7950 Boost slower than 670 or 660ti in Crysis 3

[GrantMeThePower]

Did you guys see this?

http://www.hardocp.com/article/2013/03/12/crysis_3_video_card_performance_iq_review/4#.UT9vrKUTvUY

 

[KevinH]

Those aren't the same settings. HardOCP's #'s can be confusing to read at first. They will up the settings to the highest playable level.

 

[lehtv]

Read up on HardOCP's methodology, OP.

http://www.hardocp.com/article/2013/03/12/crysis_3_video_card_performance_iq_review/2

 

[Grooveriding]

TXAA - The available modes are TXAA Medium 2XT 1 and TXAA High 4XT 1. TXAA only works on NVIDIA GPUs and is NVIDIA's invention and is called Temporal AA. This mode takes a large performance hit, but it also causes extreme texture blurring, ruining your gameplay experience. This mode should be avoided.

The very first thing you will notice is that turning on TXAA causes everything to be very blurred. You will notice loss of detail on the building, blurring of the grass and trees, everything. It does a lot to reduce aliasing, but at the cost of severe blurring. In the second screenshot we again see that it does a great job removing aliasing, but that tree is just way to blurry. In the third screenshot again, all the extreme angled lines are smoothed out to a great extent, but the textures are showing less detail.

This last screenshot demonstrates up close how detail is lost in the high resolution textures due to the blurring. The stone wall simply has less detail with TXAA enabled. Also, the vegetation here is extremely blurry with TXAA enabled.



Overall, TXAA should be avoided. We don't know how this quality of TXAA got implemented into the game and left there. It seems like somebody took the day off when quality control of TXAA image quality rolled around that day. TXAA destroys the gameplay experience in Crysis 3.

:thumbsup:

 

[Lonyo]

NM misread their AA nomenclature.

 

[GrantMeThePower]

Those aren't the same settings. HardOCP's #'s can be confusing to read at first. They will up the settings to the highest playable level.

I understood that. If you look at the settings, they are much higher for the 670 than the 7950 and the 7950 is still slower, even at the reduced settings.

 

[Xarick]

This is not news. Remember.. everyone acknowledges the 7950 is slower than the 670. Until you overclock.. that is where the value comes in.. cheaper and overclocks past the 670.

 

[notty22]

I understood that. If you look at the settings, they are much higher for the 670 than the 7950 and the 7950 is still slower, even at the reduced settings.

Here's Apple to Apple

​

(Click Graph for Larger Image)​

This is where things change, the Radeon HD 7950 Boost as previously discussed seems to be underperforming in this game. The HD 7950 Boost is 23% slower than the GeForce GTX 670.

In fact, the GeForce GTX 660 Ti is 12% faster than the Radeon HD 7950 Boost! There is either something wrong performance on the HD 7950 and HD 7870, or these video cards are just slow for this game and the GTX 660 Ti and GTX 670 are simply better.

 

[Rvenger]

And it is faster than the 670 clock for clock. If you match the clocks on the 670 and 7950 you will find that it outperforms it by about 5%.

 

[notty22]

And it is faster than the 670 clock for clock. If you match the clocks on the 670 and 7950 you will find that it outperforms it by about 5%.

A new metric.

 

[GrantMeThePower]

This is not news. Remember.. everyone acknowledges the 7950 is slower than the 670. Until you overclock.. that is where the value comes in.. cheaper and overclocks past the 670.

Right...but for the 660ti to be so much faster than the boost version of the 7950 is pretty surprising...no?

 

[Redmenace]

I noticed an odd thing about a Boost Sapphire 7950 I just added as the second card in a crossfire set up with a vanilla Sapphire 7950 card. I have the clock set to 970 (limited by the vanilla card) but noticed that the Boost card will occasionally drop down to 850, even though card 1 is locked at 970 (per GPUz log). Is that the expected behavior of a Boost card in crossfire?

 

[BallaTheFeared]

Win some, lose some, sold my code anyways console port ahoy!

7950 would need nearly 1200 and a mem bump to much the 670 I believe at least in this title.

Btw when people say they're close, it's overall some games nvidia is faster some AMD.. Some kind of delusion is taking place where every game is neck n neck and it's simply not the case.


Looks like AMD didn't continue to refine TXAA after they bought out EA, can't imagine why >.<

 

[3DVagabond]

Right...but for the 660ti to be so much faster than the boost version of the 7950 is pretty surprising...no?

It's one game. There are games where the opposite is true and lower tiered AMD cards are faster than nVidia cards. While interesting, it's not a big deal, really.

 

[Rvenger]

A new metric.

I don't know what that means but I tested it out myself. So its truely not BS.

 

[Tweak155]

I don't know what that means but I tested it out myself. So its truely not BS.

Matching the clocks is a new metric... what's not to get?

 

[notty22]

Nvidia is allowed to lead performance in some games, the cards do trade blows in other games?
http://www.techspot.com/review/642-crysis-3-performance/page4.html

 

[sublime420ec]

i play crysis 3 at very high settings 16xaa and 2x msaa 1920x1080p and i hardly ever dip below 50fps i get up to 80-81fps in the cavern map on mp... i'm not even on beta drivers on 13.1 card is slightly overclocked stock voltage... smaa seems to really tax any card been in the 30-45fps range with 2xsmaa... lol >_<

card is HIS HD7950 turbo 3gb

 

[Rvenger]

Matching the clocks is a new metric... what's not to get?

I always thought matching the clocks was a good perspective of performance when it comes to overclocking. The GTX 670 is good card, but if you want to overclock to squeeze every last drop of performance out of a card, the 7950 is incredible and wayyy better Perf/$.

 

[thilanliyan]

And it is faster than the 670 clock for clock. If you match the clocks on the 670 and 7950 you will find that it outperforms it by about 5%.

No point in going clock for clock if the architectures are so different. Remember the "double pumping" of shaders?