but *at launch*, with a P2 233 or 266 (the fastest you could get), there was no difference.
That's because back then:
(1) CPUs were far too slow.
(2) Memory bandwidth was too low.
(3) T&L didn't exist.
(4) Traffic from the CPU to the GPU was extremely tiny
Just because there was no difference at all
at the time it does not mean that AGP is not faster than PCI
AGP was promised to be faster than PCI, but it took until the V3/TNT2 series to show any difference at all.
In otherwords AGP eventuallty delivered on its promise, just like AGP x8 will do in a few years.
I was referring to your statement about the Geforce 256 being materially faster in it's AGP form vs. PCI.
I still don't see any evidence to back your 1-2% comments.
It's not that large. Really, it isn't.
Of course it's not that large - old games on a slow CPU on a slow GPU.
But there is still a
significant performance difference, even in such unfavourable test conditions. What do you think would happen on a P4 3.06 GHz with a 9700 Pro? The difference would be massive, probably in the vicinity of 50%.
BTW, nice backpedal - first there was no difference and now it's "not that large".
And, like I mentioned before, AGP 1X was useless, other than the fact that it gave us a free PCI slot. Whee...
It's clear to me now that you're not interested in the truth at all and you're just trolling. Why don't you run some tests for yourself, maybe that will convince you that you're wrong.
Oh and for the article? The D3D benchmarks were useless, due to driver issues. Gawd knows what issues the OGL drivers had.
I could find plenty more articles to back my claims in addition to using data from my own tests but it's clear to me that you're not interested in the truth at all.