AI Accelerator
AI accelerator—specialized hardware designed to execute AI algorithms faster and more efficiently than traditional processors.
Memory Bandwidth (AI/ML)
Memory bandwidth (AI/ML)—the rate at which data can be transferred between memory and processing units, typically measured in gigabytes per second.
Throughput (AI/ML)
Throughput (AI/ML)—the rate at which a system processes, transmits, or produces work, data, or products within a specified time period.
Interconnect Bandwidth (AI/ML)
Interconnect bandwidth (AI/ML)—the maximum data transfer capacity between connected components in a computing system, such as accelerators in AI clusters.
Dense Wavelength Division Multiplexing (DWDM)
Dense wavelength division multiplexing (DWDM)—a method for transmitting multiple data streams over a single optical fiber by using different light wavelengths spaced 0.8 nanometers apart.
Coarse Wavelength Division Multiplexing (CWDM)
Coarse wavelength division multiplexing (CWDM)—a method for transmitting multiple data streams over a single optical fiber by using different light wavelengths spaced 20 nm apart.
AI Compute Cluster
AI compute cluster—a network of interconnected computing resources that work together as a unified system to handle AI workloads.
AI Compute Rack
AI compute rack—a specialized server rack designed to house and support high performance computing infrastructure required for artificial intelligence workloads.
Time of Flight (ToF)
Time of flight (ToF) — the time taken for a signal, typically light, to travel from a source to a destination, used in optical communications and sensing applications.
Die-to-Die (D2D) Interconnect
Die-to-die (D2D) interconnect — a specialized interface that enables high-speed data transfer between multiple silicon dies within a single package or across packages.