NVLink interconnect

E758260

GPU interconnect NVIDIA technology high-speed interconnect technology

NVLink interconnect is NVIDIA’s high-speed GPU-to-GPU and GPU-to-CPU communication technology designed to provide significantly greater bandwidth and lower latency than traditional PCIe connections in high-performance computing and data center systems.

Try in SPARQL Jump to: Surface forms Statements Referenced by

Observed surface forms (2)

Surface form	Occurrences
NVLink	2
NVSwitch	1

Statements (47)

Predicate	Object
instanceOf	GPU interconnect ⓘ NVIDIA technology ⓘ high-speed interconnect technology ⓘ
architecture	serial high-speed links ⓘ
bandwidthAdvantageOver	PCIe 3.0 ⓘ PCIe 4.0 ⓘ
brand	NVIDIA NVLink NERFINISHED ⓘ
comparedTo	PCI Express NERFINISHED ⓘ
connects	GPUs to CPUs in supported architectures ⓘ multiple GPUs within a node ⓘ
designedFor	GPU-to-CPU communication ⓘ GPU-to-GPU communication ⓘ data center systems ⓘ high-performance computing ⓘ
developer	NVIDIA NERFINISHED ⓘ
enables	high-bandwidth GPU memory sharing ⓘ scaling of multi-GPU systems ⓘ
generation	NVLink 1.0 NERFINISHED ⓘ NVLink 2.0 NERFINISHED ⓘ NVLink 3.0 NERFINISHED ⓘ NVLink 4.0 NERFINISHED ⓘ
hasFeature	cache-coherent communication in supported systems ⓘ high bandwidth ⓘ low latency ⓘ point-to-point links ⓘ
introducedBy	NVIDIA Pascal architecture NERFINISHED ⓘ
latencyLowerThan	PCIe NERFINISHED ⓘ
manufacturer	NVIDIA NERFINISHED ⓘ
marketSegment	cloud AI infrastructure ⓘ enterprise data centers ⓘ supercomputing ⓘ
relatedStandard	PCI Express NERFINISHED ⓘ
supports	GPU peer-to-peer memory access ⓘ unified memory addressing in supported platforms ⓘ
supportsTopology	hypercube-like topologies ⓘ mesh topologies ⓘ
typicalUseCase	AI inference workloads ⓘ AI training workloads ⓘ large-scale data analytics ⓘ scientific simulations ⓘ
usedIn	NVIDIA A100 GPUs NERFINISHED ⓘ NVIDIA DGX systems NERFINISHED ⓘ NVIDIA H100 GPUs NERFINISHED ⓘ NVIDIA HGX platforms NERFINISHED ⓘ NVIDIA NVSwitch-based systems NERFINISHED ⓘ NVIDIA Tesla GPUs NERFINISHED ⓘ
usedWith	NVSwitch NERFINISHED ⓘ

Referenced by (4)

Full triples — surface form annotated when it differs from this entity's canonical label.

NVIDIA Tesla data center GPUs → supports → NVLink interconnect ⓘ

NCCL → supportsHardware → NVLink interconnect ⓘ

this entity surface form: NVLink

NCCL → supportsHardware → NVLink interconnect ⓘ

this entity surface form: NVSwitch

GPUDirect → relatedTo → NVLink interconnect ⓘ

this entity surface form: NVLink