Bottleneck

[xnoobftw]

Would my motherboard be a bottleneck if i were to add another 290x to my rig? Rig in sig

@huilun02 said that it wont but i just wanted to clarify! 

[zappian]

I think we should replace bottleneck with another word.

 

Like vaseneck , im tired of seeing that word.

 

Anyways it wont vaseneck.

[harrynowl]

The 2nd lane is a 2.0 X4 to the chipset, so you have 1/8th the bandwidth of the first card effectively.

 

The performance impact I'm unsure of, if there is it'd be minimal at best. Results online show PCIe 2.0 x8 being 1-2 FPS slower than x16 in most tests, but that's very close to the margin of error.

[xnoobftw]

I think we should replace bottleneck with another word.

 

Like vaseneck , im tired of seeing that word.

 

Anyways it wont vaseneck.

 

The 2nd lane is a 2.0 X4 to the chipset, so you have 1/8th the bandwidth of the first card effectively.

 

The performance impact I'm unsure of, if there is it'd be minimal at best. Results online show PCIe 2.0 x8 being 1-2 FPS slower than x16 in most tests, but that's very close to the margin of error.

But i remember seeing somewhere that i would get like a 20% decrease in perf 

[zappian]

But i remember seeing somewhere that i would get like a 20% decrease in perf 

 

Its more like 5% max lol.

[harrynowl]

But i remember seeing somewhere that i would get like a 20% decrease in perf 

I cant find any material to either support or deny that number unfortunately, but I doubt it's that high. Maybe 5%?

 

Either way you'll see a benefit from upgrading to 290X X-Fire in graphically demanding applications.

[zappian]

I cant find any material to either support or deny that number unfortunately, but I doubt it's that high. Maybe 5%?

 

Either way you'll see a benefit from upgrading to 290X X-Fire in graphically demanding applications.

 

OFFTOPIC:LOL harry you changed your member title.

[harrynowl]

OFFTOPIC:LOL harry you changed your member title.

I did yes.

[xnoobftw]

Its more like 5% max lol.

 

I cant find any material to either support or deny that number unfortunately, but I doubt it's that high. Maybe 5%?

 

Either way you'll see a benefit from upgrading to 290X X-Fire in graphically demanding applications.

Hmmmm i saw that crossfiring causes tons of stuttering..... Should i even consider it? 

[harrynowl]

Hmmmm i saw that crossfiring causes tons of stuttering..... Should i even consider it? 

Where? In some games in times gone by there was frame pacing issues. This was like 18 months ago.

 

Are you not happy with your current gaming performance? Are you playing at 1440p/1600p/4K?

 

At 1080P a single 290-X ought to be enough for anything TBH.

[DemcoX99X]

Not sure if it's already mentioned: the card will only have half the speed (8x) so you would have to make your mind up if it's worth it :-)

[msvelev]

I think we should replace bottleneck with another word.

 

Like vaseneck , im tired of seeing that word.

 

Anyways it wont vaseneck.

Funniest shit I've seen all day

I'll use that word from now on!

[xnoobftw]

Not sure if it's already mentioned: the card will only have half the speed (8x) so you would have to make your mind up if it's worth it :-)

Hmmm but i heard that even a pci 2.0 4x is more than enough to saturate even a 290x 

[dragoon20005]

PCI-E 2.0 x4 will affect the performance of the 290X greatly since it depends on the PCI-E lanes to communicate with the CPU and other GPUs

[xnoobftw]

PCI-E 2.0 x4 will affect the performance of the 290X greatly since it depends on the PCI-E lanes to communicate with the CPU and other GPUs

^

 

I doubt so. Even if there was a bandwidth issue it would be minor.

 

It would be pointless to have a second card if just playing at 1080p, even at 144Hz.

 

Dual 290X is only really worth it when pushing very high resolutions.

 

OFFTOPIC:LOL harry you changed your member title.

 

The 2nd lane is a 2.0 X4 to the chipset, so you have 1/8th the bandwidth of the first card effectively.

