November 17, 2026 Maddie Quach

Octane Kernels and Sampling in Cinema 4D: Path Tracing, Photon Tracing and Fireflies

KernelsWhich Octane kernel and sampling settings to start with
From OTOY documentation

OTOY’s kernel guide follows one order: kernel, depths and clamp, then sampling. The short version:

  • Path Tracing: the default for realistic renders without important caustics
  • Photon Tracing: when caustics matter; test it on multi-GPU and network renders
  • Max Samples: start at 64–256 and let adaptive sampling and the denoiser do the rest
  • GI clamp: lower it only after fixing lights; it removes energy as well as fireflies
  • Animation: keep Coherent Ratio at 0 and avoid PMC for sequences

Most slow or noisy Octane renders in Cinema 4D come from kernel settings, not from the GPU. A scene can render for twenty minutes at 16,000 max samples when 256 samples, adaptive sampling and a denoiser would give the same picture in a fraction of the time. On a render farm the cost of a bad setting multiplies by every frame.

This guide follows the order OTOY recommends in its kernel guide: choose the kernel, set depths and GI clamp, then tune sampling. Every value below is from OTOY’s documentation; treat them as starting points to test on your own frames.

Checked against the OctaneRender for Cinema 4D manual (Kernels, Path Tracing, Photon Tracing) and OTOY’s Help Center kernel guide, November 2026.

1. Which Octane kernel should you use?

Octane has four main render kernels: Direct Light, Path Tracing, PMC and Photon Tracing. OTOY’s guide places them by what the shot needs. Direct Light is the fast option for simple scenes without caustics or heavy scattering. Path Tracing is the general choice for realistic images when caustics are not a key feature. Photon Tracing is built for caustics: OTOY’s C4D manual says it renders caustics much faster and with less noise than PMC, using GPU photon mapping and path guiding. PMC handles very complex lighting and refraction but is much slower and less predictable from frame to frame.

iRender
Choosing an Octane Kernel
Schematic
What does the shot need?Simple light, no causticsDirect Lightfast; max samples 128 defaultRealistic GI, minor causticsPath Tracingdefault choice; 16,000 maxCaustics matterPhoton Tracingenable caustics per materialVery complex refractionPMCslowest; weak for animationDo look development in Path Tracing or Photon Tracing first; test the kernel on a few frames before an animation
usual defaultcausticsspecial cases
Source: OTOY Help Center, kernel guide

Two notes matter for farm rendering. OTOY’s guide warns that Photon Tracing adds overhead and can cause problems in multi-GPU or network rendering, so test it on the actual multi-GPU machine before committing a sequence. And because PMC can change from frame to frame, OTOY calls it a poor fit for animation; use it for stills.

For caustics with Photon Tracing, OTOY notes that you also enable caustics in the IOR channel of the reflective and refractive materials that should create them. Our guide on Spectron lights and photon tracing caustics covers that setup.

2. Depths, GI clamp and the settings that cost the most time

Depth values set how many times a ray may bounce. OTOY’s guide notes that the Path Tracing defaults of 16 diffuse and 16 specular bounces are a good start, while Direct Light’s default glossy depth of 2 is usually too low. Lower depths step by step until the image starts to change, then go back one step.

SettingsStarting values

Octane kernel settings that change render time the most

Starting points from OTOY’s kernel guide; tune per scene.

SettingStarting valueWhat it trades
Max Samples64–256 for Path or Photon Tracing; 16–64 for Direct LightNoise against render time; keep low and let the denoiser finish
Diffuse / Specular depth16 for Path Tracing (default); 5–8 glossy for Direct LightBounces of light; lower until the look changes
GI clamp1,000,000 (default) to startFireflies against light energy
Parallel Samples32; 16 on older cards or heavy scenesSpeed against VRAM use
Coherent Ratio0 for animationSpeed against blotchy, flickering dark areas
Adaptive Samplingon; noise threshold 0.03 default, 0.02 cleanerTime spent on clean pixels
Workflow:set kernel, depths and clamp first, then sampling and the denoiser last (OTOY).

