Efficiency terror : AMD Radeon R9 290X CrossFire performance leaks out

csbin

Senior member
Feb 4, 2013
895
577
136
http://videocardz.com/46990/amd-radeon-r9-290x-crossfire-performance-leaks


AMD’s new flagship Radeon R9 290X will no longer require CrossFire fingers to operate in multi-card configuration. Instead only PCI-Express slots will be required to make the cards run in such combination.
According to previous leaks this new approach would eliminate most of the frame pacing problems for which AMD was criticized since the topic became popular. However AMD went a step further and decided it would be a good idea to eliminate the necessity of the a separate CrossFire connection. It was possible since PCI-E interface is already capable of delivering enough bandwidth for multi-gpu communication.
The new slide supposedly leaked from AMD Reviewers Guide for R9 290X revealed that this card will scale much better than expected. The R9 290X CrossFire configuration will deliver up to 2.0x the performance of the single card. There is no game on the list that would not scale lower than 1.8x, but we are more than sure that these games were chosen here for a reason.
Additionally to that we have a chart with all the frame rates listed. I updated it with the CrossFire to single-card comparison.

5bKj9b4.jpg


From the same source we have a chart comparing Uber Mode to Quiet Mode. These results are very interesting, especially since there’s almost no difference between them. The average difference is just 1,54%. If that’s the case then I don’t really see the point of this mode. I guess everyone will just run it in quiet mode.

QVXfz8V.jpg
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
Gonna call shennigans on this. I remember when HD 7970 and there were similar claims of improved scaling, since I guess HD 69xx series had bad scaling or something.

And we all know where that scaling went. There are a handful of games where I experience the following:
1 card FPS meter ~40s, super duper stutter
2 cards FPS ~60, super stutter

Same game, on GTX 680
FPS meter ~30s, smooth as butter

Times like these I really want to drop CFX and I'm hoping 290X with OC will be enough to drive 1440p because one 7970 can't and two tends to be more bother than it's worth.
 

BrightCandle

Diamond Member
Mar 15, 2007
4,762
0
76
It looks like it's nothing more than anti-throttling mode

Of makes sense. One of the first debugging things people suggest is increasing power tune and then they start to get into XML file modifications for fixing the clock speeds. Dynamic clock speeds are a big problem for consistent performance and a lot of users have real world problems with it.

I have seen similar scaling graphs from AMD re the 7970 and that turn out horrible. Although this time with frame pacing at launch maybe it will all be OK. But I can shake the feeling that I want g-sync more than I want pretty much anything else and a new faster GPU is just no where near as exciting as having a zero lag, no tear and no stutter screen. The impact on visual quality is going to be enormous.
 

Carfax83

Diamond Member
Nov 1, 2010
6,841
1,536
136
Crossfire has always had excellent scaling, better than SLI truth be known. But good multi-GPU performance is more than simply good scaling.
 

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
5xxx series had bad scaling (iirc), 6xxx had better scaling but likely similar stutter issues as 7xxx was exposed to having.
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
5xxx series had bad scaling (iirc), 6xxx had better scaling but likely similar stutter issues as 7xxx was exposed to having.

I recall the scaling horrors following 6x since it was minor tweaks to 5x. I'd have to look it up, but I do specifically remember the 7x series having those "AMAZING CFX SCALING" charts and what not.
 

GaiaHunter

Diamond Member
Jul 13, 2008
3,656
234
106
I recall the scaling horrors following 6x since it was minor tweaks to 5x. I'd have to look it up, but I do specifically remember the 7x series having those "AMAZING CFX SCALING" charts and what not.

The 6850 is slower than a 5850.
The 6870 is slower than a 5870.

But if you look at the reviews at launch, the 6850CF and the 6870CF were beating/equaling the 5850CF and the 5870CF.

33214.png

33220.png


And even they lost, the CF lost for less than the single cards.

33229.png
 
Last edited:

Lonyo

Lifer
Aug 10, 2002
21,938
6
81
Remember that there is a different hardware setup for Crossfire, so there's no reason not to think that Crossfire scaling might also be different.

One would assume AMD wouldn't use a new method which uses additional chip hardware/etc unless there were benefits, such as improved scaling.
 

Teizo

Golden Member
Oct 28, 2010
1,271
31
91
Uber mode? LOL! The names these marketing people come up with to try to make things sound so special...

Hopefully the numbers are accurate though, and I imagine from here on out with new cards the CF/SLI bridge may become things of the past.
 

cmdrdredd

Lifer
Dec 12, 2001
27,052
357
126
Gonna call shennigans on this. I remember when HD 7970 and there were similar claims of improved scaling, since I guess HD 69xx series had bad scaling or something.

And we all know where that scaling went. There are a handful of games where I experience the following:
1 card FPS meter ~40s, super duper stutter
2 cards FPS ~60, super stutter

Same game, on GTX 680
FPS meter ~30s, smooth as butter

Times like these I really want to drop CFX and I'm hoping 290X with OC will be enough to drive 1440p because one 7970 can't and two tends to be more bother than it's worth.

Yeah I'd love to see a single card that can drive 1440p at the same framerate I enjoy with SLI. I'm not even 100% happy with overclocked 780 performance right now across all titles.
 

zlatan

Senior member
Mar 15, 2011
580
291
136
... what happened with the Ubermode lol.
I have heard that Ubermode is just for tuning. It's only work marginally different that quiet mode with stock settings, but with quiet mode the card don't support overclocking.
 

PCunicorn

Member
Oct 18, 2013
63
0
61
I think I call BS. No CFX bridge any you get at least a 80 percent improvement in the newest triple A games? Not gonna happen.
 

Lonyo

Lifer
Aug 10, 2002
21,938
6
81
I think I call BS. No CFX bridge any you get at least a 80 percent improvement in the newest triple A games? Not gonna happen.

There is a reason there is no CFX bridge... they changed the way the compositing works. Losing it won't reduce scaling performance just because they took it away...
 

Will Robinson

Golden Member
Dec 19, 2009
1,408
0
0
I'ts pretty obvious that moving the crossfire data over the PCI bus is not only good enough,its also allows better scaling than with the external bridge.
That's a step forward in dual card rendering architecture.
I'm pleased to see the reports of both the 280 and 290X feeling noticeably smoother than the last gen.
That's welcome news and can't wait to try one.
 

tential

Diamond Member
May 13, 2008
7,348
642
121
I see what they did there.

Lol good point out there. Gave me a good laugh.

I'll wait for actual reviews though. This looks too good to be true.
Resolution/Games probably chosen for a reason but well who knows.

Looks like AMD is firing on all cylinders but man could we use an ACTUAL REVIEW!!!!
 

imaheadcase

Diamond Member
May 9, 2005
3,850
7
76
They did seem to get all ducks in a row when dealing with 2 card installs. I suspect when nvidia comes out with new architecture they will go the same route.
 

VulgarDisplay

Diamond Member
Apr 3, 2009
6,188
2
76
Lol good point out there. Gave me a good laugh.

I'll wait for actual reviews though. This looks too good to be true.
Resolution/Games probably chosen for a reason but well who knows.

Looks like AMD is firing on all cylinders but man could we use an ACTUAL REVIEW!!!!

Yes, they chose all the standard benchmark games in use today for nefarious benchmark cheating purposes.