Kepler test build 2.58e Standalone (for GTX 6xx cards)

[madcoo]

Just wanted to say Thank you to Roeland for this Kepler release.

I made the stupid mistake of NOT reading the smallprint in the Commercial download page, and bought a GTX 680, thinking I would divide by 2 or 3 my rendertime...


At least this release enables me to keep working, and I've had no problem with it so far.

Cheers !

[huebner]

don't worry..... i was even more stupid and bought two gtx 680

[madcoo]

don't worry..... i was even more stupid and bought two gtx 680

Oh my goodness !!!
Well, I think it'll be ok if we wait some more...
Most new CUDA cards are based on Kepler architecture (at least, that's what my local reseller told me) : future is in our hands !!!

[mbetke]

How much slower is the 680 for you guys?

[madcoo]

Unfortunately I didn't keep any records of the Ms/sec I had while using my former 2 x GTX560
The latest project I worked on with my GTX 680 rated 19 Ms/sec.

One thing I'm definitely happy with is the ability to load 2Mb scenes now, instead of 1Mb previously

[huebner]

depending on the scene gtx 680 is about 60-70% as fast as a gtx 580

[MattRM]

Hi all,

I will buy a zotac GTX670 4 giga of ram, will Octane support this card ? And the kepler build is the same as the standart build ? like the same improvement and features ?

Thanks

Matt

[madcoo]

Hi Matt, I'm no expert at all but for now you should stick to the GTX5xx series IMHO, as Octane's Kepler Build is only a test build, and will not exploit your card to its fullest capacity.
BUT I'm stressing the fact that I'm not an expert, so ask the other guys in this topic...

Good luck !

[MattRM]

Hi Madcoo,

I have an opportunity for this card, that's I'll take it. But since this is 1 year that the 600 series are released, so the support for the 500 series will be end soon. Hope one DEV can answer me, but when the standard version and the kepler build will be the same ?

Thanks,

As usual, octane rocks

Matt

PS : it's really 4 giga for one gpu, not 2x 2 giga

[roeland]

We don't know yet. The reason why we have a separate build for Kepler support is that it's significantly slower on Fermi cards than the regular build. We haven't found out the cause of this speed drop yet.

--
Roeland