Vray/VrayRT vs Octane

[stevesideas]

Am I right in thinking that Octane pretty much does what RT does, but a bit better ?

Will Chaosgroup be rising to the challenge of getting RT on par with Octane ?

The reason I'm asking is because after listening to the Twistedpoly interview he basically said that the reason Octane is great is because it's instant feedback. Isn't that what RT is supposed to be ? I've not used Octane myself so just wondered if anyone had and what they thought about it when compared to vray or vrayRT ?

[glorybound]

I worked with a company that uses Octane; actually I took a job over for them. They requested 6k and I told the what the render times would be, for me, using V-Ray (13 hours). He uses Octane and said that it's a 15 minute process.

[Neilg]

I worked with a company that uses Octane; actually I took a job over for them. They requested 6k and I told the what the render times would be, for me, using V-Ray (13 hours). He uses Octane and said that it's a 15 minute process.

Surely you must know that this is a pointless comparison? I bet i could make a vray render that takes 15 minutes at 6k too.

[glorybound]

I think he said that they paid 15 thousand for their video cards. Their renders were okay; nice and clean, but nothing fancy.

Surely you must know that this is a pointless comparison? I bet i could make a vray render that takes 15 minutes at 6k too.

[vlado]

Like with most things, it would be best to try and see what you get. Both are interactive and adjustments are pretty quick. V-Ray RT can use the light cache, which makes interiors much easier to render. V-Ray RT also supports interactive DR, if you want to use something like the nVidia VCA.

Best regards,
Vlado

[Morbid Angel]

Yeah I bet they have 10 gpus in there rig which of course would render anything in 15 min. Octane as a renderer being always developed for gpu, will probably be always more advanced then vray RT. I find it really complicated though, I'm not sure if it plugs in directly into max, but its a standalone with a million of options of its own - now I already have to keep max, maya, houdini, phoenix, mel, python, nuke, photoshop, mudbox, zbrush and whatever else in my head - having to learn another UI, another language, another renderer is just getting too tiring and too old! That's just my opinion of course.

[Mokiki]

There is an Octane version that plugs directly into max. I think I tried the demo at one point and as far as I remember it was definitely cool to play with. The way they handle bump maps seems to be better/differently at least I managed to get effects like this: http://www.rgbstock.com/cache1nvaQ1/...00/mf4OOnQ.jpg directly through a bumpmap without playing with any anisotropic rotation maps which I thought was pretty cool.

But like Morbid Angel said octane is progressive gpu render from the ground up so it shouldn't be a big surprise that it does things better in the "realtime" department.

[Neilg]

Octane's light field format seemed interesting, i've had my eye on it for VR since early last year.
They're not letting people store the rendered files locally though. to access them it requires an always online connection to stream the data from the image format after it's rendered, so whatever. I had a call with them recently asking when it would be available (late this year, their 'octane vr' out soon for free just renders cubemaps lol) and they're going even more cloud focused in the future so it's off the cards for us.

Vlado, any chance you could figure out how their light field image format works and copy it? If it takes 4 days to render and results in a 200gb file that is of no concern to us.

[3LP]

I think he said that they paid 15 thousand for their video cards. Their renders were okay; nice and clean, but nothing fancy.

Well I can easily imagine that GPU is faster but as allways, it all depends on how much render power you trow at it, for any renderer.

For the time being, you can easily add multiple gpu in nodes and get a really great render farm for cheap, but you'd be limited with the features of the renderer.

We are currently building a 7x Titan X render node that will be up and running in a couple of weeks, and our estimates are that it will rock, but it will not be faster that our cpu farm (50 nodes) so again, all depends on the power you allocate for you frame to render.

But with few of those node, I can easily understand that GPU would render in couple of minutes where your workstation would be few hours.

Stan

[kosso_olli]

All I can say is that I was really surprised how fast RT/GPU is nowadays with the newest version. Had a car in 8k resolution. Nothing fancy in terms of complexity, only an HDRI and the car. But with mblur and GI. Got an accepable clean render in 30 minutes on a single Titan. I remember trying this a year ago or so, and a lot of features were missing and the speed was bad. So thumbs up to the dev team, great work!
However, you have to be really carefull with GPU memory. As soon as you hit the video memory limit, the render comes to a halt. I tried turning on displacement and the render didn't even start. But I am sure that there a cards coming up soon with loads of video ram.

[vlado]

That's very good to read I hope that the next release will be even better.

But I am sure that there a cards coming up soon with loads of video ram.

You can bet on that

Best regards,
Vlado

[savage309]

Really glad to hear that feedback kosso_oll,
I am having pretty much the same observations and as far as I am concerned, V-Ray RT GPU should be the fastest and the most feature-complete GPU raytracer out there (at least the version in the nightlies, which will soon become official) - compared to the others which are being hyped and I've tested. It is unfortunate, that people are sometimes comparing old versions (or same versions on different hardware) and making conclusions out of that.