Talk:Gigabit Ethernet
From Wikipedia, the free encyclopedia
Archives | |||
|
|||
About archives |
Contents |
[edit] does anyone actually make gigabit hubs?
they are certainly allowed by the spec (there is a specific section on repeaters which talks about joinging 2 or more segments) but does anyone actually make them? Plugwash 12:05, 26 March 2007 (UTC)
- Are you sure they're allowed by the spec? I don't have a copy to check, but I was under the impression that 1000BaseT was the first 802.3 standard to /require/ full-duplex on all links. At least some of my NICs don't advertise 1000BaseT/half. As far as I've heard, any gear that supports this is extending the specification. I could very well be wrong, of course. Do you have a reference?--ktims (talk) 18:00, 27 April 2008 (UTC)
[edit] Forcing Gigabit
The article says that several implementations allow the user to force gigabit operation to eliminate autonegotiation issues. I belive this is misleading. Autonegotiation is *required* for gigabit transcievers because it sets parameters other than just speed and duplex. These implementations that purport to allow 'forcing' operation really only eliminate non-gigabit modes from the candidates offered by the negotiation mechanism. Autonegotiaion still happens, it just won't allow modes other than gigabit to be negotiated.
[edit] Line Codes
This article says 1000BASE-T expands 8 bits into four 3 bit symbols, while the 8B10B article explains how the same 1000BASE-T expands 8 bits into a single 10 bit symbol. How do they fit together? Also both articles should have links to each other. CannibalSmith 07:25, 1 June 2007 (UTC)
- Edit: 8B10B says Gigabit Ethernet, not 1000BASE-T specifically. So is 8B10B for optics only? CannibalSmith 07:29, 1 June 2007 (UTC)
[edit] What is actual data thoughput?
Does the Giga actually move exactly 1000Mb/s, and how much of that data is reserved for error correction and packet addressing (overhead)?
Word is some 20% is for overhead. --Flightsoffancy (talk) 15:50, 28 January 2008 (UTC)
-
- There is no error correction built into GigE itself. Other software or hardware layers would have to perform that function if it is desired.
- The packet addressing/overhead (by which I assume you mean things like MAC address, pkt type, CRC checking (which is error detection), is the same for all variants of Ethernet. See Ethernet#Physical_layer for details.
- There is 25% overhead for the line coding added to the packet addressing/overhead, as discussed in the article. Are you saying that you don't find that section of the article clear?
[edit] Data transfer limitations under PCI buses?
I just upgraded my network to 1000 Mbps Ethernet and I'm seeing transfer rates at around 30% of maximum. Is this because I am using PCI (not PCI Express) cards? --John Bahrain (talk) 18:40, 5 May 2008 (UTC)
No. Something is wrong there. —Preceding unsigned comment added by 81.216.200.10 (talk) 17:22, 19 May 2008 (UTC)