3 ms·
Are you sure? We aren't in the world of USB 1.1 latency any more. Things have improved vastly since then.
by kristopolous 7y ago
Are you sure? We aren't in the world of USB 1.1 latency any more. Things have improved vastly since then.
- kevin_thibedeau 7y agoThere's still latency getting data DMAed through buffers from Host to slave no matter how fast the link is. It's also non-deterministic.
- userbinator 7y agoI found two people who measured latencies, one a PCIe parallel port and the other USB 3.0: https://stackoverflow.com/questions/41987430/what-is-the-lowest-latency-communication-method-between-a-computer-and-a-microco https://stackoverflow.com/questions/41987430/what-is-the-low... https://stackoverflow.com/questions/13831008/what-is-the-minimum-latency-of-usb-3-0 https://stackoverflow.com/questions/13831008/what-is-the-min... PCIe parallel port: 4-8us USB 3.0: 30us I believe a regular PCI or even ISA parallel port can be below 1us. Those are "real" buses, unlike USB and PCIe which are more similar to packet-switched networks.
- FPGAhacker 7y agoI don’t know about ISA, but pci (not e) can easily be under 1us. It would be an odd design to have even that much latency. As measured on the bus.
- eqvinox 7y agoTheoretical one-way latency on PCIe 3.0 ×1 for a 256-byte packet is < 0.5µs. A single register write should be < 0.1µs.