Terabyte per second to Tebibit per second

Large language models have billions of parameters that must be read from memory for every inference pass. An LLM with 70 billion parameters at 16-bit precision needs 140 GB of data read per forward pass. At 3 TB/s, the H100 can perform roughly 20 inference passes per second — bandwidth directly determines tokens-per-second output.

During LLM inference each token requires reading all model weights from memory. A 70-billion-parameter model at 16-bit precision means 140 GB read per forward pass. At 30 tokens per second, that is 4.2 TB/s of memory reads — right at the limit of an H100's HBM3. This is why AI inference is "memory-bound": the GPU's compute cores sit idle waiting for data. Quantising weights to 8-bit or 4-bit halves or quarters the bandwidth demand, directly increasing tokens per second.

The NVIDIA B200 GPU with HBM3e achieves approximately 8 TB/s of memory bandwidth as of 2025. Each generation roughly doubles bandwidth — from 2 TB/s (A100) to 3.35 TB/s (H100) to 4.8 TB/s (H200) to 8 TB/s (B200). The trajectory suggests 16+ TB/s within a few years.

About 16.7 minutes. A petabyte is 1,000 terabytes, so at 1 TB/s, the math is simple division. For context, the Library of Congress contains roughly 10–20 petabytes of data. Transferring it all at 1 TB/s would take about 3–6 hours.

Yes — petabytes per second (PB/s). Experimental optical interconnects and photonic computing architectures are pushing toward PB/s-class bandwidth. Some supercomputer storage systems already aggregate into the PB/s range when all nodes operate simultaneously. It is the next frontier for AI training clusters.

Almost exclusively in HPC (high-performance computing) documentation, supercomputer benchmarks, and IEC-compliant academic papers. If you are reading a spec sheet for a Top500 supercomputer's interconnect fabric, you might encounter Tibps. Consumer technology never reaches this scale or uses this unit.

Almost 10% — 1 Tibps equals 1.0995 Tbps, or about 99.5 Gbps more than 1 Tbps. At this scale, that 10% gap is roughly equal to a data center's entire edge bandwidth. Confusing the two in a procurement document could mean a six- or seven-figure cost difference.

Yes. A modern exascale supercomputer like Frontier has tens of thousands of GPUs that must exchange data constantly during parallel computations. The internal network fabric operates at aggregate bandwidths in the tens of Tibps to prevent communication bottlenecks from dominating computation time.

Neuroscientists estimate the human brain processes roughly 10-100 Tbps equivalent of internal signalling across ~86 billion neurons. In binary terms, that is roughly 9-91 Tibps — comparable to a mid-range supercomputer interconnect. The brain achieves this on about 20 watts of power.

Not for individual connections in the foreseeable future. A single human cannot consume Tibps of data — there is nothing to do with it. Even holographic video and full-sensory VR are estimated to need at most low Tbps. Tibps will remain the domain of infrastructure and computing systems, not end-user links.