Loading...
> >

New Radeon Pro Duo coming down the pike... 32 GB GDDR5 for $999 MSRP!

2»

Comments

  • kyoto kidkyoto kid Posts: 42,408
    scott762_948aec318a said:
    kyoto kid said:

    ...so like the old (terribly overpriced) Titan Z it is actually two separate GPUs permanently linked together.  I remember Nvidia touting the Titan-Z as having 12 GB when for rendering purposes, it only had  6 (though twice the cores of a Titan Black).

    Still at 999$ it is 1,500$ less than Nvidia's only 16 GB card which is in the more expensive Quadro line (P5000 which retails for 2,500$) and 200$ less than the 12 GB Titan Xp. Will be interesting to see how Nvidia responds.

    May need to dust off Reality and Lux.

    I'm very interested to see how they respond.  Very interested to see how fast the Titan Xp is with Iray, of course the price is a tough pill to swallow.  This AMD card may drive the prices down.  :)

    The scenes I'm rendering might be too much for the 1080 Ti - I have Nvidia Inspector which shows mem usage, but I'm not sure how accurate.  Some scenes with 4 Genesis 3 figures reporting 11GB usage.

    ...I'm in the same boat as I create highly detailed scenes. For example I cannot abide by city streets that look clean enough to eat off of so I litter and dirty them up with all sorts of rubbish props and grime overlays which add to the geometry and texture load. If you render in large format for printing, texture detail becomes more important.

    Technically, for my needs a Quadro P5000 would be sufficient to render 98 - 99% of my scenes, but 2,500$ is a pretty steep price.  I really don't see releasing a 16 GB Enthusiast card with a competitive price with the Pro Duo, especially after the 1080 Ti pretty much undercut the Titan-X. 

  • Kevin SandersonKevin Sanderson Posts: 1,643

    Yes, scripted rendering does KK.

  • tj_1ca9500btj_1ca9500b Posts: 2,057

    So I just grabbed ths product: https://www.daz3d.com/scene-optimizer

    It's part of the Barbarian sale.  For those of us that have scenes with a lot of stuff in them, it looks like it'd be quite helpful  We shall see if I can figure it out and how well it works.

     

  • CoolBreezeCoolBreeze Posts: 207
    edited April 2017

    .

    Post edited by CoolBreeze on
  • kyoto kidkyoto kid Posts: 42,408
    Kevin Sanderson said:

    Yes, scripted rendering does KK.

    ...thanks, that's what I as hoping.

  • linvanchenelinvanchene Posts: 1,386
    edited April 2017

    edited: kyoto kid is right, there is no information so far that combining VRAM of dual cards or two separate cards will be supported

    - - -

    @ AMD and Cuda

    enter "amd cuda" in google and you will find that a lot of different groups have been working on compatibility.

    https://www.google.ch/search?q=amd+curda&sourceid=ie7&rls=com.microsoft:en-GB:IE-Address&ie=&oe=&gfe_rd=cr&ei=cnsEWcO7KYHZ8Aep-biQBg#q=amd+cuda&spf=375

    Allready in 2016 Otoy have figured out a way to run Cuda applications on AMD GPU.

    compare:

    https://www.bit-tech.net/news/hardware/2016/03/11/otoy-cuda-compiler/1

    This means:

    @ AMD and Otoy OctaneRender

    - Sooner or later OctaneRender could be run with most of the features working on AMD cards

    The latest post on this was four days ago on April 25th 2017

    "This morning we validated that Octane 3.00 AMD test builds are working with the current/latest public win10 drivers (and resolved why it was breaking before - which is good). We're now focused on updateing the test builds to 3.06+. We found one of the features in 3.04/5/6 is not running and are looking into it. We think there is a workaround. We have reported the issue to AMD. We hope they fix it soon. [...]."

    source: https://render.otoy.com/forum/viewtopic.php?f=9&t=60274&p=311853&hilit=amd#p311853

    @ Nvidia Iray and AMD GPU?

    Otoy and Nvidia seem to be exchanging information. As far as you can gather Nvidia did not object to Otoy making Cuda available for AMD cards.

    This means in theory nothing is preventing Nvidia to include basic AMD GPU support for Nvidia Iray in the future as well.

    - - -

    Commentary

    Will all companies involved deceide to cooperate and provide products and solutions that are useable for their customers?

    I hope Nvidia, AMD and Otoy deceide on cooperating with each other to make GPU rendering more viable.

     

    Post edited by linvanchene on
  • kyoto kidkyoto kid Posts: 42,408

    ...again the card is a dual GPU so like the Titan-Z, only half the advertised VRAM (in this case 16 GB) is available for rendering. 

  • Daywalker DesignsDaywalker Designs Posts: 3,586
    kyoto kid said:

    ...again the card is a dual GPU so like the Titan-Z, only half the advertised VRAM (in this case 16 GB) is available for rendering. 

    It all depends on how the render engine is programmed, I suspect. Iray doesn't, yet, split VRAM loading if it can't fit all the data on one card, but if ProRender can, that changes things.

  • tj_1ca9500btj_1ca9500b Posts: 2,057

    The rumormill has been speculating that NVidia might release another dual GPU later this year, but I haven't seen anything concrete on this as of yet in my tech news reading rounds.  The Titan Z is getting a little long in the tooth.(released in 2014) and it appears that NVidia hasn't released a dual GPU card (for gamers anyways) since the Titan Z was released, at least according to this article (see last paragraph):

    https://arstechnica.com/gadgets/2017/04/amd-puts-two-gpus-and-32gb-of-ram-on-its-latest-radeon-pro-duo-graphics-card/

    There's been a few new standards/features added to the GPU world sine 2014...

     

  • kyoto kidkyoto kid Posts: 42,408

    ...as I understand, there is a 20xx series in the works, which is mostly updating/refining the current 10xx series cards.  Not sure if there will be any memory boosts as to support 16 GB the processor needs to be a GP104 such as used in the standard 1080 and Quadro P5000.  Discussions I read make no mention of HBM2 memory for the 20xx cards so they will probably still be using DDR5/5X.

    Volta isn't due out until late this/early next year and most likely the first release will be a Tesla compute device followed by a Quadro series GPU. 

Sign In or Register to comment.