Megabit per second to Terabyte per second

Because ISPs advertise in megabits (Mb) while download managers show megabytes (MB). There are 8 bits in a byte, so 100 Mbps ÷ 8 = 12.5 MB/s. Your connection is working perfectly — it is just a unit mismatch that has confused people for decades.

Netflix recommends 25 Mbps for 4K, YouTube suggests 20 Mbps, and Apple TV+ needs about 25 Mbps. In practice, 50 Mbps gives comfortable headroom for one 4K stream plus normal browsing. A household streaming on multiple devices simultaneously should aim for 100+ Mbps.

Wi-Fi shares bandwidth among all connected devices, loses throughput to interference from walls and other electronics, and uses half-duplex communication (it cannot send and receive simultaneously). A 300 Mbps Wi-Fi router might deliver 100–150 Mbps to a single device in practice, while Ethernet gives you the full rated speed.

Download Mbps measures data coming to you (streaming, browsing), while upload Mbps measures data you send (video calls, cloud backups). Most home connections are asymmetric — 100 Mbps down but only 10–20 Mbps up. Fiber-to-the-home plans increasingly offer symmetric speeds.

Surprisingly little — most online games use only 1–3 Mbps of bandwidth. What gamers actually need is low latency (ping), not high throughput. A 10 Mbps connection with 15ms ping will outperform a 500 Mbps connection with 100ms ping for gaming every time.

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.