Welcome, Guest. Please login or register.

Author Topic: 1200 networking speed report.  (Read 3995 times)

Description:

0 Members and 1 Guest are viewing this topic.

Offline Zac67

  • Hero Member
  • *****
  • Join Date: Nov 2004
  • Posts: 2890
    • Show all replies
Re: 1200 networking speed report.
« on: July 25, 2012, 06:23:35 PM »
USB adapters are very prone to stalling in certain conditions. When a network packet is received by the adapter it's buffered locally, then the host needs to poll the adapter to find out there's data to collect. The data is read over a - supposedly - slow bus (don't think clockport offers any significant throughput) which empties the buffer and enables the next packet to be buffered. A larger buffer (=better, more expensive adapter design) provides for receiving packets while transferring previous ones to the host, but there's only so much you can do.

Even with a much faster USB 2.0 host adapter on a modern machine it's hard to reach Fast Ethernet limits, let alone Gigabit - there's simply too much overhead and all the latencies add up. Plus, USB doesn't really help missing overlapped I/O, decent interrupts, etc.

I remember my experiments with a Fast Ethernet ISA(!) card many years ago: it simply had no point.

A native Zorro II card (or Z III - I'm dreaming) should be able to get somewhat higher speeds by lowering the overhead and increasing bus speed. However, to get near-maximum throughputs you need to give your NIC direct memory access on a low-latency no-hassle basis (like PCI or PCIe).