V-Ray Render Time Optimization: Sampling, GI Settings, and Denoiser Workflow

V-Ray Render Time Optimization: Sampling, GI Settings, and Denoiser Workflow

I do a lot of render optimization for studios that come to me with complaints about 4-hour render times per frame. After reviewing their settings, I can usually cut that to 1-1.5 hours without any visible quality difference. The problem is almost never hardware — it's that artists are using brute-force settings where smarter approaches would work just as well.

Principle 1: Don't Brute-Force Noise — Denoise It

The single biggest render time waste I see is artists setting Noise Threshold to 0.001 and waiting hours for the renderer to clean up every last pixel of noise. V-Ray 6's denoiser is good enough that you can render at a higher Noise Threshold and let the denoiser handle the rest.

My standard approach:

• Noise Threshold: 0.01 (default) instead of 0.003
• Add VRayDenoiser render element with Strength 0.7
• For RTX cards: use NVIDIA AI Denoiser for even better results

This alone cuts render time by 40-60%. The denoised image is visually indistinguishable from a non-denoised render at 0.003 threshold in most cases. I always do an A/B comparison with the client before committing to this approach, and they've never been able to tell the difference.

Principle 2: Optimize GI Settings

Global Illumination is the second biggest time sink. The default GI settings in V-Ray 6 are good, but they can be optimized based on scene type.

Interior Scenes

• Primary GI: Irradiance Map (not Brute Force)
• Secondary GI: Light Cache
• Irradiance Map Preset: Medium (not High — the difference is invisible in most cases)
• Light Cache Subdivs: 2000 (not 1000 — this is worth the small time cost)
• Light Cache Sample Size: 0.02 (slightly larger samples = faster, slightly softer shadows)

I see many artists using Brute Force for both primary and secondary GI on interiors. This is the most expensive combination. Brute Force is accurate but slow — Irradiance Map + Light Cache produces virtually identical results in a fraction of the time.

Exterior Scenes

• Primary GI: Brute Force (exteriors have simpler GI, so BF is fast enough)
• Secondary GI: Light Cache
• Light Cache Subdivs: 1000 (exteriors need less GI bouncing)

Product/Studio Renders

• Primary GI: Irradiance Map
• Secondary GI: Brute Force (for accurate reflections in studio lighting)
• Irradiance Map Preset: High (product renders need clean reflections)

Principle 3: Limit GI Bounces

Every GI bounce adds exponential render time. V-Ray defaults to 3 bounces for diffuse, which is fine for most scenes. But I've seen artists set it to 10+ "just to be safe."

My bounce limits:

• Diffuse bounces: 3 (interiors), 2 (exteriors)
• Reflection bounces: 2 (most scenes), 3 (glass-heavy scenes)
• Refraction bounces: 3 (glass), 2 (most scenes)
• Total bounces: 5-6 max

Going from 10 diffuse bounces to 3 cuts render time by about 40% with no visible difference. After 3 bounces, the light is so diffused that the contribution is negligible.

Principle 4: Use Render Elements for Compositing

Instead of getting everything perfect in the render, I render with slightly lower quality settings and fix issues in compositing. This requires rendering with the right render elements:

• VRayReflections: Lets me boost or reduce reflections in post
• VRayRefraction: For adjusting glass transparency
• VRayGI: For adjusting indirect lighting intensity
• VRayLighting: For adjusting direct lighting
• VRaySpecular: For adjusting specular highlights
• VRayDenoiser: For noise cleanup
• VRayCryptomatte: For per-object adjustments

With these elements, I can fix lighting, reflection, and noise issues in Nuke or Photoshop in minutes — changes that would take hours to re-render.

Principle 5: Optimize Materials

Material settings have a significant impact on render time. The most expensive material properties are:

Reflection Glossiness

Every subdivision in reflection glossiness doubles the samples needed for that material. A material with 32 subdivs is 4x more expensive than one with 8 subdivs.

