GK110 versus R9 290X - Why Developers Matter

[the54thvoid]

Despite the release of AMD's new card and it's excellent showing against GTX Titan, it really doesn't matter these days who has the fastest card. The fact is, as these two very simple slides will show (4K testing), it depends on the game coding.

First UP, AMD biased game - Grid 2 (uses fx better suited to AMD)

What you get here is an 18% lead for the 290X over the Titan - that's huge.

Now what happens if you look at a classic Nvidia centric game - Batman (Arkham Origins).

What you get in this title is a 18% lead for the Titan over the 290X.


(http://www.kitguru.net/components/g...r9-290x-review-ultra-hd-4k-testing-part-2/19/) and BTW, before any people start hitting with crap, KitGuru recommend the 290X as their market leader recommendation. FTR it is a 290X review, so the OC section of it doesn't bother to OC the Titan.

The point is, in the olden days, the premium card was just that. The last card to have a pure performance lead was the GTX 580? The 680 and 7970 traded well, with 7970GHz, tending to win more often.

But with all the furious brand bashing of recent weeks it is worthy to point out that the games you play almost matter more than the card you own. I cherry picked two games from that review that i knew were Brand centric before anyone gets all touchy about their beloved team.

In W1zz's review the Titan and 290X are a few % apart across the board. But from the two extreme examples above it's pretty clear you can be deceived by the clamouring fanatical voices that one card is the be all and end all.

If anyone cares to reply to this thread, please keep it civil and you know, even talk more about game development and coding issues rather than just say "drivers suck this" or "overpriced that".

 

[Mathragh]

If anyone cares to reply to this thread, please keep it civil and you know, even talk more about game development and coding issues rather than just say "drivers suck this" or "overpriced that".

Don't think there's much to discuss about tbh, you've hit the nail exactly on the head, or whatever the English version of that saying is

The only thing left is something outside of our own power to influence: which road developers will take in the future regarding optimisations, perhaps mantle.

Thnx for this post!

 

[Sempron Guy]

That Batman AO result for the 290x probably needs more driver work. Grid 2 has been out for years already so that's most likely end of the line in terms of driver optimization for nvidia cards. Batman AO however is newly released. 2 frame difference between the 280x is questionable.

 

[Fourstaff]

That Batman AO result for the 290x probably needs more driver work. Grid 2 has been out for years already so that's most likely end of the line in terms of driver optimization for nvidia cards. Batman AO however is newly released. 2 frame difference between the 280x is questionable.

2 alternatives, both old games.

 

[TheHunter]

Check 1920x1080 or 2560x1440 tests to get the bigger picture.

4K UHD is a joke imo, no gpu is capable to run at decent fps, nvm more demanding games..

 

[heky]

Its not that the 290x is the be all end all, but for that much less cash compared to the titan it competely crushes it.

 

[erocker]

Just make sure to buy that 780ti when it comes out. That extra $150 bucks is totally going to be worth the extra few FPS!!!

I'm being sarcastic.

It would be nice if both companies kept their pricing relative to each other. The only reason I can see to buy a similar card with with minimal performance over the other comes down to marketing.

Of course, someone will have their reasons. I'll never get it.

 

[MxPhenom 216]

Its not that the 290x is the be all end all, but for that much less cash compared to the titan it competely crushes it.

+-5FPS isn't really crushing it. They trade blows.

 

[heky]

+-5FPS isn't really crushing it. They trade blows.

Read my post once again. If you compare the prices, the 290x crushes the titan.

 

[the54thvoid]

This thread is not meant for card bashing. I guess it's already run it's purpose or it can't be channelled in such a way, i.e. development of games using brand specific coding.

Can a mod just shut it before it becomes another tired fight between AMD and Nvidia loyalists?

 

[crazyeyesreaper]

Why are people still harping on Titan OMG ITS $1000 OMG 290X beats it

OMG every single GTX 780 available now thats non reference also beats Titan

OMG NVIDIA GTX 780 Windforce / DirectCu II / Twin Frozr / Lightning / Jetstream / Douchenozzel editions are way faster than Titan. Why would you buy a Titan ????? 1011010101101010 1337

their I just summed up all the ignorance in this thread as of this moment and in the future. It was all good until we hit post #6 lol

Titan is basically a geforce branded quadro it actually has more in common with the Quadro cards then the GTX 780 etc. It was a cash in card by Nvidia they even basically said so themselves with Titan outselling the GTX 690 in just the first few weeks of it being available.


the 290x has a competitor it is the GTX 780 not Titan. Yet everyone ignorantly clings to Titan because of the price avoiding the facts and the real picture.
780 Aftermarket is 4% faster than Titan at 2560 while 290x is 5% faster than Titan, huh.... what do you know Nvidia already crushed Titan themselves.......

Lets see where a 780 custom card comes in out of box

 

[Serpent of Darkness]

Re:

GRID 2 @ 2160.

All these figures are @ reference specs. This is looking at it from the averages. Figures down below are all theoretical with some degree of uncertainty...

RX9-290x Correlation: 0.000019 fps per (sp x core).
GTX Titan Correlation: 0.000018 fps per (CC x core).
GTX 780 Lightning Correlation: 0.000019 fps per (CC x core).
RX9-280 (Sapp Toxic Edition) Correlation: 0.000018 fps per (sp x core).

GTX Titan to GTX 780 Correlation Difference: 5.84% (GTX 780 > GTX Titan).
RX9-290x to RX9-280 Sapp Toxic Edition Correlation Difference: 1.42% (RX9-280 > RX9-290x).

For RX9-280 to reach 53 FPS, Core ~ 1350 Mhz.

GTX 780 Ti = 53.5 fps to 48.8 fps. + c; 10.25% difference.

