770 vs 7970 vs 7950

a 7950 or 760 are both all the gpu you'll ever need for 1080p

No none of them can max Crysis 3, which takes as much power as you can throw at it.

With Metro LL, you will want to turn on the PhysX effects, so its best to go with either the GTX 770, 670, or 760 in that order of descending price/performance.

The GTX 770 costs $400.

http://www.techpowerup.com/reviews/MSI/GeForce_GTX_760_TF_Gaming/14.html


http://www.techpowerup.com/reviews/MSI/GeForce_GTX_760_TF_Gaming/19.html

Here's HARDOCP's review of the MSI GTX 760 overclocked.

On their test bench comparing the AMD 7950 and GTX 760 @1080p for Far Cry 3 and Metro:



We pushed Far Cry 3 up to 4X MSAA to stress all the GPUs. You can see that as we move the MSAA level upwards the Radeon HD 7950 starts to take bigger leads over the MSI N760 OC in raw performance. Yet, the performance is still so close that in-game we can't tell a difference.



In the above comparison we are comparing Metro: Last Light at 1920x1080 with "Very High" settings and "Very High" tessellation enabled. This is the highest in-game settings. Once again the AMD Radeon HD 7950 has a slight advantage in performance. The MSI N760 OC overclocked is a lot faster than the PowerColor HD 7870 MYST though.

Messed up on Crysis 3 in the above post. Here's Crysis 3 -



In this graph we are comparing Crysis 3 at 4X SMAA at "High" settings now. This increased SMAA setting puts a big burden on shader performance required for this game. The Radeon HD 7950 overclocked performance starts to take a bigger lead over the MSI N760 OC at this higher, more demanding shader heavy setting. The difference is only 6%, and the game felt the same between these, but it is larger than it was at 2TX SMAA. There is no doubt the MSI N760 OC overclocked is a lot faster than the PowerColor HD 7870 MYST overclocked.

first of all no one plays with x4aa, simply turn that off and you're set, they turn that crap on just to bench cards because otherwise the cards would all look pretty much identical... no one plays with AAx4 on, no one. second even if you wanted to play at aax4, 40fps is good enough as long as the frames are well paced

You can afford to turn down graphic eyecandy a tiny bit to max out your system to 60fps in crysis 3. everyone does it. I still stand by the assertion the 7950/760 is all the gpu you'll ever need at 1080p.

as a sidenote, i will die a happy man if i can enter a convo about gpus and not get spammed by a techpowerup graph ever again. Their gpu ranking graphs are so far outside of every other benching site's results i can't believe people still quote them for gpu rankings as their results are clearly outliers for reasons i'm not quite sure. and that they seem to be quoted in every single gpu thread blows my mind.

the 770 is a nice gpu but it's about equal with a 7970GE (probably a little better, depending on overclocks involved), which is a card the 7950 well overclocked can come within an eyelash of matching... there is something massively wrong with techpowerups gpu graphs and i suspect the problem lay with driver issues. They do the same thing andtech does, which is mix driver versions in their graphs. for example, an older card will get tested on a game with the most recent driver when the game comes out, they save the result, then 4 months later they test a new card with even MORE recent and optimized drivers on that same game, then the new card benches significantly higher then the older cards because of a driver issue at the time the older cards were benched.

Those bench results from the techpowerup graph with the 760/770 are with drivers perfected for crysis over a number of months, while those of the older cards were of those cards when the game came out. as i said... while both are very nice gpus the 760/770 aren't significantly or even noticeably better then a 7950/7970GE respectively.

To answer the Original Question, here's some analysis on the 770 VS the 7970 Ghtz edition:

Metro Last night 2560x1600:



In this apples-to-apples graph we are comparing at 2560x1600 with "very high" settings and "very high" tessellation. The MSI GTX 770 Lightning is 13% faster than the GeForce GTX 680 and 16% faster than the Radeon HD 7970 GE.

Crysis 3 @1080p -



This graph shows that all three video cards are playable at 1080p with "very high" settings. The MSI GTX 770 Lightning is 15% faster than the GeForce GTX 680.





