How Much Better Is A 4870 In PCI-E 2 Over PCI-E 1.1?

[Nick1881]

Hmmm, upgrade time?

 

[Bizarre]

Running a gts g92 on a x48 ds4 netted me just over a frame extra in the crysis test compared to running the test on the same rig using a p5b deluxe.

I'm back using the p5b deluxe now as I prefer it over the ds4 for a number of reasons, and atleast going by my own testing, the x38/pci-e 2.0 chipset offered me little to no performance gains. Infact my current p5b + 4870 combo seem to keeps pace with the best of them on ocuk's crysis leaderboard.


P5B Deluxe - PCI-E 1.1 - 53.615 fps



X38 DS4 - PCI-E 2.0 - 54.775 fps

 

[Uriel]

P43, P45, X38 and X48

And on the AMD side: 770, 780, 790 series chipsets. Not sure about NVIDIA.

 

[fornowagain]

None looks like the PCiI-e lane splitter chip is still v1.1 at least on the hardware making the rounds atm

http://www.pcper.com/article.php?type=expert&aid=590&pid=2

PCI sig confirms that the spliter chip the "PEX8647" PCI-e 1.1

http://www.pcisig.com/developers/compliance_program/integrators_list/pcie/

which seems abit sad to see

Depends which review you read. > PEX86BB

(PCPer also got the framebuffer wrong, says 1Gb on their GPU-Z shot. Not even sure there is a 512Mb version, Anandtech have since updated their review to show 2x1Gb)



http://techreport.com/articles.x/15105
http://www.legitreviews.com/article/745/2/

In the center of the board sits a PLX PCIe switch chip, flanked by a pair of RV770 GPUs. We couldn't find this particular model of PCIe switch listed on PLX's website, but when installed, the card shows up as a PCI Express 2.0 x16 device. Each GPU shows up as being PCIe 2.0 x16 capable in AMD's driver control panel, as well. So our best guess is that we're looking at a PCI Express Gen2 switch that has 48 total lanes—16 routed to each GPU and 16 connected to the PCIe slot.

For ref PLX site > http://www.plxtech.com/products/expresslane/

 

[marc2003]

In my simple testing, I dropped almost 2500 3DMarks with the board in PCIe 1.1 mode and a single 4850 on my JW-X48D2-Extreme (12000 vs 14500).

06 at default settings? i can't believe a 4850 would be hampered (that much) at such a low res with no aa/af??

 

[willhub]

If thats true it means my score should be 16200 with an overclocked 4870 at 790/1100 and a E6600 at 3.4

 

[Overslop]

06 at default settings? i can't believe a 4850 would be hampered (that much) at such a low res with no aa/af??

Perhaps his mobo can't cope with lowered PCI-E speed?

 

[Deleted member 66701]

Okay, got the new Maximus so I've had a chance to compare a few games - 1st score is the Commando (PCI=e 1.0a), 2nd score is the Maximus (PCI-e 2.0):-

All done @ 1680*1050 with 4xAA, 16xAF, recorded through FRAPS, average three runs. Max settings in game. Bare Vista SP1 install, patched up to date, Cat 8.6 hotfix drivers.

Medieval 2:Total War, Agincourt scene = 67.8fps, 71.2fps
AoC, Pyramid 1 min run = 79.8fps, 84.6fps
Grid demo = 78.4fps 78.4, 82.9fps
World in Conflict = 39fps, 38.7fps

Sorry - no Crysis benchmark - FPS aren't my thing.

So, averaged out, tha v2.0 board is 4% quicker than the v1.0a board - although I suspect this could be down to better board design/better bios than actually an increase in PCI-e bandwidth.

 

[Greebo]

Okay, got the new Maximus so I've had a chance to compare a few games - 1st score is the Commando (PCI=e 1.0a), 2nd score is the Maximus (PCI-e 2.0):-

All done @ 1680*1050 with 4xAA, 16xAF, recorded through FRAPS, average three runs. Max settings in game. Bare Vista SP1 install, patched up to date, Cat 8.6 hotfix drivers.

Medieval 2:Total War, Agincourt scene = 67.8fps, 71.2fps
AoC, Pyramid 1 min run = 79.8fps, 84.6fps
Grid demo = 78.4fps 78.4, 82.9fps
World in Conflict = 39fps, 38.7fps

Sorry - no Crysis benchmark - FPS aren't my thing.

So, averaged out, tha v2.0 board is 4% quicker than the v1.0a board - although I suspect this could be down to better board design/better bios than actually an increase in PCI-e bandwidth.

Pretty good results. Good job there. To be honest it is what I would have expected at that res and a single card.

The crunch question though is, if a single 4870 is hit be around 5%, how much would a 4870x2 be hit by the pci-e 1 slot?

 

[DarkMatter]

yeah, i think that is the question on everyones mind. As my power supply is on the Very very very edge of the boarderline of powering a 4870x2 i am considering a 4850x2 instead.

 

[SS-89]

Pretty good results. Good job there. To be honest it is what I would have expected at that res and a single card.

The crunch question though is, if a single 4870 is hit be around 5%, how much would a 4870x2 be hit by the pci-e 1 slot?

yep but surely pcie 1.1 has more bandwidth than a 1.0a slot.

 

[Greebo]

yep but surely pcie 1.1 has more bandwidth than a 1.0a slot.

Nope. They are both 16x and same bandwidth.

Only pci-e 2 doubles the bandwidth.

 

[SS-89]

ah ok fair enough.

 

[Rangler]

Excellent read this, thanks to all involved.