Crossfire r9 290 - pcie 3.0x16 & 2.0x4 bottleneck with 4k res?

Toggle sidebar Toggle sidebar
C

Cheddle

Reputable
Sep 25, 2014
8
0
4,520
not looking for an assumptive responses - please reference benchmarks or personal experience

running a 4k (or triple 1440p) monitor configuration with crossfire r9 290's on a mobo (I have a z97m-g43) that runs a pcie 16x 3.0 for primary card and pcie 4x 2.0 for secondary card.

I am interested to see if there is a significant performance difference between 3.0 16x / 2.0 4x and a 3.0 8x/8x or 3.0 16x/16x using this configuration of 4k and r9 290 in crossfire.

There is negligible loss at 1080p and certainly a large increase in frame rates @ 4k - I will investigate this tonight via benchmarks with single card / crossfire @ 1080 & 4k.

For now id love to hear some experience on the matter.
 
Sort by date Sort by votes
C

Cheddle

Reputable
Sep 25, 2014
8
0
4,520


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

So far this is the best I can find online supporting that 16x/4x is a bottle neck with bridgeless crossfire

Im selling my Z97M-G43 (pcie 3.0x16 / pcie 2.0x4) and buying an ASUS Z87 Gryphon (pcie 3.0x8/ pcie 3.0x8)

I will report back with comparison before/after benches

so far the games not playing nice with crossfire on my current set up are:
Farcry 3 - more dropped frames and stutter than actual showed frames, reported FPS not bad - unplayable
Counter Strike GO - stuttering with crossfire
Grid Autosport - global illumination results in flicking headlights (strobing with v-sync / scanning with out v-sync)
planetary annihilation - stuttering with crossfire
skyrim - poor performance / stuttering - unplayable
watchdogs - large stutters - similar to farcry 3 - unplayable
Assassin's creed 4 - lighting/shading bugs with fixed horizontal lines of 20-30 pixels being darker on certain surfaces - average performance
lichdom: battlmage - poor performance / stuttering - unplayable

games playing nice:
tomb raider
crysis 3
crysis 2
metro last light (nvidia physx MUST be disabled (else 4fps lol))

benchmarks (synthetics) all so far work perfectly...

In general I find I have more poor performing titles than good ones... But when it works... 4k gaming with the settings cranked is like nothing else - as soon as a single card capable of 4k gaming @ reasonable frame rates is released ill go right back to a single card - unless more effort is invested in making crossfire/sli more reliable and polished.

hoping stuttering is caused by xDMA over chipset/southbridge (as opposed to cpu/pcie bus/northbridge) as such new mobo will resolve. I belive that its less a bandwidth issue and more a latency issue (it takes longer for the cards to communicate than it does to render a frame as such the data is lost??) as some engines do not have an issue yet others do. (although lichdom uses cryengine 3 yet crysis is perfect??)
 
Upvote 0 Downvote
P

philllip

Reputable
Sep 30, 2014
2
0
4,510
Really interested in this too. I've managed to get 2x 290's for just $500 and i'm keen to see whether my Sandybridge Asus Z68 (1x16x & 1x4x) MB will cause bottleneck issues with a Samsung 4k monitor.
 
Upvote 0 Downvote
P

philllip

Reputable
Sep 30, 2014
2
0
4,510
Found this: http://www.pugetsystems.com/labs/articles/Impact-of-PCI-E-Speed-on-Gaming-Performance-518/

Seems legit.

Our testing has pretty clearly shown that for gaming using either PCI-E 2.0 or PCI-E 3.0 will give you nearly identical performance. Oddly, in some benchmarks PCI-E 2.0 was actually faster than PCI-E 3.0. At the same time, x16 was not consistantly faster than x8. Again, x8 was actually faster than x16 in many cases. So unless you care about getting up to 1.5 FPS better performance, you might actually want to manually set your video cards to operate at x8 speeds - although we really would not recommend doing so.
 
Upvote 0 Downvote
C

Cheddle

Reputable
Sep 25, 2014
8
0
4,520
I bought a z87 based SLI certified board - pcie 3.0 x 8 on both slots.

the difference in performance is measurable:

tomb raider:
z87 = 66.7
z97 = 65.9

metro last light
z87 = 47.39
z97 = 42.86

unigine valley
z87 = 98.8
z97 = 94.8

firestrike gpu score (may be slightly different clocks on core in these (less than 3%) and z97 was with 14.7rc3 - z87 with 14.9)

z87 = 20,768
z97 = 18,881
 
Upvote 1 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom