Kepler test build [Obsolete]

[glimpse]

..if the cost of 'good enough' Tesla will going to be equal or even less then building new rig with multiple gtxs - i'll step for buying one =) otherwise start to look for590 to replace my 460 =) - with that i can cut times to one third or even fourth of what i have now =)

i'm in love with simplicity & straight forwardness of this package! Just hope it will remain to be as good as it is or even better.. =) time will show, for sure!

[manalokos]

Why is the Kepler build slower? Is the hardware in fact slower, or is it the Octane render not yet completely compatible/capable of using 100% of the possible speed with these new boards? If they are really slow, they wont have much of a market in the gaming industry.
If they can be super fast in games, they can be super fast in Octane for sure!

Best regards
Filipe Alves

[glimpse]

Why is the Kepler build slower? Is the hardware in fact slower, or is it the Octane render not yet completely compatible/capable of using 100% of the possible speed with these new boards? If they are really slow, they wont have much of a market in the gaming industry.
If they can be super fast in games, they can be super fast in Octane for sure!

Best regards
Filipe Alves

For games they are faster..but games and computational work takes different things into acount, that's why playing games with Quadro or tesla isn't that good of idea - you pay more, but they perform less..than GTX cards made on the same architecture =)

It isn't known for sure right now, but it (kepler) has a bit different architecture compared to Fermi, You can google up those reviews..- time will show..but it seems (though we can be wrong) that nVidia is trying to separate their lines as much as possible in order to sell some pro line models for a bit higher prices =) simple differenciation - You get what You pay for.

Maybe cuda5 will be optimised better to talk with new Kepler's architecture, let's say in more efficient way and we're going to see some performance boost when Octane will be rebuild on that =) who knows?..i'm not a programer to talk about that..

[marcio_max]

I think the best option to replace the GTX 680, is to buy a GTX 580 3GB and GTX 590.
So technically you would have the processing power of the GTX 680. It has been decided! That's what I do! =/

[newmountain]

Roeland, The Kepler build is running great on my new laptop.
Asus G75V, I7-3610AM / 8gb ram / Nvidia GTX660M / 17" 1080P display
I wanted the the Samsung Series 7 Gamer with the GTX680M (rebranded Fermi gtx580m) but could not get it to land on my doorstep in time, as I needed a machine for a business trip.
The Asus is ugly as sin, but is proving to be a hell of a machine for the price, and is running cool and quite.

Octane is performing well on this machine, considering it is a laptop.
.81 ms/s with benchmark scene, pathtracing, default camera position.
2.83 ms/s direct lighting
Minimal to no lag while navigating the viewport.
Hopefully it will prove useful for material creation, preliminary scene set ups, and working away from the office.
Thanks for this version!

[roeland]

I posted an improved build. It is not on par with the regular builds on the Fermi cards yet but it should be significantly faster than this build.

--
Roeland

[huebner]

hello

thanks for the update!
i have done some tests with two scenes. keppler testbuild 02 seems to be about 1.5 x faster then testbuild 01.
see the pictures below.

[bb3d]

Thank you very much for this test. Now it would be interesting to make a comparison based on the benchmark scene, so we could see the relation to the Fermi-based cards in this thread:
http://www.refractivesoftware.com/forum ... f=9&t=7383

[huebner]

gtx 680

....benchmarkscene 2.5 Ms/sec.

[PeterCGS]

That sounds pretty reasonable considering Roeland said the performance is about 60% of a 580 GTX... 60% of 4.14 is 2.484 Once its over 3.0 I think I'll go for a couple of 690 cards (that will hopefully have 8GB RAM by then LOL)... Double amount of textures, and lesser power draw is pretty important to consider!