• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Multi Gpu Performance & Scaling | 1080P - 4K

LtMatt

LtMatt

LtMatt

Caporegime
Joined
12 Jul 2007
Posts
43,700
Location
United Kingdom
This review is a few months old now, but they test a lot of gpu setups so it's pretty interesting. Also of great interest (to me anyway) is how XDMA Crossfire/SLI scale from 2-4 gpu's at 1080P-4K. Really extensive review, loads of games and setups tested. Well worth a look.


4K - Gpu Scaling & Performance

QHAhITm.png


JYwaCLw.png


1440P - Gpu Scaling & Performance

o3ImC1t.png


vRdiU9N.png


1080P - Gpu Scaling & Performance

L6XpVQI.png


RDhiaxe.png

Source
http://udteam.tistory.com/585
 
Last edited:
XDMA works no better than SLI and with 4 GPUs can actually hinder performance/stability if the CPU is not up to the job. Even using a SB-E with quadfire XDMA is asking for trouble and only an IB-E or better should be considered.
 
Kaapstad said:
XDMA works no better than SLI and with 4 GPUs can actually hinder performance/stability if the CPU is not up to the job. Even using a SB-E with quadfire XDMA is asking for trouble and only an IB-E or better should be considered.
Click to expand...

You'll be pleased to know they used Ivy-E and it's clocked to 4.5Ghz. 290Px4 does pretty well too which is good to see.
 
Also looking at the graphs the ones @1080p are wrong. According to them 3 x 290Xs performance better than 4 x 290Ps but scale worse which is impossible.
 
LtMatt said:
You'll be pleased to know they used Ivy-E and it's clocked to 4.5Ghz. 290Px4 does pretty well too which is good to see.
Click to expand...

See my post above.:D

The graphs are meaningless.

It would have been more useful to give the numbers rather then colourful graphs.
 
Kaapstad said:
See my post above.:D

The graphs are meaningless.

It would have been more useful to give the numbers rather then colourful graphs.
Click to expand...

Take a second look, it lists performance scalability numbers after each game tested. Sometimes negative scaling/tri-quad gpu issues becomes apparent under certain titles it seems. Probably explains a few weird results at 1080.

gsmi9F5.png
 
LtMatt said:
Take a second look, it lists performance scalability numbers after each game tested. Sometimes negative scaling/tri-quad gpu issues becomes apparent under certain titles it seems. Probably explains a few weird results at 1080.

gsmi9F5.png
Click to expand...

3 similar cards should never perform better and scale worse than 4 cards that is basic maths.:p

I also don't like the review because they are not pushing the hardware that much with the settings. People need to know that the 780ti's will fail on a lot of those games maxed and the 290s will fail on a few, not telling people this is less than helpful when looking to go 4K.

Put 4 Titan's in there turn up the settings to max and they will beat the faster 290Xs and 780ti's with ease @4K.
 
Kaapstad said:
3 similar cards should never perform better and scale worse than 4 cards that is basic maths.:p
Click to expand...

It comes down to how well the 3/4 cards work in some titles compared to others though I'd have thought. Looking at the results when 4 cards win, it wins big. In other instances 3x290X's are faster than 4x290's at 1080P indicating negative scaling or something else. Looking through the results it looks to me like at 1080P with 4 cards there are just a whole lot of worms with regards to performance and scaling, depending on the title.
 
780 SLI - 78FPS in BF4 at ultra + 4xmsaa over 1080p? Really? Its a rare day when it drops below 100fps, let alone averages out to sub 80fps! No CPU excuses either, he's using a 4930k at 4.5ghz. I think he even tries to blame memory bandwidth? Completely fudged results on that. He also says something about it not supporting Tri-Quad-SLI, it does.

Crysis 2, looks to have a frame limit cap, maybe he should have removed this before checking for scaling, it makes the results almost pointless.

Besides those two every thing else looks pretty much spot on results wise, I always saw better scaling on my 7970's then I do my 780's (at 1080 anyhow)
 
8 Pack said:
Xfire is much more efficient than SLI for sure. I have tested this many times.
Click to expand...

In games it is pretty even, if anything I would say that SLI have got the edge with 4 cards.

Try loading up a copy of Tomb Raider or Sleeping Dogs and bench it on 4 290Xs, it is not a pretty sight. Tomb Raider has got micro stutter quite bad and Sleeping Dogs has got Macro stutter. Quad SLI on the other hand is a lot smoother.
 
LtMatt said:
That's what all the benchmark sites usually show too, especially at higher res. :cool:
Click to expand...

Actually it is the other way round on the synthetic benches where AMD do very well with multi GPU scaling, this advantage decreases as the resolution increases.

The reason the 290s do well in games benches at high resolutions is down to their 512bit bus not the GPU scaling, unfortunately when max settings are used the 6gb Titan's come out on top.
 
You must log in or register to reply here.
Back
Top Bottom