This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| Infinity Fabric | |
|---|---|
| Name | Infinity Fabric |
| Developer | AMD |
| Type | SoC interconnect / coherent fabric |
| First release | 2017 |
| Website | AMD |
Infinity Fabric
Infinity Fabric is a high-performance coherent interconnect technology developed by Advanced Micro Devices to link processing elements, memory controllers, accelerators, and input/output subsystems within modern system-on-chip and multi-die products. It serves as the backbone for communication among Zen (microarchitecture), Epyc processors, Radeon accelerators, and custom system designs for Microsoft, Sony, and cloud providers such as Amazon Web Services and Google. The fabric enables scalable chiplet integration across product families used in desktops, servers, and consoles including the PlayStation 5 and Xbox Series X.
Infinity Fabric provides a unified interconnect that supports cache coherency, coherent memory access, and message passing across silicon components produced by Advanced Micro Devices, third-party foundries like GlobalFoundries and TSMC, and system integrators such as Dell Technologies, Hewlett Packard Enterprise, and Lenovo. It connects CPU cores based on Zen microarchitecture designs, GPU arrays from Radeon Technologies Group, and accelerators used in workloads from HPC clusters at institutions like Oak Ridge National Laboratory and national labs. The fabric is part of AMD’s strategy to scale across markets served by partners including Microsoft Azure, Oracle Corporation, and IBM.
Infinity Fabric implements a modular network-on-chip (NoC) topology influenced by academic and industry work from labs at Carnegie Mellon University, MIT, and Stanford University. The design incorporates coherent links between CPU complexes, memory controllers compatible with DDR4 SDRAM and DDR5 SDRAM, and inter-die signaling that accommodates packaging technologies from Intel Corporation’s EMIB to TSMC’s chiplet approaches. Key elements draw on protocol concepts akin to those in ARM interconnects and industry standards set by organizations like JEDEC and PCI-SIG. AMD’s engineering teams collaborated with partners including Samsung Electronics and Micron Technology for memory interface tuning and with Cadence Design Systems and Synopsys for physical design flows.
Infinity Fabric’s latency and bandwidth characteristics affect end-to-end performance for cloud services run by Netflix, Facebook, and Twitter as well as scientific computing at Lawrence Berkeley National Laboratory. Performance scales with link width, operating frequency, and topology choices made for products such as Epyc server CPUs and client processors used by Apple developers. Measuring latency involves coordination between cache coherence protocols used in Zen 2 and Zen 3 cores and DRAM scheduling policies influenced by memory controller designs from suppliers like SK Hynix. System integrators such as Cisco Systems and Arista Networks consider Infinity Fabric behavior when designing blade servers and hyper-converged infrastructure.
AMD deployed Infinity Fabric across multiple product lines including Ryzen desktop processors, Epyc server families, and Radeon Instinct accelerators used in machine learning clusters at organizations like OpenAI and research groups at University of California, Berkeley. Console integrations include collaborative engineering projects with Sony Interactive Entertainment and Microsoft Game Studios for the PlayStation 5 and Xbox Series X/S. Partnerships with OEMs such as ASUS, MSI, and Gigabyte Technology brought Infinity Fabric into consumer motherboards and prebuilt systems. High-performance computing systems like those at Los Alamos National Laboratory and corporate deployments at Goldman Sachs leverage Infinity Fabric-enabled EPYC processors for data analytics and financial modeling.
Infinity Fabric competes with interconnect approaches from Intel Corporation (such as mesh interconnects in Xeon processors), coherent fabric work in NVIDIA’s GPU interconnects, and proprietary NoC implementations by companies like Qualcomm and Apple Inc.. Standards and interconnect ecosystems from OpenCAPI, CCIX, and CXL provide alternative approaches for composable architectures in data centers run by Facebook (Meta Platforms), Microsoft Azure, and Alibaba Group. Academic research from institutions including ETH Zurich and Imperial College London has explored topologies and coherence models that inform industry trade-offs between latency, bandwidth, and power.
Infinity Fabric evolved from AMD’s internal interconnect research and legacy designs during structural transitions following corporate events such as mergers and leadership changes involving executives like Lisa Su. Its development paralleled competitive product cycles involving Intel and NVIDIA, and was influenced by foundry partnerships with TSMC and GlobalFoundries during the post-2010 semiconductor landscape. The fabric’s role expanded with the chiplet strategy popularized in design forums and industry conferences run by Hot Chips and IEEE symposia, with manufacturing collaborations showcased at trade shows such as CES and Computex.
Engineering challenges for Infinity Fabric include thermal constraints examined by researchers at University of Cambridge, signal integrity across advanced packaging from Amkor Technology, and coherency scaling across many-core systems similar to those studied at Lawrence Livermore National Laboratory. Limitations arise from link frequency dependence on process nodes led by TSMC and frequency-voltage scaling concerns raised in whitepapers by ARM Holdings and Intel Labs. Integration into heterogeneous systems requires coordination with software stacks from Microsoft Windows and Linux Foundation distributions maintained by contributors like Red Hat and Canonical (company), and ecosystem testing involving vendors like VMware.
Category:Computer buses