New gtx 680

[nildoe]

Sweeeettt...

i can't wait for t_3 to come here with his experience!!

Nildo

[MrMikeEsq]

Hmm, I'm curious as to what the performance is like! I have a 3gb 580 at the moment, direct lighting on the demo is good speed considering the visual quality, but path tracing / mlt just isn't usable in the timeframes I work with. I wonder what difference a 680 would make, either instead of my 580 or paired with it! It may be an incentive for me to buy octane

[t_3]

sadly no "real" numbers for now. octane detects the card but doesn't render; the gpu just fails in octane without shooting a single ray... so i guess we need to wait for a new octane build

in the meanwhile some more synthethic benchmarks (sisoft sandra 2012, raw cuda sp shader performance):

Code: Select all

gtx 560 ti @ default: 1065mpix/sec
gtx 570 @ default:    1520mpix/sec
gtx 580 @ default:    1680mpix/sec
gtx 580 @ 850mhz:     1850mpix/sec (also vram oc)
gtx 680 @ default:    2750mpix/sec = 1.64 times a gtx 580 or 2.58 times a gtx 560 ti at default clocks
gtx 680 + 300mhz:     3250mpix/sec (+15%)

cool thing -literally- are the temps - even after running some benchmark cycles the card didn't get past 70°C with the standard fan rpm curve (maxing out at around 70%) and is quit silent at the same time. looks like these will stack up nicely think i will keep it, even if the wait for a new octane build may last a little...

... so what about a nightly octane build?
just kidding



ps: i have two gtx 570 2.5gb for sale, drop me a pm if interested

[michaelkdaw]

Thanks for posting the numbers. They provide a bit more optimism after seeing the OpenCL results.

You haven't had any problems running it at a 165 mHz overclock?

I hope you get a chance to use it in Octane soon...

[t_3]

You haven't had any problems running it at a 165 mHz overclock?

at least not while the rather short benchmark cycles. but expect to see this change, if pushing it with octane. but if it is really that powerful *praying* it won't be necessary to oc anyway...

[Jaberwocky]

T_3

have you downloaded the latest drivers from the Nvidia web site?

I think they just came out at launch.Maybe that might help get Octane moving.

If that doesn't help, then the radical change in architecture on the chip is probably responsible for Octane not working.

Perhaps Radiance and the team need to invest in a GTX680 for their office Test bed before they release the next version.

[lixai]

You haven't had any problems running it at a 165 mHz overclock?

at least not while the rather short benchmark cycles. but expect to see this change, if pushing it with octane. but if it is really that powerful *praying* it won't be necessary to oc anyway...

hey, thanks for the numbers, can you try to render with iRay from Max, or RT from Vray, i think if you can benchmark using those renders, the resoults must be the same, they both use brute force to compute. thanks

[t_3]

T_3 have you downloaded the latest drivers from the Nvidia web site?

yep 301.something, dating 22.03. most probably there is a new cuda build necessary. would be interesting to have a small or even tiny statement from behind the newzealandian wall of silence, but i wouldn't mind if they feel better need to tell us about instancing and the likes first

hey, thanks for the numbers, can you try to render with iRay from Max, or RT from Vray, i think if you can benchmark using those renders, the resoults must be the same, they both use brute force to compute. thanks

sorry have no access to it. but i guess they will need another build based on an updated cuda sdk too?

[lixai]

t_3 you can download 3ds max trial version if you want, it comes with iRay.

http://usa.autodesk.com/3ds-max/trial/

[t_3]

t_3 you can download 3ds max trial version if you want, it comes with iRay.

http://usa.autodesk.com/3ds-max/trial/

i'm really sorry, but i can't do that i used to use max when it was called 3d studio and running as dos program. they never reworked their ui basics since then. my head would likely explode if i need to operate it now

just kidding; i'll see, what i can do...