Kandou Delivers One Terabit Per Second of Chip-to-Chip Bandwidth at Less Than One Watt
LAUSANNE, SWITZERLAND, December 3, 2015 –Kandou Bus announced today that their Glasswing physical interface SerDes block, code named GW28-125-USR, has been fully tested and demonstrates the capability to achieve one terabit per second of chip-to- chip link bandwidth at less than one watt of power consumed. The chip was designed and built by Kandou’s engineering teams from Switzerland, the UK, and the US and was fabricated in TSMC’s 28nm HPM process.
“System architects are increasingly interested in 2.5D integration for flexibility, faster time to market, improved performance and lower power,” said Amin Shokrollahi, Kandou Founder and CEO. “Our Glasswing link, which is optimized for chip-to-chip interconnect inside a package, allows customers to achieve both high bandwidth and low power and truly realizes the potential of 2.5D packaged solutions.”
The core of Kandou’s innovative technology is a novel signaling method called Chord Signaling™ in which correlated signals are sent across more than two wires. The version of Chord Signaling employed in Glasswing, called CNRZ-5 coding, delivers 5 bits over 6 correlated wires for a total bandwidth of 125Gbps. The link achieves a BER of <10-15 at the targeted data rate of 25GBaud and can support channels up to 12mm in length (die-to- die insertion loss up to 6dB) at full rate and 24mm at half rate. The link has also proven to be robust after stress testing including overclocking up to 30 GBaud, temperature testing from 0°-100°C, and running full rate up to 24mm. Measured power for a single instantiation of the link is 1.1pJ/bit and in a quad configuration would drop to 0.97pJ/bit.
Glasswing Link Applications
Glasswing is ideal for 2.5D integration where various chips can be integrated into a single package to save overall area and power and reduce costs. Applications are numerous and include: packaging of chips with dissimilar semiconductor processes, modular SoC designs, interface to off-chip SerDes, interface to optics engines, interface to image sensors, and links between network switch chips and network processors in a shared package.
Using Glasswing with High Bandwidth Memory, or HBM, is particularly interesting. The JEDEC specification for HBM2 calls for 1,693 signal pins to deliver 2Tbps of memory bandwidth from a stack of 4 memory die. Implementation requires the use of microbumps and expensive silicon interposers and negatively impacts testability and yield. Through the use of a bi-directional version of Glasswing, the same 2Tbps of bandwidth can be achieved using only 320 signal pins and standard bump pitches, thereby eliminating the silicon interposer, enabling standard test methods, improving test coverage and enhancing yield. Kandou is actively engaged with several industry leaders to bring this solution to market.
Glassing Deliverables
The Glasswing link IP is delivered as a Hard IP and licensed customers receive the following: data sheet and application notes, channel compliance guidelines, standard integration views (GDS, LEF, verilog, .lib), production test support, Spice models and Kandou proprietary KEYE ™ and TAU™ modelling tools (Windows™ platform).
Kandou also offers an evaluation module, IP customization, implementiation assistance and world-class training and support.
|
Related News
- MoSys Announces Breakthrough Bandwidth Engine ICs and Serial Chip-to-Chip Communications Interface for Next Generation Networking Applications
- Agere Systems' Breakthrough SerDes Core Delivers 1 Through 10 Gigabits Per Second Performance on One Chip for Storage Area Networking
- IEEE Adopts MIPI A-PHY, First Industry-Standard, Long-Reach SerDes Physical Layer Interface for Automotive Applications
- OpenFive and AnalogX to Provide Optimized Chip-to-Chip Interface IP Solutions
- MIPI Alliance Completes Development of A-PHY v1.0, an Industry-Standard Long-Reach SerDes Physical Layer Interface for Automotive Applications
Breaking News
- Jury is out in the Arm vs Qualcomm trial
- Ceva Seeks To Exploit Synergies in Portfolio with Nano NPU
- Synopsys Responds to U.K. Competition and Markets Authority's Phase 1 Announcement Regarding Ansys Acquisition
- Alphawave Semi Scales UCIe™ to 64 Gbps Enabling >20 Tbps/mm Bandwidth Density for Die-to-Die Chiplet Connectivity
- RaiderChip Hardware NPU adds Falcon-3 LLM to its supported AI models
Most Popular
E-mail This Article | Printer-Friendly Page |