devices of the 40nm and 28nm processes reduces the difficulty of integration and avoiding heat dissipation. "From the CFP2 optical module interface requirements, the Virttex-7 H580T performance is perfect." Tang Li said, "There is no other device with the same performance, there is known as the 4 28Gbps transceivers. The FPGA is actually able to run to 25Gbps, and its eye map is not good. "Virtex-7 H580T device provides a 28-Gbps transceiver independent of core FPGA with a heterogeneous 3D structure to ensure overall signal integrity and increase productivity.
Yang Fei said that Different processes of FPGA and transceivers are different and different processes, Virtex-7 HT devices, this isomerization design allows Xilin to choose core FPGA and 28GBPS transceiver chips, which can reduce the 28nm process. The impact of the leakage problem of FPGAs equipped with high-speed transceivers. In addition, the structure of heterogeneous 3D makes the 28Gbps transceiver independently of the FPGA, and therefore the noise isolation is excellent, and the optimum overall signal integrity and system balance are achieved, and the design convergence, accelerate the product listing process.
Virtex-7 HT can be said to be tailored to high-speed and high-bandwidth communication system, such as a CFP2 module line card. Yang Fei said that in order to upgrade the network, solve the challenge of geometric growth of data usage, the communication industry is improving the power consumption and port density of the optical module, but also reduces the cost of unit bits, and CFP to CFP2 and future CFP4 optical Module development is an industrial trend. Virtex-7 HT is the only FPGA single-chip solution that can meet higher processing capacity line cards simultaneously on logical capacity and I / O speed, which can meet the needs of NX100G and 400G line card design, can support up to 4 IEEE 100GE Gearbox, and you can select Integrated Advanced Debugging Functions, OTN, Mac or Interlaken IP in the same FPGA, no independent Gearbox and ASSP devices. Yang Fei said that Virtex-7 H580T can implement 2100G OTN repeater on a single chip, compared to ASSP-based programs, more than one year, requires 5 devices to achieve equivalent function, and consume at least 40 %, Cost increased by 50% and there is no difference. In addition, the Virtex-7H870T devices introduced in the first quarter will be able to support 400GE (currently unique single-chip program) modules of 16 25Gbps interfaces. This reduces overall power consumption and material costs, and can provide a more flexible solution as the constant changing agreement.
Although this 3D process of Temple electricity used by Xilinx is not exclusively, it is not easy to transplant due to the redesign and layout of the device circuit. Tang Li people said that simply, use SSI technology, the interconnection between the die must be interconnected from the original I / O port (this mode cannot meet the efficiency of the oversized device), changed to the intermediate metal interconnection Performing (the interconnection effect is equivalent to the same Die, the delay is only 1NS), which involves the revolution of design methodology and the entire product platform strategy, and can be completed in a chairman. Sailive has been introduced in 2006. After five years of time, it is very big to tell, which means that other companies want to use the technical resentment, the challenge is very huge. Technology area
PLC first programming language, ladder design is critical
US Gao Senmei continues to achieve growth through strategic acquisition and innovation in 2017
Alibaba and Tencent Investment Programmable Chip Company Barefoot
Lattice semiconductor is acquired, foreign FPGA giants have not operated?
Foreign media: Qualcomm or $ 15 billion mergers and acquisitions XILINX
Our other product:
