Nvidia FERMI cards are coming... tomorrow !

[pixelrush]

Really I am disappointed with the specs of these cards..
Very hot ,very noisy and not as fast as they should have been, also more expensive than the competition.
95 deg C, 70db, 850W for SLI system under load!...perhaps I'll hold off for a die shrink...
Still thankfully the 470 is 9.5" and requires 2x 6 pin plugs so it will fit my case/power supply.
I see raytracing is 8x faster, and folding@home is 3.5x faster for 480 than 280 - so maybe the Octane performance gain will make the proposition *passable*

BTW I see various reviewers lack a benchmark for Cuda performance comparison. Radiance might like to send out Octane with a std test scene to see if it could be adopted- good for publicity too

[radiance]

Really I am disappointed with the specs of these cards..
Very hot ,very noisy and not as fast as they should have been, also more expensive than the competition.
95 deg C, 70db, 850W for SLI system under load!...perhaps I'll hold off for a die shrink...
Still thankfully the 470 is 9.5" and requires 2x 6 pin plugs so it will fit my case/power supply.
I see raytracing is 8x faster, and folding@home is 3.5x faster for 480 than 280 - so maybe the Octane performance gain will make the proposition *passable*

BTW I see various reviewers lack a benchmark for Cuda performance comparison. Radiance might like to send out Octane with a std test scene to see if it could be adopted- good for publicity too

i'd love to but i'm %200 occupied finishing beta2.

is there anyone interested in coordinating something like this ?
contacting hardware review sites and giving them a demo version of beta2 once it's released, with a decent test scene ?

Radiance

[pixelrush]

Well its certainly too good an opportunity to pass up.
I think you will need to provide something a bit slicker than a toy aeroplanes scene though...

See here 'Tech Report' typical page of review benchmarking 3d modelling and rendering for processor comparison
http://techreport.com/articles.x/18581/13
Q. Who uses pov ray anyway??

or part of the AnandTech GTX480 review http://www.anandtech.com/video/showdoc.aspx?i=3783&p=6

[radiance]

i've got heaps of complex scenes, but they all have commercial bought models in them, so they can't be redistributed publically.

maybe we can use one of the CG talk lighting challenge scenes ?

Radiance

[pixelrush]

I dont think it need be very complex, only original and good eye candy for print/web.
You see in the linked examples the scenes are not staggeringly epic..
Probably you dont want to distribute too large a benchmarking package either.

The distinguishing feature I think is that it should be heavy enough to produce a time to say 4000 samples that it can be used meaningfully to compare GPU say for the next 5 years as speed increases, and also not take too long so that the reviewer can do their tests quickly. So say some scene that is *finished* in 20 mins? (currently that is, on a GTX260).
I realise 20 mins is not especially flattering of Octane now and many images can be done in less but you need to think of a 10-20 fold speed up in the future..

[pedrojafet]

OpenCL benchmarks (ATI can run openCL too) shows that the new GTX480 are like 2x faster than the higher end ati card

2x faster??? What does it means in the overall software running process, like i.e., with Octane???

[pixelrush]

Looking around it seems EVGA have the best Fermi solns for *serious* Octane purposes ATM, if you dont mind the extra expense - watercooled single slot cards - http://www.evga.com/articles/00539/
...the radiator with the heat of 3/4 cards going to it will need quite a lot of air though...still thats 1440/1920 cores blasting away...

[Sam]

Before speaking about benchmark scene
We should wait for MLT

Its not good to have multiple versions of a benchmark utility
Look at Cinebench, they update it only every years (when they release major C4D updates)

[RayTracey]

if you want to see how fast Fermi can do raytracing and pathtracing be sure to check out this video: http://www.youtube.com/watch?v=nEXeTXmCfdE

Can't wait to see Octane running on one of those