PCIe Bandwidth Calculator

Calculate raw and effective PCIe bandwidth for Gen1 through Gen6 with configurable lanes, encoding, and protocol overhead.

Live

Input

PCIe Generation

Number of Lanes

Payload Size (bytes)

Additional Protocol Overhead: 5%

Accounts for ACK/NAK, flow control, and other link-level overhead. Typical: 3-8%.

Use binary units (GiB/s, MiB/s)

Example Configurations

Effective Bandwidth

6.75 GB/s

54.0 Gb/s aggregate

Raw Aggregate Bandwidth

7.88 GB/s

63.0 Gb/s aggregate

Parameter	Value
Raw data rate (per lane)	16 GT/s
Raw aggregate rate	64 GT/s
Per-lane bandwidth (raw)	1.97 GB/s
Per-lane bandwidth (effective)	1.69 GB/s
Encoding	128b/130b
Protocol efficiency (payload)	90.1%
Total efficiency	85.6%

Calculation:

Raw = 16 GT/s × 4 lanes × 98.5% encoding ÷ 8 = 7.88 GB/s

Effective = Raw × 90.1% (payload) × 95.0% (additional) = 6.75 GB/s

Example input:

Gen4 x4, 256-byte payload, 5% protocol overhead → ~7.87 GB/s effective

Results

Engineering Notes

Encoding Overhead by Generation

Gen1/Gen2 use 8b/10b encoding (20% overhead, 80% efficiency). Gen3/Gen4/Gen5 use 128b/130b encoding (~1.54% overhead, ~98.46% efficiency). Gen6 uses PAM4 with 1b/1b encoding (no encoding overhead, but FLIT mode introduces different protocol overhead).

Protocol Overhead Estimation

The protocol overhead percentage accounts for TLP headers (12-20 bytes for PCIe), DLLPs (6 bytes), framing symbols, and ACK/NAK traffic. Typical values range from 3-10% depending on payload size. Larger payloads have lower relative overhead.

Read vs Write Throughput

For reads, the completion overhead (data returning with CplD headers) reduces effective bandwidth compared to writes. Writes typically achieve higher effective throughput because the payload only travels in one direction.

Common Mistakes

Confusing GT/s with GB/s

GT/s is the raw transfer rate. To get GB/s, you must divide by 8 and account for encoding efficiency. A Gen4 lane at 16 GT/s with 128b/130b encoding gives ~1.97 GB/s per lane of usable data before protocol overhead.

Ignoring protocol overhead

Raw bandwidth is NOT what applications see. TLP headers, DLLPs, and framing reduce effective throughput. Use the protocol overhead field (typically 3-10%) for a realistic estimate.

This tool is intended for engineering analysis and educational purposes. Values and interpretations should be validated against the applicable JEDEC, PCI-SIG, NVM Express, MIPI, vendor datasheet or other authoritative specification.

Frequently Asked Questions

Related Tools

PCIe LTSSM Decoder

Paste LTSSM state logs to analyze transitions, detect repeated recovery events, and get engineering debug guidance.

DDR / LPDDR Bandwidth Calculator

Calculate theoretical bandwidth for DDR4, DDR5, LPDDR4, LPDDR5, and LPDDR5X with configurable data rate and bus width.

SERDES Calculator

Calculate symbol rate, unit interval, Nyquist frequency, and aggregate bandwidth for NRZ and PAM4 links.