Date: Sun, 15 Sep 1996 11:46:35 -0700 (PDT) From: "Rodney W. Grimes" <rgrimes@GndRsh.aac.dev.com> To: grog@lemis.de (Greg Lehey) Cc: freebsd-hardware@FreeBSD.org Subject: Re: Any Pentium boards with more than 4 PCI slots? Message-ID: <199609151846.LAA04679@GndRsh.aac.dev.com> In-Reply-To: <199609151230.OAA00535@allegro.lemis.de> from Greg Lehey at "Sep 15, 96 02:30:14 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
> Does anybody know of Pentium boards which support more than 4 PCI > slots? No, but if you find anything please do let us all know about it! > One of my customers has a requirement for up to 8 or 10 > slots. That would either require 2 host to PCI bridge chips, and I don't know of any Pentium Host to PCI bridge chips that allow more than 1, though in theory you could probably do it with some glue logic and the 82439HX chip since it at least understands 3 devices being on the host bus. Or 2 PCI-PCI bridges giving a primary PCI bus with 2 slots open and 2 secondary buses with 4 slots each. > My understanding is that this is only possible with a PCI > bridge--is this correct? No, that is now the only way it could be done, see above. But this is the most economical way it could be done. > If so (and even if not, I suppose), how transparent is the PCI bridge > to the software? Once configured correctly a PCI-PCI bridge is totally transparent to both the software and the hardware as far as functionality goes. > How much slower is it than the direct side of the PCI bus? Thats a ``Complicated Question'', both sides of the bridge still run at full speed, but when a transaction must cross the bridge the chip must aquire the other bus (and the other bus may be busy due to having multiple masters on it), after the aquire the bridge looks like a 1 clock delay. But then, some of the bridges (DEC DC21050, go dig around on http://www.dec.com for full data sheets, app notes, and even schematics and order numbers for a evaluation board that would add 4 PCI slots to any standard motherboard) have posting buffers that would eliminate or atleast make transparent some of this loss in performance. > Are there any other gotchas? Yea, most chipsets don't have a way to route more than 4 interrupts to PCI slots, thus your going to be forced to share interrupts unless you go with a custom design. How ``important'' is the need for this product, ie, would it be worth the ~$250,000 NRE charges to have a custom solution designed? Remeber with enough money you can have almost anything designed and built that is technically possible to due. -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Reliable computers for FreeBSD
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199609151846.LAA04679>