Absolute Benchmark Results

Cool review.

"only the latest GeForce 9 graphics cards are PCIe 2.0 compliant."

thats incorrect. the G92 based 8 series cards are also PCIe 2.0

*sigh* looks like it's a new mobo for me. it was a good run GA-965P-DQ6!

What's up TOM's? PCIe 2.0 compliant? Only 9 series. Like mentioned... G92 is PCIe 2.0!

Just prooves two things - first that even PCIe 1.1 x8 is good for most cards & applications (same bandwith as x4 PCIe 2.0) as you lose 0-4% in all but heaviest apps like Flight Simulator and Crysis with AA/AF enabled; and second, that lower end graphics card can be used on PCIe 2.0 x4 slots without any problems, and for entry level you can make PCIe x1 cards just like the PCI ones that are sometimes used for smaller servers and such..

So I just wonder, will they ever start making x1 and x4 cards.. And I mean with _physical_ x4 or x1 slots, as you really can't plug x16 card in x1 slot unless you use a saw
It would also make multi-monitor setups much cheaper than using MBOs with 4x PCIe x16 slots and I bet x1 PCIe graphics cards would be much cheaper as they'd have less complicated and smaller PCB as well.

Nothing ground-breaking, but interesting read nonetheless...

Oops, my bad, they already exist :S Though I've never seen any in pricelists in my country
And these x1 cards that I've found (X1550 & 8400GS) aren't really cheap for 100+$, as x16 cards are less than half the price. Kinda kills the main reason of NOT SPENDING too much money.. as for extra 50-60-70$ you can get a better MBO in the first place

i think compare also be with hd 3870 x2

i think if thay use ati 3870 x2 it will be give more results than geforce 9800 x2

I have an MSI k9a2 platinum now I can either have 2 cards running at 16x each or 4 cards running at 8x
So what would be better 2 x ati 3870x2 or 4 x 3870
and would the 2x ati 3870x2 run at 8x or 16x
because if they run at 8x then i might be better off with 4 x 3870 runing at 8xor would the two 2 x 3870x2 run faster then 4x3870
at the moment i have 2 1900xtx

I think not.

It sounded like the article was recommending people move from their pci-e 1 motherboards to pci-e 2. The 1 speeds were 99% as fast most times. Flight simulator was the only game that saw a significant improvement going to new interface. I don't think that small difference in performance warrants the time and effort if you already have pci-e 1.0 or 1.1 motherboard.

The gains from PCIe 2.0 are minimal right now, but they will likely be more pronounced as newer cards are released, especially for multi-GPU configurations which seem to be all the rage. At this point I am finding it more and more difficult to recommend PCIe 1.x boards with the P4X chipsets on the horizon and the 750i already out.

Epic fail... not so much the article itself but some of the things in it. Actually I liked the overall article for the insight it provided because it reinforced what I've been telling people for months. I guess the most glaring of the problems is that I get a 404 error when I try to view the conclusion page. (err... I just went back to try to read something and am getting a 404 for the whole article). Not that I can't draw my own conclusions from what I read, I just like to see someone else say it... and of course I like being able to read the whole article.
That aside, someone goofed the batchwork on the graphs. The AMD graphs have Nvidia in the key but the Nvidia graphs are separate, and about half the graphs are "normal" .png files and the others are poorly compressed .jpg. Not that this is a big deal, but it only takes a minute to fix any of these quibbles and it takes away from the polish of the article... otherwise it seems to fit with the "Tom's is on an upswing" comment I think I read about yesterday's article about the VelociRaptor.

-mcg

OK... the 404 error disappeared. I read the conclusion page, and I don't like how they ignored the fact that lowered speeds sometimes outperformed higher speeds. In a few cases 8x and 4x slightly outperformed 16x. I wish they had explained that this was most likely due to "margin of error and signified no change" or something like that because I can completely see a noob quoting the article and asking if Crysis will perform better at 8x than 16x because several of the article benches showed this.

-mcg

Edit: HA!! I didn't realize that the "talkback" under the articles was actually posting on the forum or I would have just edited my first message.

I was hoping to see PCIe x16 2.0 (X38 chipset) versus PCIe x16 (P35).

excellent info. To go back even farther to a pci-e convertor sitting on top of an agp fitted card....What we see as written is just that in this article. Written.Just like CPU's and labels of "core duo". There are many items out there, and have been out there for years, with no name- doing what is now written. This is all for the sake of knowing it has a name. No panics for upgrades, if you do...there is not all that much surprise.

The last page says "9900 GX2"

So SLI PCIE 2.0 vs PCIE 1.1 that's something I would have like to have seen.