Bandwidth is the budget with the most moving parts: every master has a demand, every link has a peak and a sustained rate, and the workload decides which link binds. NanoSoC's link table is fully worked — 384.0 MB/s peak per fabric link, 361.4 sustained with long bursts, 19.2 through the slow peripheral bridge — and this topic teaches how to turn workloads into demands, read that table, and name the bottleneck before simulation does. For the memory-system side see Memory Bandwidth and Latency; for the maths of outstanding transactions see Little's Law.

Suggested order: top to bottom. Price the demand first, then the supply, then compare them. For interface-level budgeting see Bandwidth & Latency Budgeting; for sustained-rate maths see Peak and Sustained Bandwidth.