3870x2, the only reason to get spider?

[taltamir]

I noticed the tweaktown review tested the 3870x2
http://www.tweaktown.com/revie...aphics_card/index.html
It pretty much lags sligtly behind the GTS on most games.
This review however was done on a PCIe v1.1 motherboard.

I have seen some comparisons claiming massive speed improvements for that card in PCIe v2.
And the tests from websites who ran it on a PCIe V2 mobo show it annihilating the ultra. (with a few rare exceptions where the drivers don't work right and the card is running like a single card, in which case it falls behind the GTS)
http://en.expreview.com/?p=219#more-219
http://www.fpslabs.com/reviews...deon-hd-3870-x2-review
http://www.insidehw.com/Review...on-HD-3870X2-1-GB.html
In Chinese:
http://www.pczilla.net/post/89.html
http://www.pconline.com.cn/diy...iews/0801/1210234.html

So we actually have a card here, that is the fastest on the market, IF you have a PCIe v2 motherboard... which makes the 790FX chipset based motherboards highly attractive.
This is the only reason I know of to buy an AMD system at the moment. Sure your CPU is slower, but you can run the X2 card at the speeds it is meant to run, getting the fastest video card on the marked at the moment.


Call of Juarez on PCIe v2:
http://www.fpslabs.com/reviews...-3870-x2-review/page-6
Call of Juarez on PCIe v1.1:
http://www.tweaktown.com/revie..._dx10_vista/index.html

 

[krnmastersgt]

Only reason I would think that the 3870x2 card would be beating the Ultra when it comes to PCI-E 2.0 is because it's running 2 GPUs that actually exceed the PCI-E 1 bandwidth limit, didn't think I'd see that till the enthusiast 9 Series cards came out, but I guess you proved that theory wrong Still we don't know how the performance will be in a CrossFire or CrossFireX set-up do we? The Ultra might still beat out the 3870x2 card if they can't get the games to fully utilize all the GPUs

 

[Avalon]

The numbers of the other cards don't match up between benchmarks, so I don't think you can draw any conclusions from comparing the two.

 

[taltamir]

Numbers rarely match from different tests. Thats because a video card does not exist in a vaccum. There is a full system in which those video cards are tested. So different scores are to be expected (not to mention that different settings are used).

Unless they are lying about the results you should be safe to look at the RELATIVE performance.

The whole point anyways is that the numbers don't match, or rather the RATIOs don't match... PCIe v1.1 and v2 get completely different performance levels.

 

[lopri]

I do think that PCI-E 2.0 cards are somewhat crippled under PCI-E 1.0 x8 bandwidth, at least. In other words, I've tested HD 3850 on 975X - which can negotiate PCI-E bandwidth to one x16 or two x8's. When utilized under x8 bandwidth (which is PCI-E 1.0), the card loses its performance by 5~10%. So yeah now we need at least full x16 PCI-E 1.0 bandwidth. I'd like to see more thorough analysis on this topic. Like:

PCI-E 2.0 card on PCI-E 2.0 x16
PCI-E 2.0 card on PCI-E 2.0 x8 (this could be left out since most PCI-E 2.0 boards seem to have at least two full x16 slots)
PCI-E 2.0 card on PCI-E 1.0 x16
PCI-E 2.0 card on PCI-E 1.0 x8