Interesting speed comparison with Octane, if you're interested [very cool! ;)]

Discussion in 'Off_Topic' started by rojharris, Jun 16, 2017.

  1. rojharris

    rojharris Member

    Hi All,

    Just finished rendering out some huge (20,000 x 20,000) pixel renders for posters for a client. The images were of aircraft, so not archvis.

    I used to render these out in Octane because I have twin Titan-X GPUs in my machine and they are super fast right? Er.....

    Small frame around 2000 x 2000 in Octane, render time to noise free: 20 seconds
    V-Ray CPU only on my crappy old 2010 mac pro (2 x 3.46 GHz 6-Core Intel Xeon) and DR to Hackintosh with intel 5960X, render time to noise free: 50 seconds.

    Huge 20,000 px frame Octane time to noise free (remember these titans have 12GB Ram so no HD use necessary): 1 hour, 20 mins **
    V-Ray huge frame render time to noise free: 20 mins! !

    So not only is V-Ray over three times faster at big sizes than twin titans, the end result looks a million times better (of course ;-) ).
    I'm super impressed by that, both the mac pro and the hack are almost exactly the same speed, so not that fast by todays standards and V-Ray still out performed Octane by a big margin!

    NB, this is with NO optimisation in V-Ray, I don't really know what I'm doing there so I just pick default GI and default Bucket AA!!

    Eat my V-Ray dust Octane **

    Can't wait to get my iMac Pro when they come out!
  2. stefan

    stefan main admin here:-)

    so octane 1h20min and vray 20min:)

    thats a nice result indeed.
    (also actuality the old mac compared to 2xtitan gpus)

    by the way, the coming 3.5 with the new unbiased adaptive lights is even faster than 3.4!
  3. Vision4Next

    Vision4Next MC2015

    GPU rendering is nice, but at large sizes there is still room for optimization IMHO. GPU gives you fast previews, but for high quality and large sizes CPU rendering is still very nice and Vray is certainly one of the fastest engines around.
  4. sirio

    sirio Well-Known Member

    A reality check against the marketing BS(like GPU renderers are 10 to 1000 times faster than CPU).
  5. Gnorkie

    Gnorkie MC2015

    I use both Octane and VRAYforC4D and I'm happy with both of them. I flip the coin when I start a work :D
  6. Josef

    Josef Active Member

    while i don't entirely disagree that some marketing teams exaggerate the speed of a gpu renderer, i think from a technical standpoint, gpu is generally a faster hardware than cpu, its the *renderer and how its optimized that makes the big difference.

    that said, a *lot can factor into a speed test like this, each renderer handles things different, for example VRay can chew through an interior render thanks to it's biased / unbiased settings while octane can takes several hours to render even on fast gpu's, it's just not optimized for it., 1/1 comparisons may reveal different results depending on what involved in a scene (lights,geo,glossy reflection etc).
    also Vram can greatly bottleneck big images with octane i found, a little surprised it made to the final rendering to begin with :)

    but regardless i think this a great showcase on what steep optimization can do to the speed of a render engine. and it's something i feel like otoy is no longer doing with octane unlike CG.
    nejck and Vision4Next like this.
  7. pxlntwrk

    pxlntwrk MC2015

    It's been a long time since, I no longer have the courage to do that!
    Dreaming(waiting) for an efficient IPR and material node editor in one of them...;)

    I try to organize my renderings at night ...
    If this calculation is long, I sleep a long time, :)

    But the time I gain with an interactive preview for creation (not tweaking settings), is much more valuable...Just my point of view!

    That said, here are some interesting perspectives on hybrid rendering (cpu / gpu)

    ps/ some news for an VrayforC4d update, night builds, etc... ?
  8. luke.letellier

    luke.letellier MC2015

    I too am tired of marketing propaganda - from both camps - as each user has their own individual needs, and CPU & GPU engines each have their strengths and their weaknesses.

    To their credit, the GPU teams appear to be much more dedicated to C4D than Chaos Group. I really appreciate all the work that Stefan & his team have done for the C4D integration, but it gets *really* tiring to be Chaos Group's annoying little brother that they shove off to the corner.

    Case in point: octane has had fully functional production-ready IPR and nodal materials for several years . The Redshift team generates several builds available to the public each month, with detailed change-logs so that users know exactly which bugs have been fixed when and in which builds. Chaos Group took ~4 months to get (AFAIK) two bugs fixed for us.

