I currently have an HD 7850 card (PCIe 3.0) in my Gigabyte board with a PCIe 1.0 slot, waiting on a board with a PCIe 2.0 slot. It's my understanding that in many cases the generation labeling is more marketing hype than reflection of actual better data flow.
With Medal Of Honor Warfighter and Battlefield 3 I find I can run on "high" settings but "Ultra" causes it to stutter.
Do you think this 1.0 slot is bottlenecking this particular card?
Is there a reliable way to test/compare the actual data flow?
Thanks
http://en.wikipedia.org/wiki/Pci_express
4 History and revisions [x16 full slot] Practical
rate
4.1 PCI Express 1.0a 250 MB/sec per lane 4GB/sec ~2GB/sec
4.2 PCI Express 1.1 250
4.3 PCI Express 2.0 500 8GB/sec ~4GB/sec
4.4 PCI Express 2.1 500
4.5 PCI Express 3.0 985 15.8GB/sec ~8GB/sec
4.6 PCI Express 4.0 ?
Anything over 1GB/sec is a good start.
Note that, transfer efficiency is also an issue, and
can cut the theoretical (physical layer) values in half.
So the 1.x slot runs ~2GB/sec on an Intel desktop. Still
twice as much as is needed. This is due to the usage of
small buffers to hold PCI Express packets when they arrive.
Chipsets can have other bottlenecks. And the history
of motherboards, is that they aren't always honest
about it, when there is one in the way.
You should also be checking the games, to see if
they are know to be "CPU limited" or "GPU limited".
The processor in your system, might actually be the
limiting factor. There are games in both those camps,
and the solution to performance problems will differ,
depending on which camp a game is in.
Some games are just poorly written, and you can pour
an infinite amount of $$$ at the problem, when really
the game engine should just be re-written.
Paul
