shamino wrote:

Seriously? Well that's a surprise.
A few years ago I had an NF7-S rev2.0, I tried to quickly test it using a PCI video card that I had handy, probably my S3 Virge card. I don't remember anymore what the exact symptoms were, but I couldn't get it to work properly. I don't remember if it failed to produce a display, or if it just wouldn't function with accelerated drivers.
So after researching that on the web, I found some info claiming the nForce2 didn't support PCI video cards. I swapped to an AGP card and then it worked perfectly, and continued to do so ever since, so I didn't question it any further.

OS might've been a factor - I know my Symmetric card will not work properly in XP, it just produces a solid green screen instead. But for booting into the BIOS or if you have an older OS loaded it works correctly. No idea why someone would say nForce 2 can't work with PCI video though... 😕

But you were able to run a PCI card in your AN7, so apparently the PCI thing was a myth. I have an AN7 myself actually (don't have the NF7-S anymore). I'll have to try a PCI card in it sometime.

As far as the bridge situation goes, I guess the question becomes why the 3rd party PCI bridge chip manufacturer(s) didn't also come up with an AGP version of those chips. Was this the result of a significant complication compared to PCI, or just because for whatever reason, the card manufacturers didn't have nearly the same demand for AGP? I'm not aware of any reason why one would be more difficult than the other, so I suppose the answer is demand, but I'm surprised there wasn't enough demand for a G80+ AGP bridge to be worth somebody's while, even if not nVidia's. Anyway, apparently it wasn't, and if such a chip wasn't produced by now then clearly it's not going to happen.

I've always wondered if it wasn't power related to some extent - 6800 Ultra uses "only" 75W and is pretty high-end for AGP, whereas something like 8800GTX uses closer to 200W. 😊

That 6800 looks bizzare; I think "back alley" is fair. 🤣

shamino wrote:

As far as the bridge situation goes, I guess the question becomes why the 3rd party PCI bridge chip manufacturer(s) didn't also come up with an AGP version of those chips. Was this the result of a significant complication compared to PCI, or just because for whatever reason, the card manufacturers didn't have nearly the same demand for AGP? I'm not aware of any reason why one would be more difficult than the other, so I suppose the answer is demand, but I'm surprised there wasn't enough demand for a G80+ AGP bridge to be worth somebody's while, even if not nVidia's. Anyway, apparently it wasn't, and if such a chip wasn't produced by now then clearly it's not going to happen.

Probably this, too. Additional features like texture acceleration and sideband addressing would need to be supported by a true AGP bridge, which would just add to the already difficult task of converting PCIe signalling to PCI.

obobskivich wrote:

OS might've been a factor - I know my Symmetric card will not work properly in XP, it just produces a solid green screen instead. But for booting into the BIOS or if you have an older OS loaded it works correctly. No idea why someone would say nForce 2 can't work with PCI video though... 😕

At the time, I had used that Virge PCI card to test a few other motherboards, and I think it was with Win2k. Only the nForce2 board had problems with it, for whatever reason. It's been too long for me to remember the episode clearly though. I might revisit it next time I have the AN7 out. I have an 8400GS PCI I want to try with it, so maybe the Virge will get another shot while I'm at it.

Standard Def Steve wrote:

shamino wrote:

As far as the bridge situation goes, I guess the question becomes why the 3rd party PCI bridge chip manufacturer(s) didn't also come up with an AGP version of those chips. Was this the result of a significant complication compared to PCI, or just because for whatever reason, the card manufacturers didn't have nearly the same demand for AGP? I'm not aware of any reason why one would be more difficult than the other, so I suppose the answer is demand, but I'm surprised there wasn't enough demand for a G80+ AGP bridge to be worth somebody's while, even if not nVidia's. Anyway, apparently it wasn't, and if such a chip wasn't produced by now then clearly it's not going to happen.

Probably this, too. Additional features like texture acceleration and sideband addressing would need to be supported by a true AGP bridge, which would just add to the already difficult task of converting PCIe signalling to PCI.

Could be, but if nothing else, couldn't those extra features just be ignored? Not an optimal solution, but it should still perform better than a PCI slot anyway, and get the card on it's own dedicated bus. Didn't some of the early 3dfx AGP cards operate basically as 66MHz PCI cards, without really implementing AGP's additional features?

shamino wrote:

Didn't some of the early 3dfx AGP cards operate basically as 66MHz PCI cards, without really implementing AGP's additional features?

That's true for all 3dfx cards. Also true for some chips from other companies, definitely Rendition, probably S3 Virge GX/2, Trio3D - Trio3D/2X might have AGP texturing.