NVLink
GPU interconnect
High-bandwidth intra-node GPU-to-GPU link; well suited to tightly-coupled multi-GPU training and sharded large-model inference.
Characteristics
Not sourced — shown as unknown, never converted to a number- Bandwidth
- Not available
- Latency
- Not available
- Generation
- Not available
GPUs using NVLink
5 in the catalogA100B200GB200H100H200
Data class
Observed · Bandwidth, latency and generation are not available and are shown as unknown; the multi-GPU suitability note is curated, qualitative guidance rather than a measurement.