Question about 290x crossfire on PCI 2.0 x16 x16 at 4k resolution

[cthomas1489]

So I have an all AMD system with the FX8350 and Asus Crosshair V Formula-Z. This board can run two cards in x16 x16 but all the slots are only PCI 2.0. I've seen many posts about how it is a very marginal loss by not using a PCI 3.0 slot but I think most of these were at 1080p resolution. I did see a graph of a triple 1080p monitor set up that seemed to show there were some significant losses on the PCI 2.0 setup in BF3.

 

Anyway, I plan to play at 4k resolution and was wondering if the losses would be major or minor?

[SkilledRebuilds]

Dunno about triple screen, but consensus I've managed is that the PCIE 2.0 16x will give you 95-100% of the power available, and 2.0 8x is the one where your at 90-95%.

 

Both cards in 2x 2.0 16x slots will be fine and sync accordingly.

Microstutter may be induced a little by using 16x/4x (which for you, isn't a thing) but 16/16 or 8/8 should be very very good.

 

Triple screen gaming should not be that much different... until we "need" PCIE 3.0" I guess.

[cthomas1489]

Okay makes sense. Yeah I wouldn't be triple screening but rather just a single 4k monitor.

[G9x001]

I don't believe any of the new single GPU cards would have a bandwidth that would exceed 4GB/s so you should be able to see no difference down to a PCI-E 3.0 x4 lanes and PCI-E 2.0 x8 lanes.

check this out.

www.pugetsystems.com/labs/articles/Impact-of-PCI-E-Speed-on-Gaming-Performance-518/

 

[cthomas1489]

Thanks for that link! I've been reading things that go back and forth non stop but I like that article above.

[VintageCake]

the 8350 bottlenecks an sli or crossfire system by quite a bit at stock, unfortunately.

 

overclock it

[cthomas1489]

How so?

[G9x001]

are your R9 290x air cooled and are they reference cards and at stock clocks? 

they could be hitting the power or thermal limits and throttling. I had an arse of a time when I first got my R9 290x. 

if you don't know how to check if it is throttling have a look at this video. I made it a while ago.

sorry about the picture quality in advance.

http://youtu.be/csegc6OUazc

I hope it helps.

[G9x001]

the 8350 bottlenecks an sli or crossfire system by quite a bit at stock, unfortunately.

 

overclock it

Facts are always good when making posts.

you would have to go all the way back too a LGA 775 socket motherboard with a Intel® Pentium® 4 Processor 524 CPU or a AMD AM2+ socket with AMD Phenom™ II X6 processors to see a bottleneck with two R9 290x's in crossfire. even then it would only be a small loss of performance if any.

 

AMD FX 8350 would handle quad cross fire with no bottle neck. the only way you could do this would be with two R9 295x2's or two HD7990's or Two HD6990's and so on.

[G9x001]

Here's  a little more info for you on the performance requirement for the r9 290x and the minimum pci-e lanes required for maximum performance.

the minimum pci-e lanes required for the card to operate at full potential is PCI Express® based PC is required with one X16 lane graphics slot available on the motherboard. So that is one pci-e 1.0 x16 slot.

each revision of pci-e you get around twice the transfer rate per lane.

 

Per lane (each direction):

• v1.x: 250 MB/s (2.5 GT/s)
• v2.x: 500 MB/s (5 GT/s)
• v3.0: 985 MB/s (8 GT/s)
• v4.0: 1969 MB/s (16 GT/s

so a R9 290x will run at full performance with no bottle neck with pci-e 2.0 x8 or pci-e 3.0 x4

Sorry about posting so much I Just want to clear the BS from the facts.

www.sapphiretech.com/presentation/product/?cid=1&gid=3&sgid=1227&pid=2085&psn=&lid=1&leg=0#

www.hisdigital.com/un/product2-777.shtml Under system requirements on both links.

[VintageCake]

Facts are always good when making posts.

you would have to go all the way back too a LGA 775 socket motherboard with a Intel® Pentium® 4 Processor 524 CPU or a AMD AM2+ socket with AMD Phenom™ II X6 processors to see a bottleneck with two R9 290x's in crossfire. even then it would only be a small loss of performance if any.

 

AMD FX 8350 would handle quad cross fire with no bottle neck. the only way you could do this would be with two R9 295x2's or two HD7990's or Two HD6990's and so on.

that's nice, have you ever actually tried games with the 8350 stock in crossfire? I am speaking out of my own experience.

It slightly bottlenecks in games, that's how it is. A nice overclock will make this go away.

[cthomas1489]

Thank you very much for the replies guys. Everything I've been researching leads me to believe I won't be hindered. I haven't noticed anything yet at least when playing.

They are XFX DD cards not reference. Everything is slightly OC'd at the moment and while they get real hot sometimes I don't notice any throttling.

[G9x001]

that's nice, have you ever actually tried games with the 8350 stock in crossfire? I am speaking out of my own experience.

It slightly bottlenecks in games, that's how it is. A nice overclock will make this go away.

I think what you are seeing is not an in crease in performance for the GPU but an increase in the CPU compute performance. A lot of games (games with a lot of things moving falling in the back ground like BF2-3-4) rely on the CPU to carry out the physics calculations so a higher CPU clock speed will increase the Physics calculations giving you a higher FPS in games. Has nothing to do with the GPU being bottlenecked.

[VintageCake]

I think what you are seeing is not an in crease in performance for the GPU but an increase in the CPU compute performance. A lot of games (games with a lot of thing moving falling in the back ground like BF2-3-4) rely on the CPU to carry out the physics calculations so a higher CPU clock speed will increase the Physics calculations giving you a higher FPS in games. Has nothing to do with the GPU being bottlenecked.

but it nerfs the entire crossfire setup by quite a lot.

 

If you see the cpu running at 100% in games, it's a bottleneck. This is what happened in BF4 with two r9 290s (OC) in crossfire.

dude, its a bottleneck.

[cthomas1489]

VintageCake, can you provide some screenshots of the bottleneck? That'd be great. I'm just interested in learning but I honestly have more proof of the other case vs yours...

[cthomas1489]

Don't mean to call you out by any means, I'm just curious what you've seen first hand. BF4 puts a few cores up around 75-85% but I've never maxed all my cores at 100% but maybe I haven't play a demanding enough of game?