7970 CF benchmarks/rant

[jnmbhj]

I was looking online and I came across this benchmark from tomshardware. 

http://www.tomshardware.com/reviews/crossfire-sli-scaling-bottleneck,3471-4.html

 

If you go and scroll down to see the battlefield 3 ultra quality defaults the 1 7970 performs a little better than 2 7970's at 1920x1080 Surprisingly enough it does much better at 4800x900 than 1080p.

 

I kinda wonder wtf is going on with the amd 7970s. The 7970's beats the 680s at 4800x900 but the 680s completely destroys the 7970 at 1080p (which most people play)  Kinda strange if you ask me. 

 

If you look at the other tests. It appears that my 8350 will pretty much do better with 1 card rather than 2. Unless your gaming over 1080p 

 

One thing they didn't mention is that in crossfire the 8350 makes the fps dips almost unplayable at 1080p. 

 

So it doesn't really make sense to go with an 8350 and 280x CF or 7970 ghz CF. It just doesn't work.

 

I am also a little upset that I got tons of misinformation on this forum, I asked about crossfire in games at 1080p. 

[Katsu-K1L]

Reminds me of how the 780 ti beats the 290x at lower res and the 290x reks the 780 ti at 4k

[Kyuubixchidori]

I 

 

If you go and scroll down to see the battlefield 3 ultra quality defaults the 1 7970 performs a little better than 2 7970's at 1920x1080 Surprisingly enough it does much better at 4800x900 than 1080p.

 

what. where do you see one 7970 beats 2 7970s?

 

whoops sorry was reading graphs wrong. intresting how it works like that. I think its just a bf3 thing

[jnmbhj]

what. where do you see one 7970 beats 2 7970s?

 

whoops sorry was reading graphs wrong. intresting how it works like that. I think its just a bf3 thing

And BF4 and crysis 3

[dragoon20005]

the FX used to be really good

 

but with modern GPUs

 

the FX has a hard time feeding data to the GPUs

 

one is a issue but 2 makes it worse

 

 

your GPU is not the issue but the CPU

[SkilledRebuilds]

I was looking online and I came across this benchmark from tomshardware. 

http://www.tomshardware.com/reviews/crossfire-sli-scaling-bottleneck,3471-4.html

 

.

 

I am also a little upset that I got tons of misinformation on this forum

 

 

This article in question is using AMD Catalyst 12.10 - http://www.tomshardware.com/reviews/crossfire-sli-scaling-bottleneck,3471-4.html

Was written in 2013, before the frame pacing improvements done with 13.8 initially, then even further progress has been done in subsequent releases like 14.1 and so on..

13.8 - http://www.hardocp.com/article/2013/08/01/amd_catalyst_138_beta_frame_pacing_crossfire_driver/1

14.1 - http://www.anandtech.com/show/7729/amd-catalyst-141-beta-drivers-now-available-mantle-frame-pacing-more

 

The game would have changed somewhat performance wise for the better,..and while there can still be issues, that particular source article of information isn't relevant anymore to cite for comparison.

While easy enough of a mistake to make,and no doubt many of us are guilty of not looking closer,.... this is exactly why there is misinformation spread, the test setup of reviews and driver verisons/article dates should be noted before you post their relevancy.

[jnmbhj]

This article in question is using AMD Catalyst 12.10 - http://www.tomshardware.com/reviews/crossfire-sli-scaling-bottleneck,3471-4.html

Was written in 2013, before the frame pacing improvements done with 13.8 initially, then even further progress has been done in subsequent releases like 14.1 and so on..

13.8 - http://www.hardocp.com/article/2013/08/01/amd_catalyst_138_beta_frame_pacing_crossfire_driver/1

14.1 - http://www.anandtech.com/show/7729/amd-catalyst-141-beta-drivers-now-available-mantle-frame-pacing-more

 

The game would have changed somewhat and while there can still be issues, that particular source of information are not relevant anymore.

It is still pretty relevant considering I still get around the same ball park scores as it was done in the tests. Also the 7970 is not AMDs main focus and not much has been done to improve them. I have also played a little Battlefield 3 a few mins before I posted this to make sure it was still the case and surprisingly it was almost exactly the same. If you want I can redo the test again and give you a more exact number. As for the other scores I am not sure how they faired, but the main games that I tested was Battlefield 3 and 4. I understand that it was pretty old drivers but scores remained about the same in BF3 (72 fps in CF) (74 fps single card)

[SkulD]

I was looking online and I came across this benchmark from tomshardware. 

http://www.tomshardware.com/reviews/crossfire-sli-scaling-bottleneck,3471-4.html

 

If you go and scroll down to see the battlefield 3 ultra quality defaults the 1 7970 performs a little better than 2 7970's at 1920x1080 Surprisingly enough it does much better at 4800x900 than 1080p.

 

I kinda wonder wtf is going on with the amd 7970s. The 7970's beats the 680s at 4800x900 but the 680s completely destroys the 7970 at 1080p (which most people play)  Kinda strange if you ask me. 

 

If you look at the other tests. It appears that my 8350 will pretty much do better with 1 card rather than 2. Unless your gaming over 1080p 

 

One thing they didn't mention is that in crossfire the 8350 makes the fps dips almost unplayable at 1080p. 

 

So it doesn't really make sense to go with an 8350 and 280x CF or 7970 ghz CF. It just doesn't work.

 

I am also a little upset that I got tons of misinformation on this forum, I asked about crossfire in games at 1080p. 

You have to disable ULPS from registry that improves FPS on all scenarios.

 

Ihad that problem with my 7970 CF build on my signature is not updated.

[jnmbhj]

You have to disable ULPS from registry that improves FPS on all scenarios.

 

Ihad that problem with my 7970 CF build on my signature is not updated.

I disabled ULPS using MSI afterburner and I got no difference in fps 

[D2ultima]

Reminds me of how the 780 ti beats the 290x at lower res and the 290x reks the 780 ti at 4k

That's pretty much architecture. The 980 beats the Titan Black above 1080p, but gets beaten at 1080p. Kepler was built for 1080p and doesn't support 1440p+ very well is all; Maxwell was built for higher res (though not exactly sure about its actual 4K support... I think Maxwell's intended sweet spot is somewhere around 3K res or so. But that's my opinion).