My approach: Use 8 subdivs for most materials and enable Use Interpolation. Interpolation caches glossy reflection results, dramatically reducing render time with minimal quality loss. I only increase subdivs for hero materials that the camera is close to.

Subsurface Scattering

SSS is extremely expensive. I've seen a single SSS material add 30 minutes to a render.

My approach: Use V-Ray Fast SSS2 instead of the standard SSS material. It's optimized for speed and produces nearly identical results. Also, set the Scale parameter correctly — an incorrect scale value can cause the SSS to calculate unnecessarily deep scattering.

Displacement

Displacement is the most expensive geometry operation in V-Ray. Every displaced triangle is subdivided at render time.

My approach:

• Use 2D displacement (landscape mode) instead of 3D when possible — it's faster
• Set Edge Length to 4-6 pixels (lower = more subdivision = slower)
• Limit displacement to objects that need it — don't apply it globally

Principle 6: Proxy Geometry for Heavy Scenes

I covered this in my 3ds Max optimization guide, but it applies to V-Ray specifically: use VRayProxy objects for any geometry over 50,000 polygons. Proxies load at render time only, keeping the scene file small and the viewport fast.

For vegetation specifically, I use Forest Pack with VRayProxy trees. A forest of 10,000 trees renders in the same time as 10 trees if they're all proxies of the same source mesh.

Principle 7: Render Region and Test Settings

For testing, I never render the full frame. I use Render Region (the region render button in the render frame window) to render a small area that's representative of the scene's complexity. This lets me iterate on settings in seconds instead of minutes.

My test workflow:

1. Set render resolution to 50% of final
2. Render a region that includes the most complex area (e.g., a corner with glass, reflections, and GI bouncing)
3. Check the VRaySampleRate render element — red areas indicate where the sampler is struggling
4. Adjust settings based on the Sample Rate feedback
5. Once the region looks good, render the full frame at 100% resolution

The VRaySampleRate render element is my secret weapon. It shows a heat map of where the renderer spent the most samples. Large red areas mean the renderer is struggling — usually with glossy reflections, SSS, or high GI bounces. I optimize the materials or settings causing the red areas, then re-test.

Principle 8: Bucket Size and Sequence

On CPU renders, Bucket Size affects performance. I use 32x32 for most scenes (the default is 24x24). Larger buckets process more pixels at once, reducing overhead.

Bucket Sequence: I use Top to Bottom for most scenes. If I'm testing and want to see a specific area first, I switch to Hilbert or set a custom bucket order.

Principle 9: GPU vs CPU Selection

V-Ray GPU is faster for most scenes, but not all. I've found:

• GPU is faster: Scenes with many lights, simple materials, high sample counts
• CPU is faster: Scenes with complex SSS, displacement-heavy geometry, or when using features not yet supported on GPU

I test both on a region render and use whichever is faster. For our studio's hardware (RTX 4090 + Ryzen 9 5950X), GPU wins about 80% of the time.

Real-World Example

A client sent me a scene that was taking 3.5 hours per frame on a 16-core Xeon with 2x RTX 3090. Here's what I changed:

1. Noise Threshold: 0.003 → 0.01 + Denoiser (saved 50%)
2. GI: Brute Force + Brute Force → Irradiance Map + Light Cache (saved 30%)
3. Diffuse bounces: 8 → 3 (saved 15%)
4. Reflection subdivs: 32 → 8 + Interpolation (saved 20%)
5. Displacement Edge Length: 2 → 4 (saved 10%)

Final render time: 45 minutes per frame. Total savings: 78%. The client couldn't see any difference in the final output.

Summary

V-Ray render optimization is about working smarter, not harder. My optimization order: enable Denoiser and raise Noise Threshold → switch to Irradiance Map + Light Cache for interiors → limit GI bounces to 3 → reduce reflection subdivs and use interpolation → use VRaySampleRate to identify problem areas → use render region for testing. This sequence typically cuts render times by 50-70%.