RX9-290x to RX9-280 Toxic: 25.0% difference in performance.
RX9-290x to RX9-280 Toxic performance difference: 23.26% difference in performance.

GTX Titan to 780 Ti: 13.50% difference in performance.
GTX Titan to 780 Ti performance difference: 23.36% to 13.49%.

GTX Titan @ 1137 Mhz Core = 55.8 fps.
GTX Titan @ 1300 Mhz Core = 63.8 fps.
RX9-290x @ 1200 Mhz Core = 63.6 fps.
GTX 780 Ti @ 1200 Mhz Core = 69.12 fps to 63.1 fps; 9.54% difference.
RX9-280 Toxic Edition @ 1250 Mhz Core = 46.7 fps.


Batman AO

GTX Titan Correlation: 0.00002 fps to (CC x Core).
RX9-290x Correlation: 0.000014 fps to (sp x Core).
GTX 780 Lightning Correlation: 0.000019 fps to (CC x Core).
RX9-280 Sapp Toxic Edition Correlation: 0.000016 fps to (sp x Core).

GTX Titan to GTX 780 Lightning Correlation Difference: 7.82%
RX9-290x to RX9-280 Toxic Edition Correlation Difference: 18.59% (RX9-280 > RX9-290x).

RX9-280 Toxic Edition would need to be @ 1150 Mhz Core to reach 39 fps.


GTX 780 Ti = 54.48 fps to 50.53 fps + c; 7.82% difference.

GTX Titan to GTX 780 Ti performance difference: 13.50% to 5.28%.

RX9-290x @ 1225 Mhz Core = 47.77 fps ~ 48 FPS to match the GTX Titan.
RX9-290x @ 1396 Mhz Core = 54.5 fps to match the GTX 780 Ti possible fps performance.

Conclusion:

What does this prove. One, GTX Titan, GTX 780, and possibly the "full Titan" aka 780 Ti will have correlations to these two games, within a small percentage difference from one another. So with that information, a theoretical value or projection could be made about the performance of the GTX 780 Ti, with respect to Batman AO and GRID 2. Second thing, this may support the point that the OP made, and that is, the FPS performance for both brands could be dependent more on the coding of the game. You can see that the correlations for both RX9-290x and GTX Titan are off by a small percentage. It's only 3.06% on GRID 2. On the other hand, in Batman AO, the correlation changes, in favor of GTX Titan, at 47.19%. The performance of the RX9-290x is significantly less in Batman AO, and this is why RX9-290x performs worst than GTX Titan on Batman AO. MSI GTX 780 Lightning performs less in FPS because it's Cuda Core Count is less than the GTX Titan, in Batman AO, but can easily bridge the gap by increasing it's core frequency by a small percentage.

The correlations could be an indication of how efficient the cards are with respect to Batman AO or GRID 2. As you can see up above, the correlations for the RX9-280 Sapphire Toxic Edition is higher than RX9-290x, in both games. The significant difference is the core frequency difference and streaming processor count between the two. The only reason why the RX9-280 is doing less in performance is because it's Streaming Core Count is less than the RX9-290x while maintaining a frequency difference of 150 Mhz. So the gap can be bridge by simply increasing the core frequency by approximately 50 Mhz, in Batman AO, and 200 Mhz in GRID 2. The same goes for the MSI GTX 780 Lightning on Batman AO. An increase of 69 Mhz core frequency could bridge the gap between the GTX Titan and MSI GTX 780 Lightning. In GRID 2, even though the GTX Titan has more Cuda Cores, it has a lower Core Frequency at Boost. The MSI GTX 780 Lightning is beating it's performance by a small percentage. So, the GTX Titan would only require roughly a 62 Mhz increase to beat the MSI GTX 780 Lightning, and a 204 Mhz core frequency increase to beat the stock reference RX9-290x.

Side Note: Please take this information with a grain of salt. There will always be other variables in play like memory frequency, CPU Limitations, etc... This is only looking at the situation based on a view that is focusing on the GPU's Shader Count and Core Frequency.

 

[Crap Daddy]

This thread is not meant for card bashing. I guess it's already run it's purpose or it can't be channelled in such a way, i.e. development of games using brand specific coding.

Can a mod just shut it before it becomes another tired fight between AMD and Nvidia loyalists?

Don't close it just yet. I am surprised that this perspective on what these cards are capable of wasn't a topic earlier. It is obvious that since GCN arch and the Gaming Evolved offensive, AMD has done everything they can to get as many developers on their side as possible. Nvidia has been doing this thing successfully for years. Thing is, I don't remember such big discrepancies and if they were, they were repaired by drivers. What we have now is something much worse. We'll be soon be needing two machines one red and one green to run well all the games we want. I don't even want to think what happens if Mantle will succeed.

On the other hand Hawaii is the closest GPU in pure performance to NVidia's flagship in years. That's good for all of us.

 

[terroralpha]

i'm not an AMD or nvidia loyalist. i own an MSI 290x and an EVGA ACX GTX 780.

given the power consumption and heat output of the reference 290x, and it's now higher price tag (a difference of $80) i really don't see a reason to buy a 290x at this point. a non reference GTX 780 runs neck and neck with a reference 290x set to "uber." now factor in that nvidia gives you 3 free games that are worth no less than $60 on ebay together, and i think the GTX 780 is the bigger bang for the money.

i got the original release of the 290x and got a free copy of BF4, but they won't be doing that anymore. if you look on newegg you'll see that the BF4 editions of the 290x are disappearing and the a la carte 290x are now taking their place and labeled as "coming soon." that's another point for nvidia.

the way i see it is that the GTX 780 has a $140 price advantage over the reference 290x with lower heat output and power draw and just about the same performance. having said that, let's wait and see if the card manufacturers can work some magic with the 290x and get it to run without the need for water cooling.