First Tomb Raider Patch here

Performance loss may be roughly the same, but the 6xx series is noticeably worse at compute than AMDs 7xxx series, so if more games start using compute for physics then AMD will be comfortably ahead. Hopefully Nvidia will learn a lesson and not screw over consumers by unnecessarily gimping a series of cards on purpose to encourage the sales of its overpriced higher end cards.

 

is the 500 series (GTX 580 specifically) better at DirectCompute then the 600 series?

 

I do believe they are, the 5xx series is also better at PhysX calculations.

 

maybe that's why it seems like 600 series owners are having more issues with Tomb Raider...with my GTX 580 it ran fine with tessellation and TressFX enabled...only issue I had was that SSAA did not seem to work

 

It could be a typo (yet the v314.07 drivers are WHQL certified) but HardOCP mention v314.70 betas in their HD 7970 review here: http://www.hardocp.com/article/2013/03/11/asus_hd_7970_directcu_ii_video_card_review/2#.UT8tBFdjj8k

*EDIT*
It has to be a mistake as they mention using v314.70 WHQL in the Video Configuration menu so they probably meant v314.07 WHQL after all.

Oh well, for a moment there I thought we were going to get a new driver today.

 

5xx series have better double precision performance, for single precision performance the 6xx series cards easily outperform the 5xx series.

Actually Kepler performs better with PhysX
http://www.geeks3d.com/20120502/physx-test-gtx-680-vs-gtx-580-vs-gtx-480-in-fluidmark/

 

so do most games make better use of single or double precision?...meaning how does this effect gaming and which is more important?

 

Double precision is for the high performance computing servers (super computers) and serious processing. Games don't need that kind of precision and the tremendous performance hit that comes with it.
You'll be seeing single precision in games.

 

We can talk about that when it gets relevant one day. It's clearly not in this case. And another question is if it will actually be a bottleneck at all, the GPU does a lot of other things in parallel when gaming. AMD is also slower in tesselation for example, it's not a bottleneck except you're going way too far though (see Crysis).

 

I've seen Kepler struggle with OpenCL, but it does seem to perform pretty well under DirectCompute:
http://www.videocardbenchmark.net/directCompute.html

also AnandTech's 680 review showed 2 DirectCompute benches where they (680/7970) traded leads.

guessing it all depends on algorithm / implementation.
currently OpenCL isn't used for gaming, so only DirectCompute and CUDA performance matters.

 

Nvidians...will suffer....TRR was meant to "putsch".....

 

this passmark bench is useless and doesnt say anything, I mean look at 8400GS - its better then 580GTX?? I highly doubt it

 

Latest beta, all @ max, MSAA 4x: I have glitches on hair...
Working good with 2x, FXAA or without patch...

 

patched up and still getting tressfx hair flickering with sli

haven't played extended period of time to see if it still crashes at random.

and lastly, any downsides to using hitman absolution profile?

 

lol, not sure how 8400gs got up there. hovering over it shows "number of samples: 1" so maybe that's one magic card

anyway, ComputeMark v2.1 results show similar performance differences.

Spoiler

so my point is still valid. while kepler's OpenCL results suck (for whatever reason), DirectX11 Compute results aren't nearly as bad.
also pcgameshardware's TressFX results in TR pretty much confirm that.

 

New patch has just gone live, they also mention new beta drivers getting released.

PC Patch Released - 1.0.722.3 - March 15th 2013

 

314.21?

Not on Nvidia's website

 

They should be soon.

 

Cool, glad they listen, i reported both issues as well.

I hated all the people with mics and always on, i muted them all xD

Also had some crashes in MP..