[Leaplist] so, dead nv part?

Kevin Korb kmk at sanitarium.net
Wed Jan 13 18:09:05 EST 2010


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Interesting.  I never even considered looking for newer video chipsets in
PCI.

IIRC the last PCI video card I bought was a GeForce MX 4000 which was a
favorite for NTSC video output.  At least until they dropped driver
support for it.

On Wed, 13 Jan 2010 15:02:58 -0800 (PST)
"Bryan J. Smith" <b.j.smith at ieee.org> wrote:
> Two (2) reasons for AGP mechanical issues ...
> 
> 1)  Intel released the _wrong_ mechanical specifications pre-release
> 
> Unlike PCI and PCIe, AGP was a "trade secret" by Intel.  Accordingly,
> only select vendors got the mechanical specifications.  Guess what?
> Intel changed them in the final spec, putting pin spacing off 25 mils
> from some specs -- a full pin -- for the placement of the slot to the
> end of the board.  So the "hard to get" specs were plagued by out-dated
> info, things that would be a no-brainer if they were part of an open spec
> available to all vendors.
> 
> 2)  AGP pin spacing is tiny
> 
> AGP is actually 32-bit PCI with all sorts of extra bus logic that fits
> in a much smaller space than PCI.  So it uses half the width (25 mils
> IIRC) of PCI, and it two-rows of pins.
> 
> [ Hence why PCI-Express uses Serialized I/O, to cut the pin-count down. ]
> 
> Additionally, AGP always had issues with the original "universal slot"
> design. It was designed to both 3.3V and 2.5V cards could be supported,
> including a possible, future drop (which would later be 1.8V).
> Unfortunately, if 1.8V card was not 2.5V compliant, the original
> "universal slot" definition would fry an 1.8V only card, and often
> destroy the slot/mainboard with it (did it myself ;).
> 
> Now as far as AGP v. PCIe video cards, it's all about the interface on
> the GPU.
> 
> AGP is 32-bit parallel PCI, with special "local bus" connections for
> special port and memory access (e.g., Direct in-Memory Execution --
> D[i]ME).
> 
> PCIe is serial PCI (up to 16 channels), and relies on different
> mechanisms for in-Memory Execution (far more reliable/compatible, with
> caveats).
> 
> They are completely different beasts.  A GPU must be either AGP or PCIe
> as they are radically different in many aspects.
> 
> Only one bridge type was ever created -- a PCIe x8 to AGP by Intel.
> It was used early on for bridging AGP cards to PCIe x8, and they were
> slower in PCIe because the AGP features were lost -- e.g., NV40.
> It can be used the other way as well, but then PCIe doesn't provide those
> AGP features either -- e.g., NV45.
> 
> However, both AGP and PCIe can be far more easily bridged to PCI.
> 
> AGP is directly just 32-bit @ 66MHz SDR/DDR/QDR (66-266MHz effective)
> and 133MHz DDR/QDR (266-533MHz effective), so slowing down to PCI
> 33MHz (or even PCI-X 66MHz SDR/DDR) is cake.  Other feature pins are
> lost.
> 
> PCIe is PCI as well, but serialized.  So it merely just needs to be
> converted from serial to parallel.  Intel sells a series of IXH chips
> that convert up to eight (8) PCIe channels into up to 64-bit PCI-X
> 1.0/2.0 at 66MHz SDR/DDR/QDR (66-266MHz effective).  So doing 32-bit PCI
> at 33MHz is nothing.
> 
> Hence why still see 32-bit PCI options with GeForce 8400 and even a 9500
> card, often universal 3.3V and 5V (because it's converted down much lower
> anyway).
> 
> 
> 
> ----- Original Message ----
> From: Kevin Korb <kmk at sanitarium.net>
> 
> Those are all good suggestions especially re-seating the AGP card.  Those
> things are way more picky than PCI cards in my experience.
> 
> I just wanted to point out that they still make AGP video cards.  I
> bought one not long ago.  They just don't make them in the latest
> greatest chipsets.
> 
> The last time I had a power failure one of my AGP video cards was dead
> when it came back :(  And yes, that system was on an APC SmartUPS.
> 


- -- 
~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~
	Kevin Korb			Phone:    (407) 252-6853
	Systems Administrator		Internet:
	FutureQuest, Inc.		Kevin at FutureQuest.net  (work)
	Orlando, Florida		kmk at sanitarium.net (personal)
	Web page:			http://www.sanitarium.net/
	PGP public key available on web site.
~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.11 (GNU/Linux)

iEYEARECAAYFAktOUpEACgkQVKC1jlbQAQe+AQCdE5TIenGL9J0ADB/X7Pq0ymtp
x+sAoOIWxO5a2BpQ1/7V3NvjehsS+EEl
=/v5z
-----END PGP SIGNATURE-----

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.



More information about the Leaplist mailing list