The 770 looks like the best performing card on HardOCP's bench.

btw: the msi lightning is about the fastest and most expensive 770 you can get your hands on. Lets not use a $500 gpu to prove some stupid benching "whos manhood is bigger" contest. i somehow doubt someone who's looking at a 7950 as an alternative is looking to spend 500 on thier gpu.

Lets keep this a little sane and give him some affordable alternatives.

TechPowerUp, the same people who created GPU-Z, use only the latest and the same drivers consistently for all of their charts. That can't be said for most other review sites. TPU is as trustworthy as they get.

Quote:
"•All video card results were obtained on this exact system with exactly the same configuration."
http://www.techpowerup.com/reviews/MSI/GeForce_GTX_760_TF_Gaming/6.html

My bad! I didn't check pricing before posting.

I am in the Camp with you that the 760 or the 7950 is the best option at 1080p; that's why I posted the specs on the MSI 760 for HardOCP since they were comparing the 2 cards overclocked directly against each other.

The 770 material was just to try and answer the original question.

Rule out the 7950 or any other AMD card if playing Metro Last Light with PhysX enabled.

Quote:
"PhysX is also a key role player for NVIDIA video cards, which is implemented effectively into Last Light. When playing with PhysX enabled, we benefit from many advanced physics effects that we otherwise miss out on. There are a wide span of PhysX effects added in this game. One of the most notable effects is destruction. For instance, if you start shooting at a concrete wall with PhysX disabled there will be some dust and a few small pieces that fall to the ground. Enabling PhysX will show the dust, and smaller pieces of concrete flying through the air and hitting the ground where they are now visually noticeable. Bigger explosions will project debris faster through the air, and even allow big chunks to collide.

PhysX also provides a more realistic smoke, steam and volumetric fog simulation in which players and AI walking through fog will cause and interaction, making the fog to billow and swirl around depending on where it was disrupted from. PhysX is also very notable with cloth. With PhysX disabled cloth may seem stagnant and non interactive with the environment around it, and in other places may remove cloth entirely. With PhysX the cloth will react to things in the environment like wind and explosions, and even fluttering through the air."
http://www.hardocp.com/article/2013/05/29/metro_last_light_video_card_performance_iq_review/#.UeboqxPn-Ul


"NVIDIA and 4A Games did a great job of implementing PhysX. It had an impact on every scene when it was enabled. PhysX brought the world around us to life by doing simple things like adding more smoke to barrels that are burning, and improving the way cloth interacts with people and the environment. PhysX also brought more particles to life. Anytime a bullet hits a wall, there's sure to be debris spewing out and cluttering together on the ground. Fog isn't just stagnant with PhysX enabled, you can clear a path through it just by walking through it. And let's not forget the large particles that come flying directly at you anytime there is an explosion. It is a shame for AMD users that they will miss out on such great technology."
http://www.hardocp.com/article/2013/05/29/metro_last_light_video_card_performance_iq_review/9

yeah. i'm not really a fan of any gpu in particular... if i had my choice and money was no option i'd be all over a titan or 780... or some other nvidia, but those gpu benches from techpowerup don't match any other benching sites reviews. when 10 review sites have one result and tpu has another you get suspicious. that those graphs always show up in a gpu debate bugs the heck out of me because their results are so far outside the normal.

overall the 760/7950 would be the best option for the op at 1080p, but yes... technically speaking if he wants to spend $150 more he can get a 770 for $400 which will do better then either, that's why it's $150 more. However it seems like a LOT to pay for a moderate gain in performance. Particularly since you can get a 7970GE for $50 less.

if the price is equal i'd get a 760 over the 7950... if the price is $20-$50 cheeper for the 7950 i'd get the 7950.

Settings used from one site to another affect the results a lot. Many sites love certain settings, in particular high levels of AA. I personally find 4x MSAA to be the sweet spot and will play with it if I can. Of course if performance is poor, then you can lower it from there, yet at the same time, many sites will use 8x MSAA, SSAA or some other high AA version that just drops performance a ton. Just because a setting exists does not mean you have to use it.