Visible to Intel only — GUID: ujn1559464046622
Ixiasoft
Visible to Intel only — GUID: ujn1559464046622
Ixiasoft
4.5. Link Rate and Bandwidth Efficiency Calculation
The F-Tile Serial Lite IV Intel® FPGA IP bandwidth efficiency calculation is as below:
Bandwidth efficiency = 64/66 * (burst_size - burst_size_ovhd)/burst_size * [align_marker_period / (align_marker_period + align_marker_width)] * [(srl4_align_period - 2) / srl4_align_period]
Variable | Description |
---|---|
raw_rate | This is the bit rate achieved by the serial interface. raw_rate = SERDES width * transceiver clock frequency Example: raw_rate = 64 * 402.812500 Gbps = 25.78 Gbps |
burst_size | Value of burst size. To calculate average bandwidth efficiency, use common burst size value. For maximum rate, use maximum burst size value. |
burst_size_ovhd | The burst size overhead value. In Full mode, the burst_size_ovhd value is referring to the START and END paired CWs. In Basic mode, there is no burst_size_ovhd because there is no START and END paired CWs. |
align_marker_period | The value of the period where an alignment marker is inserted. The value is 81920 clock cycle for compilation and 1280 for fast simulation. This value is obtained from the PCS hard logic. |
align_marker_width | The number of clock cycles where a valid alignment marker signal is held high. |
srl4_align_period | The number of clock cycles between two alignment markers. You can set this value using the Alignment Period parameter in the IP Parameter Editor. |
The link rate calculations are as below:
Effective rate = bandwidth efficiency * raw_rate
You can get the maximum user clock frequency with the following equation. The maximum user clock frequency calculation assumes continuous data streaming and no IDLE cycle occurs at the user logic. This rate is important when designing the user logic FIFO to avoid FIFO overflow.
Maximum user clock frequency = effective rate / 64