How to build the ULTIMATE Render PC?

Hi, 

We are trying to build the ULTIMATE RENDER PC! (And you might have some ideas how to make it happen.)

Rendering High Quality animations takes a lot of time. Currently we have 2 PC's only rendering animations.

ANIMATION PC CONFIG:
GPU - 4x 1080ti (connected with risers)
RAM - 32 GB
CPU - i7 8700k
PSU: 2x 1000 wat 

The problem is that wen we connect more than 4x 1080ti Daz3D becomes unstable and when rendering it usually crashes or resets.
Has anyone successfully and stably ran Daz3D with 5 or more cards? 
If so what can you please tell us how? And what's your CONFIG? 
Do you know what could be the bottle neck causing the PC with more cards than 4 cards to crash?
Do you have any other suggestion on what we should update for the PC to render faster? (Please don't suggest 2080ti it's not optimized for Daz3D. And dont suggest optimizing the scene. )

Comments

  • Richard HaseltineRichard Haseltine Posts: 108,028

    Is 32GB of RAM enough to feed four 11GB cards?

  • incetongamesincetongames Posts: 1
    edited May 2019

    Yes 32GB is enough since Daz3D only uses the VRAM of one GPU.

    Post edited by incetongames on
  • Richard HaseltineRichard Haseltine Posts: 108,028

    No, Iray uses the memory on each GPU separately.

  • kenshaw011267kenshaw011267 Posts: 3,805

    When loading the scene onto the cards Daz does use lots of RAM per GPU. To add more GPU's you need more RAM.

  • DrNewcensteinDrNewcenstein Posts: 816

    And your motherboard has to have enough PCI lanes to support multiple GPUs. If PCI slots are throttled down as more are populated, you're not getting the full benefit of the GPUs. Your mobo specs regarding that can be found on the manufacturer's site, or in the manual.

    Assuming all the hardware is kosher, then the last step is to throttle back the clock speeds on the GPUs. These are gaming cards, intended to run at a higher rate for the data transfererence between the PC and the game and monitor. Rendering detailed scenes with hi-res meshes and textures and realistic lights requires a slower clock speed. That's one of the major differences between the GTX and their Quadro counterparts - the Quadros run more slowly, to increase stability.

    Apps like MSI Afterburner or EVGA Precision X can adjust clock speeds, though their lowest settings may still be faster than a Quadro.

     

  • kenshaw011267kenshaw011267 Posts: 3,805

    And your motherboard has to have enough PCI lanes to support multiple GPUs. If PCI slots are throttled down as more are populated, you're not getting the full benefit of the GPUs. Your mobo specs regarding that can be found on the manufacturer's site, or in the manual.

    Assuming all the hardware is kosher, then the last step is to throttle back the clock speeds on the GPUs. These are gaming cards, intended to run at a higher rate for the data transfererence between the PC and the game and monitor. Rendering detailed scenes with hi-res meshes and textures and realistic lights requires a slower clock speed. That's one of the major differences between the GTX and their Quadro counterparts - the Quadros run more slowly, to increase stability.

    Apps like MSI Afterburner or EVGA Precision X can adjust clock speeds, though their lowest settings may still be faster than a Quadro.

     

    Not true at all.

    Graphic cards can use 1 PCIE lane. It just means data is transfered to and from them more slowly, miners did this quite frequently.

    You also don't need lower clock speeds for more stable renders. 

  • PsyckosamaPsyckosama Posts: 508

    Wait for Ryzen 3. Don't go Intel.

  • Richard HaseltineRichard Haseltine Posts: 108,028

    How are the Ryzens for single core/thread performance? There are several CPU-only processes in DS or plug-ins that are single-threaded, so that is potentially quite important.

  • kenshaw011267kenshaw011267 Posts: 3,805

    How are the Ryzens for single core/thread performance? There are several CPU-only processes in DS or plug-ins that are single-threaded, so that is potentially quite important.

    Ryzen 2000's were slower than Intel CPU's of equivalent vintage, before the latest round of CPU vulnerabilities became known. When those mitigations are deployed who knows?

    Ryzen 3000 are expected to be officially announced Monday in Taipei but won't be in the hands of independent reviewers for a month or so. Expectations are that they will be roughly on par with Intel CPU's.

Sign In or Register to comment.