Dave said:
MY prediction:
Video cards/GPUs will get fast enough to almost take advantage of the
extra bandwidth of PCI-Express X16, right around the same time that some
new (and larger bandwidth) bus technology is fielded. -Dave
Here is an article comparing X1950Pro AGP versus PCI Express. They are
still in the same ball park. The bus interface doesn't seem to be making
much of a difference. (There might have been another article like this,
but I could not find it.)
http://www.firingsquad.com/hardware/sapphire_x1950_pro_agp_ultimate_review/page5.asp
This article charts how much of an effect the PCI Express lane count, makes
on performance. (The links to the English version of the article appear to be
broken, so I had to use the German version.) Page 10, shows SpecViewPerf, which
is an X-Windows benchmark. The games benchmarks on page 11 are less affected
by the bus performance. By using cello tape on the pins of the PCI Express
video card, they can test x1, x2, x4, x8, and x16 bus performance.
http://hardware.thgweb.de/2004/11/05/pci_express_durchleuchtet_alle_modi_im_test/page10.html
http://hardware.thgweb.de/2004/11/05/pci_express_durchleuchtet_alle_modi_im_test/page11.html
*******
And finally, there is a newer standard coming out soon. PCI Express 2.0
allows a lane to go faster, but it is unclear what positive effect this
will have for video.
"PCI Express 2.0 nears completion"
http://news.com.com/PCI+Express+2.0+nears+completion/2100-1006_3-6123758.html
http://en.wikipedia.org/wiki/PCI_Express
"PCI Express 2.0
PCI-SIG announced the availability of the PCI Express Base 2.0 specification
on 15 January 2007.[2] PCIe 2.0 doubles the bus standard's bandwidth from
2.5Gbps to 5Gbps, meaning a x32 connector can transfer data at up to near 16GBps.
PCIe 2.0 is still compatible with PCIe 1.1, so older cards will still be able
to work in machines with this new version.
The PCI-SIG also said PCIe 2.0 also features improvements to the point-to-point
data transfer protocol and its software architecture.[3] Intel is expected to
release its first chipsets supporting PCIe 2.0 in the second quarter of 2007
with its 'Bearlake' family. AMD starts supporting PCIe 2.0 from its RD700
chipset series. NVIDIA has revealed that the MCP72 will be their first
PCIe 2.0 equipped chipset.[4]"
The problem with fattening up the PCI Express bus, is that the rest of the
busses needed, may not be ready for it. For example, on an Athlon64 system,
video card traffic travels over the HT bus, to get to the processor and its
memory controller. That bandwidth is just enough to match a single PCI Express
x16 rate interface. Increasing the effective lane clock rates would be a waste
of time, if the processor interface remains the same. (As it is, there were
motherboards in the past, with two x16 slots, but the slots could not run
full blast simultaneously.)
And on the Intel architecture, a question would be, what bandwidth can the
memory controller sustain, if there was a PCI Express 2.0 x16 video slot
DMA transferring to it.
It will be interesting to see what new power level the Northbridge chips
will have, with PCI Express 2.0 lanes on them. As it is, while processor
power consumption in idle is improving with time, modern Northbridge
chips only seem to be growing in power consumption. PCI Express 2.0 won't
help matters.
HTH,
Paul