Parallel Samples affects both speed and memory. OTOY suggests starting at 32 and dropping to 16 on older cards or memory-heavy scenes, and checking VRAM in the Live Viewer readout. On a 32 GB card there is usually room for the higher value; on a scene close to the VRAM limit, the lower value can be the difference between rendering in-core and running out of memory.

GI clamp and fireflies

Fireflies are single bright pixels from rare light paths that carry a lot of energy, often through glass or small bright lights. GI clamp caps the contribution of each path. OTOY’s Path Tracing page explains the trade: reducing the clamp removes fireflies by removing energy, so the image can get darker and lose some indirect light.

iRender
What GI Clamp Does to Fireflies
Schematic
Brightness carried by individual light pathsGI clamprare, very bright paths = firefliesLowering GI clamp cuts the bright spikes, so fireflies go away, but it also removes light energy. Fix lights and caustics first.
normal pathsfirefly pathsclamp value
Source: OTOY, Path Tracing kernel

OTOY’s guide therefore keeps the clamp high (1,000,000) at the start and lowers it only after lights and caustics are fixed in other ways. Simple scenes can use values of 100 or 10. If caustics cause the fireflies and you need them, Photon Tracing is the better fix than a low clamp.

3. Adaptive sampling: stop spending samples on clean pixels

With adaptive sampling on, Octane stops sampling pixels that are already clean and keeps working on noisy ones. OTOY’s guide recommends keeping it on. The noise threshold sets what counts as clean: 0.02 is clean, 0.03 is the default and a little faster, and 0.04 or higher is faster but noisier. Minimum samples defaults to 256, so every pixel gets at least that many before adaptive sampling decides.

iRender
How Adaptive Sampling Spends Samples
Schematic
Samples per pixel as the render runsmaxmin 256flat wallnoise below thresholdstops near the minimumsoft shadowkeeps samplinguntil it is cleanglass and causticsnever clean enoughstops at max samplesNoise threshold 0.02 is clean, 0.03 is the default, 0.04+ is faster but noisier (OTOY)
stops earlyruns to max samples
Source: OTOY Help Center, kernel guide

Expected exposure tells adaptive sampling how bright the final image will be, so it judges noise in dark areas correctly. OTOY’s guide says it starts at 0 and should match your Imager exposure if you changed it. Pixel grouping defaults to 2×2.

Adaptive sampling changes where time goes, not the maximum. Areas such as glass and caustics can still run to Max Samples, so keep Max Samples reasonable rather than at 16,000.

4. Settings for animation and multi-GPU renders

A still image can tolerate settings that break an animation. OTOY’s guide lists three to watch for sequences:

  • Coherent Ratio: speeds up rendering but causes blotchy or noisy dark areas; keep it at 0 for animations.
  • Denoiser: keep it off during look development and turn it on near the end, then lower Max Samples step by step. Test a few frames or a render region before the full sequence; see our guide to denoising Octane animations without boiling.
  • PMC: avoid it for sequences because results vary from frame to frame.

Kernel settings are identical on one GPU or eight: more GPUs render the same samples faster. That means you can tune samples on your workstation and expect the same picture on a multi-GPU machine, with Photon Tracing as the one kernel worth a separate test, as noted above.

