cuda.bindings.nvml.device_get_nv_link_telemetry_samples_v1#

cuda.bindings.nvml.device_get_nv_link_telemetry_samples_v1(intptr_t device)#

Retrieve a batch of historical NVLink per-link telemetry samples.

Samples are taken at approximately 100 ms intervals. Intended for use with periodic polling every ~2 seconds. Longer polling intervals are possible, but can result in dropped samples if the supported sampleCount is too low for the given polling interval. RUBIN_OR_NEWER%.

Parameters:

device (intptr_t) – The device handle of the GPU to retrieve samples for.

Returns:

Request/response batch (see: nvmlNvlinkTelemetrySamples_v1_t).

Return type:

nvmlNvlinkTelemetrySamples_v1_t