View Single Post
Old 03-22-12, 10:48 PM   #97
i SPY
007
 
i SPY's Avatar
 
Join Date: Apr 2007
Location: You were sayin'
Posts: 290
Default Re: Official GeForce GTX 680 Review and Discussion thread

^
Yeah that demo looks nice, cant wait to try it out

Quote:
Originally Posted by Johnny C View Post
680 is a mixed bag. I read through a few reviews and it seems like most of the time it's a little faster..and at a cheaper price it's a good card.....but it's not a game changer. I want cards that are 400 bucks for top end. (amd and nV)

Yeah, if it were ~ 320-350€ i might consider it too, but not for ~490-499€ (that's what over 640$ lol

I checked techpowerup review as the last one and i had a hunch its gonna be like that. Its only 16% faster then 580 GTX, Ok it can be little more here and there.. But that made me think for a bit, exactly that's nothing special for next-gen "high-end", same goes for AMD 79x0..
I payed 320€ for mine, ok 250€ -tax (got a good deal) and now to pay 2x more just to have ~15-20% better perf. VS this 570 at 900mhz OC? -_- Im at lower non HD reso anyway..


http://www.techpowerup.com/reviews/N...TX_680/27.html

The only thing that made me itch was BF3, but imo that's to be expected. They knew its top sheit and made some tweaks especially for that (Fillrate, FP16 +aa) lol no doubt about that.
But then to have lousy openCL perf. in other stuff and overall "similar cuda mix" like GF114 is a bit lame tradeoff. Yeah they had to make that approach to eliminate no hotclock freq.. but only 8 polymorph 2.0 tessellation engines? at least the same 16 would be nice, esp. for that "high-end" price.






Anyway this looks cool, but its not available in directX API yet..
Quote:



Prior to Kepler, for the GPU to reference a texture, it had to be assigned a “slot” in a fixed-size binding table. The number of slots in that table ultimately limits how many unique textures a shader can read from at run time. Ultimately, a shader was limited to access of just 128 simultaneous textures in Fermi (which aligned with the API limits of DX11).

With bindless textures in Kepler, this additional step isn’t necessary: the shader can reference textures directly in memory, making binding tables obsolete. This effectively eliminates any limits on the number of unique textures that can be used to render a scene: over 1 million unique textures are possible.
Besides dramatically increasing the number of available textures, another added benefit of bindless textures is reduced CPU utilization:

At this time, bindless textures is only exposed in OpenGL. But in the future it’s possible this could be exposed in DirectX via NVAPI, or future versions of DirectX.

//from whitepaper
__________________
intel Q9450 @ 3.656Ghz [1.3875v, LLC off]| GA-X48-DS5 [Memory Enhance: Turbo]|MSi N570GTX TwinFrozrIII OC/PowerEdition|Kingston HyperX 4x2GB PC 8500 @ 1097Mhz [5-5-5-18, 2.25v]| Creative X-FI Pro [SB046A]| Tagan PipeRock 600w [48A]

i SPY is offline   Reply With Quote