5. Compare two settings on the same frames, in the background

Tuning by eye in the Live Viewer works for one frame. For a sequence, the fair test is the same short range rendered twice, once with each version of the settings, then compared side by side for noise, flicker and time. Save the scene twice (for example shot_A.c4d with your current samples and shot_B.c4d with the change), turn on Stamp Info so every frame records its samples and render time, and let the Commandline renderer do the rest while you keep working:

  • Maxon’s -frame takes a start, an end and a step: 140 144 1 renders five consecutive frames, enough to see boiling between frames;
  • g_logfile writes each run’s console output to its own text file;
  • OTOY’s example adds g_modulePath so the Octane plugin loads in a background render.
iRender
Render the same test frames with two Octane settings
cmd
1@echo off
2:: test_frames.bat - 5 frames in a row at the hardest point, once per settings version
3set C4D="C:\Program Files\Maxon Cinema 4D 2026\Commandline.exe"
4set MOD=g_modulePath="%%{g_startupPath}/corelibs;%%{g_startupPath}/plugins"
5
6for %%S in (shot_A shot_B) do (
7 %C4D% %MOD% -nogui -render "D:\proj\%%S.c4d" -frame 140 144 1 g_logfile="D:\proj\logs\%%S.txt"
8)
Pick the frame range where the shot is hardest: glass, caustics, fast motion-frame, -nogui, g_logfile: Maxon · g_modulePath: OTOY

Give the two scenes different output paths so the runs do not overwrite each other, then flip between the two sets of frames. Keep the change that removes the noise you can see at the lowest time on the stamp. OTOY notes that background rendering needs a valid command line license.

Sources: Maxon: Cinema 4D and the Command Line, OTOY: Command Line Rendering.

6. SaaS or IaaS render farm for tuning Octane samples?

CompareNeutral guide

Which farm type fits an Octane sampling workflow

Both render the same kernel; the difference is whether you can adjust it on the machine.

SituationUsually a better fitWhy
Samples and denoiser already tested on framesSaaS (managed farm)Submit and render
Still tuning Max Samples or GI clampIaaS (remote workstation)Change settings in the Live Viewer and re-render a frame
Photon Tracing on several GPUsIaaSTest the kernel on the same multi-GPU machine first
Very long sequence at final qualityEither, after a test batchMeasure time per frame before the full run
Short one-off stillSaaSNo setup time

Rule of thumb: tune kernel and samples where you can see the Live Viewer. Once the settings are fixed and tested on a few frames, any farm type will render them the same way.

Tuning Octane samples on an iRender server

iRender
Works 1:1 like your own PC: we keep your setup, you just get more GPUs
IaaS render farm
Your workstationscene.c4d1 GPUYour iRender serverscene.c4dup to 8 GPUs=same setupWindows or Ubuntu desktopYour software, in the version you chooseYour plugins and scriptsYour licensesYour scene and render settings, unchanged1 GPUGPU power: the only difference1, 2, 4 or 8 × RTX 5090 / 4090✓ Only the GPU count changes: open your scene and render the way you do at home
Same on both machinesTip: upload the full project folder with relative paths so textures are found

On iRender you rent a whole Windows workstation with Cinema 4D and OctaneRender installed and sign in with your own OTOY account, so you tune in the Live Viewer exactly as at home:

  1. Upload the project through iRender Drive and start a 2x, 4x or 8x RTX 5090 server.
  2. Open the Live Viewer and render the hardest frame with your kernel, depths and clamp.
  3. Turn on adaptive sampling and the denoiser, then lower Max Samples until noise appears and go back one step.
  4. Render three to five frames in a row and check for flicker before the full range.
  5. Write down the time per frame and multiply by the frame count to estimate the job.

Octane kernel checklist

Use this as a final check before you start the full range on any machine.

ChecklistOctane kernel checklist
0 / 6 done
Tip: tick each item as you set up the job; the ticks reset when the page reloads.

Frequently asked questions

1. Which Octane kernel is fastest?

Direct Light is the fastest, but it is meant for simple scenes without caustics or heavy scattering. Path Tracing is the usual choice for realistic renders, and Photon Tracing is faster than PMC for caustics.

2. What is the difference between Path Tracing and Photon Tracing in Octane?

Path Tracing is the general kernel for realistic lighting. Photon Tracing adds GPU photon mapping and path guiding so caustics render faster and with less noise. OTOY notes it adds overhead and can cause problems in multi-GPU or network rendering.

