would the maxwell cards ever beat the kepler cards(in rendering)

my main point is not a definite guess, but rather what makes the maxwell cards weaker than the kapler cards(specifically the GTX 780 and the GTX 970)
one thing that i know is that the maxwell cards has less cuda cores and lower memory bandwidth,
but since they are physical things they could not be changed. that i my reason of doubt, and so i ask, is it possible that the render times in the future of the maxwell cards beat the kepler cards

My 2 cents, I think that the speed increase that the Maxwell cards get for gaming is that they compress the memory transfers in the cards. This is why they can use the smaller 256 bit bus but still be faster than the Kepler cards with larger buses. I suspect that once Cycles and other renderers start using compression for memory transfers that the performance will increase. What I don’t know is how this compression is done, software or is it special hardware in the Maxwell chips? Take this with a grain of salt, as I don’t really know. :wink:

i’m quite the weird hardware mania, so i know quirky stuff, so i can say, the if the slow render times are caused by the compression, then that i unlikely to happen.
it takes a company time and money to invest in it, and though it could be done in software, it would be slowed down, and NVIDIA wouldn’t hand over the compression method that easily, because it mainly matter of concept, so if you got it, you can make your own version, and they can’t sue you.
so no, they wouldn’t hand over the compression coding. it done in the firmwares
and i don’t think they would be pressurized to make a firmware update for the compression coding, as AMD’s side is really not putting up a fight.
so let’s just hope for the gtx 980ti(i doubt so)

or let’s hope it’s not about the memory bandwidth

This is just stuff I read from Nvidia’s press releases, so like I say, I don’t really know how well this stuff will work with a renderer. Obviously it makes a massive difference with games. Here they call it “color compression”:

Edited to add: I’m thinking that the newer or a new versions of CUDA will give the devs access to this compression functionality.

the color compression doesn’t help with renders.
renders are all kept internally, and doesn’t have a massive bandwidth output.
the colour compression thingy is about compressing the output from the GPU to the memory buffer, to the ports, and this compensates having lower bandwidth, but keeping the same bandwidth to the ports.
so as long as you’re not updating your screen 60HZ with very different frames each refresh, it’s useless

but good for games, because it happens almost 99% of the time when you’re playing a game
but not renders