PCI Express cards...

[darksun9210]

I found a single chip solution to bridge PCI to PCI-Express this morning. from the looks of things, it shouldn't be too difficult to build an adapter board to plug into a Grex/Mediator/Promethius PCI slot to get a PCI-Express 1x slot.

just wondering on the OpenPCI project, how easy it would be to write or get written drivers for such cards?

PCI to PCI-Express bridge

[motorollin]

Would the Amiga's PCI solutions be fast enough? Or would this solution combine two PCI slots in to one PCI-X slot?

--
moto

[Oli_hd]

Err you read that wrong, its a PCI-E to PCI chip, it cant work in reverse.

(Title of page: PEX 8111 PCI Express-to-PCI Bridge)
Its like the PCI to ISA chips that are/were used to add ISA slots to a PCI design.

That said there are  Xilinx designs for PCI-E controllers but I doubt we will see one in the Amiga any time soon.

[darksun9210]

it doesn't really matter how fast the PCI solution is, all the interconnects are there, everything is talking to each other via hardware (no software needed), so it is only as fast as the slowest componant in the chain.

Theoretically speaking, PCI is a shared bus with 133Mb/s bandwidth.
PCI Express 1x is a dedicated link of 250Mb/s up and 250Mb/s down.
It should be possible to combine multiple adapters in multiple PCI slots to create an electrically compatable 1x,2x, or 4x PCI Express slot, but as these are all hanging off of the same 133Mb/s PCI bus, there isn't any gain in performance, just a gain in expense.

N.B.
A PCI Express Graphics card with a 16x connector will run in 1x mode if that is all it has connected, and (in theory) will be quite able to eat up all the bandwidth a PCI bus can trow at it even in 1x mode...
to fit an 16x card in a 1x slot, just take off the end of the slot allowing the card to fit "through" slot itself.

[alexh]

You can still get PCI graphics cards which already have PCI-E to PCI adapter chips on them.

E.G. Radeon X1300

http://www.theinquirer.net/default.aspx?article=32741

http://www.xbitlabs.com/news/video/display/20060630145245.html

Getting the hardware has never been a problem, getting a decent software engineer who know's Amiga OS, Warp3D, OpenPCI etc. to port the drivers from Linux who can put the hours in is the difficult part.

[darksun9210]

true...
i've wondered about doing it myself, but unfortunatly, my programming knowledge goes as far as drawing graphics in pascal, unix C scripting, and a tiny bit of Arexx.
how many hours background i would have to put into learning before i could even attempt a device driver.
Depending if could i even get hold of the docs to tell me how to program the device in the first place?

knowing me i'd probably get bored and frustrated and make a little demo instead...

[Piru]

how many hours background i would have to put into learning before i could even attempt a device driver. Depending if could i even get hold of the docs to tell me how to program the device in the first place?

Considering I've closely observed the development of the 3d drivers for R100 and R200 series of ATI chips (http://3d.morphos-team.net/), the lacking quality of the ATI documentation and other considerations, I can say with some confidence that you could expect it to take years, not hours.

[darksun9210]

Piru
*laughs* i kinda figured as much  :lol:

anyway this PCI/PCIe adapter i was just thinking of as PCI seems to be disappearing, so it might extend the life of PCI systems...

in the meantime, i'm going to have a look at this openPCI malarky and see what thats all about...