3. How do I get rid of fireflies in Octane?

Fix the cause first: very small bright lights, glass and caustics. Then lower GI clamp step by step. A lower clamp removes energy as well as fireflies, so the image can get darker.

4. What noise threshold should I use for adaptive sampling?

OTOY’s guide gives 0.02 as clean, 0.03 as the default and slightly faster, and 0.04 or higher as faster but noisier. Minimum samples defaults to 256.

5. How many max samples does Octane need?

OTOY’s guide starts Path Tracing and Photon Tracing at 64–256 samples and Direct Light at 16–64. With adaptive sampling and the denoiser, most scenes do not need the 16,000 default.

6. Is PMC good for animation?

OTOY calls it a poor fit: it is much slower and less predictable from frame to frame. Use it for stills with complex refraction and use Path or Photon Tracing for sequences.

7. Do Octane kernel settings change on a multi-GPU machine?

No. The same settings give the same image; more GPUs render the samples faster. Test Photon Tracing on the target machine because OTOY warns it can cause problems in multi-GPU or network renders.

Render your Cinema 4D + Octane scenes on iRender

iRender is an IaaS GPU render farm: you rent a bare-metal server with 1, 2, 4 or 8 NVIDIA RTX 4090 or RTX 5090 GPUs, an AMD Ryzen™ Threadripper™ PRO CPU, 256 GB ECC RAM and a 2 TB NVMe SSD, connect by remote desktop and work on it like your own workstation, with full administrator access. Cinema 4D and OctaneRender are installed; sign in with your own OTOY account (OTOY’s Studio+ plan has no GPU limit for multi-GPU servers). Your server image is saved between sessions.

Multi-GPU servers are billed per minute (1-GPU servers have a one-hour minimum), and sessions of 3 hours or more get 10% off. Our 4-GPU and 8-GPU servers are liquid-cooled; in our own measurements the GPUs hold 60–70°C under sustained render load, so they keep their boost clocks. Elite support is available 7 days a week by hotline, live chat, Skype, WhatsApp and email.

This quick start shows how to start Cinema 4D and Octane on an iRender server:

On a 4x RTX 5090 server, a 300-frame Cinema 4D + Octane animation rendered in 1 hour 18 minutes. The second clip runs the same kind of scene on 4x RTX 5090 and 4x RTX 4090:

New users get a 100% bonus on the first top-up (minimum $50). See plans on the Octane render farm page, download the iRender GPU app, or read the Octane AI Denoiser for animation.

Accelerate your production pipeline with high-performance IaaS render farm services and dedicated bare-metal workstations powered by up to 8x NVIDIA GeForce RTX 5090 (32GB GDDR7), high-frequency AMD Ryzen™ Threadripper™ PRO processors, 256GB ECC RAM, and dedicated 2TB NVMe SSD storage tailored for your pipeline:

Maddie Quach

Maddie Quach, from iRender's customer support team, has written more than 100 articles for the iRender blog on Maya, Houdini, Arnold, Karma XPU and Redshift.
Contact

iRENDER GPU

Render Farm Pricing
Quick Start
Multi-GPU Benchmarks
iRender GPU App
Blog
Documentation
The Render Farm
The Company
Terms & Conditions

INTEGRATIONS

Autodesk Maya
Blender
Cinema 4D
Houdini
Karma XPU
Lumion
KeyShot
Unreal Engine
Twinmotion
Redshift
Octane
And many more…

iRENDER TEAM

Elite support 7 days a week
SATURDAY – SUNDAY: 6:00 AM – 11:59 PM
(UTC+7)
Hotline: (+84) 912-785-500
Skype: iRender Support
Email: [email protected]
Address 1: 68 Circular Road #02-01, 049422, Singapore.
Address 2: No.22 Thanh Cong Street, Hanoi, Vietnam.

Contact
이 페이지는 한국어로도 제공됩니다. 한국어로 보기 →