NVIDIA Iray is a progressive pathracing engine, and uses your computer's graphics card in conjunction with the CPU to accelerate rendering when exporting Podium Walker animations. This approach means we can no longer simply recommend "a system with a fast processor and a lot of RAM" and call it a day.

In order of importance, these are the things you should prioritize when building a system for GPU accelerated rendering with Iray.

  • Supported NVIDIA graphics card with Iray drivers - Not required, but necessary for GPU acceleration. Rendering animations without supported hardware will be slow.
  • CPU speed - Faster clock speed and greater number of cores generlaly yields faster rendering.
  • RAM - Recommend at least 16GB, though less will often suffice.

  Graphics cards with Iray support/drivers:

NVIDIA has finished adding Iray support for most higher-end NVIDIA cards from the past few generations of GTX and QUADRO cards. New/future chip architectures may not have Iray drivers immediately, but the current generation Pascal architecture is supported.

Currently supported NVIDIA GTX series cards:

Pascal: GTX 1060, 1070, 1080, 1080 Ti, TITAN X, plus mobile variants.

Maxwell: GTX 950, 960, 970, 980, 980 Ti, GTX 750, 750 Ti, plus mobile variants.

Kepler/Fermi: GTX 500, 600, 800 series, including mobile.

Currently supported NVIDIA workstation class cards:

Quadro: P6000, M6000, P5000, M5000, M4000, 6000, 5000

Tesla: P100, K80, M40, K40m, K10, C2050, K20c, M2050

Note: These cards are not strictly required to use Podium Walker with Iray, they are simply required for GPU acceleration during rendering. If a supported graphics card is not present, Podium Walker will fall-back and use the CPU to complete pathtracing calculations.

Recommended hardware:

For still visualization: GTX 960 and better (GTX 960, 970, 980, 1060 6GB, 1070, 1080). Our lowest spec recommendation would be the GTX 1050 though you can get away with slower cards if you're patient. SU Podium is a better visualization option for users with a low-end GPU, but fast CPU.

For animation: GTX 970 and better, anything lower will be prohibitively slow in Iray photoreal mode. With the render mode set to Id (Iray interactive), you gain back a significant amount of speed but we still recommend the GTX 970 (or 1060 6GB, which is comparable).

Keep in mind, a 1 minute animation requires approximately 1500 frames for smooth motion. So we can roughly estimate total render times:

  • 3 minutes per frame - 75 hours
  • 1 minute per frame - 25 hours
  • 40 seconds per frame - 16.5 hours
  • 20 seconds per frame - 8.25 hours
  • 10 seconds per frame - 4.25 hours

ProWalker lets you set the number of seconds per frame in the settings dialog, so you can limit the maximum time ProWalker spends on an animation if necessary. However, bear in mind that a 40 second frame will look different from card to card—image quality in ProWalker is dependent on the number of samples Iray is allowed to calculate for that specific frame.

In other words, a 1 minute render from a GTX 1080 will look significantly less noisy (more refined) than 1 minute render on something like a GTX 960 (or lower). So what kind of speed does a reasonably priced, current generation card give us?

Mid-range case study - GTX 1060m | 450 samples | 3 min 33 seconds

Size: 1920x1080 — Render Mode: ID — Render Time: 3min 33sec — 450 samples

So this tells us that the GTX 1060m is definitely powerful enough for still visualization—we have a very appealing image after only 3.5 minutes, and doubling the samples would likely eliminate most of the visible noise.

For animation the results are less promising. We can extrapolate and arrive at a rough estimate: Aaproximately 88 hours for 1500 frames. Potentially doable if you have a dedicated render machine, but for the sake of efficiency we would not recommend a single GTX 1060 for rendering animation with ProWalker.

If you need to render animation on a lower-end card, there are several ways to decrease render time:

  1. Render at a lower resolution.
  2. Switch to Qck Mode instead of Id or Pr.
  3. Decrease the number of samples.
  4. Decrease the number of artificial lights.
  5. Render fewer frames per second (not really recommended), or speed up your animation paths
  6. Consider purchasing Podium Walker standard instead of ProWalker GPU.

Note: If the concept of progressive (samples based) rendering is new or unclear to you, please see here for more information about samples and image quality in Iray.

  Iray Performance Benchmarks:

Migenius Iray benchmarks: Performance benchmarks are occasionally published at Migenius for a selection of cards and cloud-based services. Currently available comparisons from newest to oldest:

Iray's scaling efficiency is very high with NVIDIA CUDA devices. In theory, multiple GPU configurations should yield a near-linear performance increase, with per-card efficiency at approximately 95% of max performance up to 8 cards. In practice, per-card efficiency has been somewhat lower than 95% in our tests, but still quite excellent.