[GPU.RU]Battlefield 4 Beta Bench

[Makaveli]

I own a 780 myself and I think you're vastly over-estimating the number of cards that do 1.3ghz. I can more or less assure you that 1.3ghz on a GTX 780 isn't common - but that is okay. Even a meager overclock of mid 1100s will generally have the 780 being in fighting range of the Titan, my own 780 is in the 1100 range.

That said, I agree with a lot of your points. Overclocking can certainly change things, but this does not change my opinion of the 290X price. 700$ is lunacy, period. 600$ is reasonable, with 550$ being a hell of a deal for that level of performance. I just don't think people will buy a 700$ card for the promise of overclocking - web reviews will hit mostly on stock performance and from everything we know, the 290X will be a bit faster than the 780 and trade blows with the Titan. To me that does not quality for a 700$+ price tag, even with overclocking - you could just as easily buy a factory OC'ed 780 to match the Titan's performance anyway for 660$ which would still be cheaper than a theoretical 700$ 290X.

I think RS means the Radeon hitting 1.3Ghz not the 780 if I read that correctly.

 

[RussianSensation]

I think RS means the Radeon hitting 1.3Ghz not the 780 if I read that correctly.

Exactly. My point is it takes about a 1.05-1.1Ghz 780 to match the Titan. If a reference R9 290X with a Turbo boost of 1Ghz is trading blows with the Titan, then the question is can R9 290X overclock to 1.3-1.35Ghz like some 780s can? If it can, then 780 will be behind once we move on to 780 oc vs. R9 290X oc scenarios. The main point is there is a hypothetical scenario where R9 290X will sell even if it costs $650 and that is if R9 290X oc beats GTX780 oc.

I also don't understand why AMD cannot sell high-end cards at prices similar to NV. ATI did it for years with 9800XT, X850XT Platinum Edition, X1950XTX. I do not like $600-1000 GPU prices, but if R9 290X ships with BF4 and it has performance of a Titan card/after-market 780s for $599-650, it is certainly a viable price level since it matches 780. I don't agree with this strategy however due to the technology curve (GPUs get faster/cheaper over time).

From AMD's point of view, they are selling a card with more VRAM, and BF4 is going to be more in demand than Batman. I think they may be baking in a price premium based on internal testing that Mantle will provide a substantial boost in BF4 and thus future titltes. Now if you launch R9 290X at $549 and Mantle brings 20% performance gain, then you will have vastly underpriced your card if your R9 290X is beating 780 by 25-30%. This could be 1 factor but it's impossible to say since we have seen no Mantle benchmarks.

Of course I still think a price of $650-700 is too much and would have much preferred $549 since 780 is now 6 months old.

--

blackened23, as far as 780s hitting 1.3Ghz+, HardOCP's MSI Lightning and Galaxy HOF both achieved that. I took both of those examples since someone spending $650 on a GPU is likely to look at those options. Grooveriding's Classies exceed 1300mhz with boost as well.

 

[Durvelle27]

https://www.youtube.com/watch?v=OIWNQIj4rRE

 

[Mastercommander]

Btw having recently run the game on both 660 Ti and 7950 I can attest that Vram usage @ 1080p no MSAA is about 1.5-1.7 gb on both

SLI is working, sort of, with almost no benefit, while Crossfire is not working at all. Well it is, perfectly, just with no fps boost

660 Ti SLI ran smoother but had occasional hiccup loading stutter turning corners which was absent with the 7950s - this may have been optimized out in patches; currently 660 Ti single card does not have such loading stutter at all

I just run single 7950 right now and use 64-bit SMAA injector for anti aliasing. I am interested in frames and would not be running high MSAA regardless

 

[hawtdawg]

So i upgraded to Windows 8, and now the game runs silky smooth with a 1.2ghz 780. I can put the scaling to 110%, turn on 4xMSAA, turn everything to ultra and run it at 1440p and maintain 60fps for the majority of the time. These benchmarks are way way way off. The numbers I get with 1 card are what this bench is showing with 2 cards.