Welcome, Guest. Please login or register.

Author Topic: Lots of PCI solutions - why did no one create a PCI+AGP solution?  (Read 3033 times)

Description:

0 Members and 1 Guest are viewing this topic.

Offline csirac_

  • Full Member
  • ***
  • Join Date: Feb 2002
  • Posts: 154
    • Show all replies
1) Data rates. Peak/burst bandwidth of PCI was around 266MB/sec... it's supposed to sustain 133MB/sec (though many implimentations do less, around 100MB/sec). Whacking a PCI bus on a 10 year old PC only serves to allow upgrade options, not performance. You're not going to get 100MB/sec transfer speeds between the Amiga's CPU and a PCI device (though you may get this _between_ PCI devices on the PCI bus if the host controller/devices involved support it).

2) Practical issues. At the time Amiga PCI expansions were being released, you could mostly get PCI versions of most video cards.  (considering the age of our Amigas, you can still get pretty decent PCI cards even now - A TNT2 is still waaay faster than my CGFX64/3D). And since AGP 1x was basically PCI rewired a bit (some of the early AGP intel chipsets actually had a pin you could set that could make it's AGP bus be another PCI bus), it wouldn't have made much sense to have an AGP slot.

3) Speed, again. From (1) we're flat out justifying PCI, let alone AGP 2x or higher. From (2) we can see that initially, AGP was not actually a *faster* bus per se, it's just that it was a dedicated one. It allowed a dedicated data path to the CPU without competing for cycles with your NIC, RAID, sound, IDE, USB, etc. From a technical point of view, it simply doesn't make sense to defeat the whole purpose of AGP by putting it back on the PCI bus where we started :-)

So nowadays, the only point of putting an AGP bus on old A4000 would be to allow us to use new cheap video cards at 0.0001% of their capacity... and who knows, maybe someone will do that, it's just as crazy as putting PCI in ;-)

Cheers,

- Paul