 

The performance impact I'm unsure of, if there is it'd be minimal at best. Results online show PCIe 2.0 x8 being 1-2 FPS slower than x16 in most tests, but that's very close to the margin of error.

[Fate]

Don't worry about it @xnoobftw only a 4% decrease with a 7970 (less powerful than a 290x, but it should be pretty similar)

http://www.techpowerup.com/reviews/Intel/Ivy_Bridge_PCI-Express_Scaling/23.html

[zappian]

Don't worry about it @xnoobftw only a 4% decrease with a 7970 (less powerful than a 290x, but it should be pretty similar)

http://www.techpowerup.com/reviews/Intel/Ivy_Bridge_PCI-Express_Scaling/23.html

 

Like i said in this thread 5% max .

No Vaseneck.

[xnoobftw]

Don't worry about it @xnoobftw only a 4% decrease with a 7970 (less powerful than a 290x, but it should be pretty similar)

http://www.techpowerup.com/reviews/Intel/Ivy_Bridge_PCI-Express_Scaling/23.html

But the 290x doesnt use a crossfire bridge... it uses the PCI lane... Doesnt that limit it even more? 

[Fate]

But the 290x doesnt use a crossfire bridge... it uses the PCI lane... Doesnt that limit it even more? 

Literally the only thing I can find: 

 

Right, so I tried playing around as you've suggested. I saw the second card react as expected and the fans were spinning constantly indicting the ULPS was disabled.

 

This did not improve anything though. So I suspected the second PCI-E spot running at 4x was bottlenecking performance.

 

I took the plunge and upgraded my motherboard to a Gigabyte GA-Z97X-SLI . This motherboard supports both PCI-E slots in 8x in crossfire as well as a few other features I was after.

 

I jumped straight into BF4 and saw a massive performance increase. I'm pulling 80-120fps on ultra on triple monitors (5040x1050)

 

Also saw an improvment of around 1.5x in a quick 3DMark run of a single test. Haven't done too much serious testing.

 

So in conclusion, it was either the H87M chipset that couldn't handle crossfire very well, or the 4x secondary slot.

 

Thanks for your help guys.

http://forums.whirlpool.net.au/archive/2269391

 

No proof, just a statement by someone who had two 290's and an H87 motherboard. 

[xnoobftw]

Literally the only thing I can find: 

http://forums.whirlpool.net.au/archive/2269391

 

No proof, just a statement by someone who had two 290's and an H87 motherboard. 

I dont really wanna drop like 400SGD just for it to get bottlenecked.... Sorry for asking so much! 

[Fate]

I dont really wanna drop like 400SGD just for it to get bottlenecked.... Sorry for asking so much! 

Understandable! I wouldn't want to either lol

Maybe you could sell your 290x for a 295x2 o.0  Just food for thought

[xnoobftw]

Understandable! I wouldn't want to either lol

Maybe you could sell your 290x for a 295x2 o.0  Just food for thought

Hmmm the 295x2 doesnt use the pci lane for CF? 

[Fate]

Hmmm the 295x2 doesnt use the pci lane for CF? 

I don't know exactly how it works, but it's one card on a single PCIe 16x lane so it should be just fine. 

[Faa]

I don't know exactly how it works, but it's one card on a single PCIe 16x lane so it should be just fine. 

Both GPU's would run of x16 lanes, GPU-z will report it like that but in reality they share the bandwidth/lanes

[dragoon20005]

@xnoobftw

 

the R9 295x2 is internal lanes for sharing data between the 2 GPUs but the link to the RAM and CPU will be using the PCI-E x16 lanes

 

the only concern for using the 295x2 will only be low end CPU like the Pentium and i3

 

i5 4670/4690K is the minimum you can use with the 295x2

 

http://www.tomshardware.com/answers/id-2312541/crossfire-290-pcie-0x16-0x4-bottleneck-res.html

 

i found proof from Cheddle that dual 290/s on a non-SLi board will clip the performance of the 2nd 290