    And in terms of render speed, Vray has it's rough areas as well: with default settings, 3.4 really struggles with black objects that have glossy reflections, like the floor in the shot I've attached. I have to crank minimum samples up to 6-8 to avoid getting bucket-error type noise/sampling errors, and that in turn leads to 45+ minute render times for 15K images (even with 7000+ Cinebench points rendering it out via DR).

    To summarize: I love working with Vray - it's my primary workhorse. I just want to be given the same attention as Maya, Max, Nuke, Modo...

    Attached Files:

    HenrikH, Josef, RwVision and 2 others like this.
  9. rojharris

    rojharris Member

    Tell me about it! I agonise over which rendering engine to use for practically every job! I came from modo to Cinema 4D two years ago. In modo we've had a super fast IPR for at least a decade and it was alway something I took for granted. When I came to Cinema I was blown away by the ease of use and the procedural modelling workflow and how it handled huge models with ease, however I was stunned at how clunky and awfully slow its rendering engine was. Since then I've been on a never ending quest to find something that works as well as modo's used to. I have tried most with varying degrees of success.
    Octane is brilliant for its instant feedback and WYSIWYG IPR, but then it can be terrible with noise and the renders look a little 'rendery'. Thea has a fantastic IPR that's both CPU and GPU together and is faster than Octane, but the renders look pants! Corona is pretty damn fast and the renders are gob-smackingly gorgeous and it integrates into Cinema better than all the others, but there's no IPR and it falls down on caustics and volumetric stuff, which can take ages to resolve. V-Ray, well, on the one hand its very very fast for most (not all) things, on the other it has no IPR (that is 100% WYSIWYG), its also brilliant at fast SSS and caustics and has tons of built-in shaders etc. The renders look fantastic, sometimes even better than Corona. I've never tried Red Shift (I'm mac based) and I briefly tried Arnold but holy-moly that was noisy.

    So at the moment I often end up using Octane if it's an urgent job and the frames aren't too big... Corona if I have more time and it's for stills and V-Ray when I have even more time and/or there are SSS or caustics or volumetric materials involved or its a massive frame size, which V-Ray handles with ease.... The last few jobs I've done I've spent a day testing the same scene in all 3 then picking the best compromise of speed vs looks! In the end its never mainly about the rendering time though as there's always overnight, its about the set up and testing of materials and for that you need a fast and stable IPR.

    I'd love to be able to use V-Ray all the time as its by far the most mature and full featured but until Chaos gets their fingers out and builds us an IPR (CPU, not GPU) that works, then sometimes I have to resort to the others.
  10. stefan

    stefan main admin here:-)

    "To summarize: I love working with Vray - it's my primary workhorse. I just want to be given the same attention as Maya, Max, Nuke, Modo..."

    Chaosgroup does work on that, and we work hard to get them to it too as soon as possible. target of Chaosgroup (and us) is to get it full same as Max/Maya V-Rays as far i can see the time tables and work they do, the gpu/ipr stuff should come same or very close within this year to max/maya vray.
    (what is needed is a special c4d node, that is full capable for interactivity).
  11. luke.letellier

    luke.letellier MC2015

  12. Josef

    Josef Active Member

    Can't see that happening without the c4d port being part of the development pipeline (in house @ CG), just a speculation!
    Rendermann likes this.
  13. stefan

    stefan main admin here:-)

    well Chaosgroup does do these development (ipr, gpu) parts we all wait on, since v3.x, so it IS already part of the pipeline already. (that did and do speed up some things a bit)
  14. Josef

    Josef Active Member

    should've phrased myself better, i know CG contributes to big parts of the development that affect the c4d vray here, i think that's common knowledge now?

    What i meant originally, as vray for for c4d being a product that only CG fully develops and distributes and handles documentation and support for, like vray for max,maya,modo,nuke etc, another product in the "line"
  15. stefan

    stefan main admin here:-)

    well i am not sure as the wait often also come from them, maybe because the Maxon market in c4d is very small in comparence, but anyway.
    we managed to get them working more on c4d needs,and still try to get more, which also shows in more and faster result in c4d specific needs the c4d version gets now from them finally. (for 3.4, 3.5,..etc).

    overall the software is now very near the maya max verisons, the missing ipr/gpu things chaosgroup works on anyway, as these parts are needed by them.
    and of course we together with CG work on adding even more features to 3.x versions and next year v4.0.

    please overall stay on the topic from the origial poster which is the speed comparison he made. this shows greatly what the new c4d brings on speed and features,thanks togetting the very recent sdks and support from CG.