The first ATI r5xx/6xx X11 driver release [LWN.net]

The first ATI r5xx/6xx X11 driver release

Posted Dec 1, 2007 20:19 UTC (Sat) by jwb (guest, #15467)
In reply to: The first ATI r5xx/6xx X11 driver release by Ross
Parent article: The first ATI r5xx/6xx X11 driver release

The funny thing about HDMI is that they "add bandwidth" by pure committee fiat.  They didn't
actually change anything about the cable, the connector, or the signalling that would actually
result in more capacity.  They did it again with the most recent revision of HDMI 1.3, which
increases the symbol rate from 165MHz to 340MHz by crossing out "165" and writing "340" over
it.

With that kind of process we should have 100-gigabit ethernet in no time ;)

(Log in to post comments)

Higher performance through revision of standards documents

Posted Dec 3, 2007 2:28 UTC (Mon) by tialaramex (subscriber, #21167) [Link]

That's essentially the process by which we got GigE on copper.

What's the difference between a Cat5 cable (suitable for 100baseT) and a Cat5e cable (suitable
for 1000baseT) ?

The Cat5e cable says "Cat5e" on it, which means it was tested to the newer standard. Did
anyone actually make cables that wouldn't have tested to the newer standard? Theory says
"Maybe". Practical experience says "Nope". They're just UTP cables, made from ordinary copper
wire. It's just not that hard when you're making thousands of kilometers of the stuff.

Obviously a process like that can't continue forever, 100 Gigabit Ethernet over Cat5 cable
isn't likely to work. But Ethernet and HDMI aren't by any means the only examples. USB 2.0
relies on the engineering choices made back in USB 1.0 to assure customers that their cables
will Just Work at the higher speed, and sure enough they do.

Higher performance through revision of standards documents

Posted Dec 3, 2007 3:06 UTC (Mon) by jwb (guest, #15467) [Link]

But the gigabit ethernet committee started with the requirement of achieving that speed on
existing installed wiring, so they invented a wholly new signalling scheme to do so.  They
didn't just overclock fast ethernet and call it a day.  Gigabit ethernet barely even resembles
fast ethernet at the electrical level.

Just for starters, gigabit ethernet requires four pairs while fast ethernet requires two.
Gigabit uses five signal levels while fast ethernet uses three.  An so forth.