Redshift out of core rendering lets a scene exceed VRAM, but the documentation attaches memory limits and caveats you should know before relying on it.
Pipeline Brief // Redshift Memory & Out-of-Core
Redshift out of core rendering: a documentation-based guide to how Redshift streams textures and geometry when a scene is bigger than VRAM, and which memory settings the manual says to leave alone.
- Streaming, Not Magic: Redshift supports out-of-core texturing and geometry. Only the needed parts of textures are uploaded, and distant textures use lower-resolution MIP maps.
- Hard Floor on Free VRAM: even with out-of-core, Redshift aborts with messages such as “cannot operate with less than 256MB of free VRAM,” mostly on 4 to 6 GB cards.
- Defaults Are Deliberate: Redshift reserves 90% of free GPU memory, and Maxon calls the memory page “only for advanced users,” warning that wrong settings can cause poor performance or crashes.
- Practical Production Fix: keep Automatic Memory Management on, then watch the texture cache numbers in the Feedback Display. For larger cards see the Cinema 4D cloud rendering guide. Change one memory setting at a time.
Redshift out of core rendering is what lets a scene use more textures and geometry than the GPU can hold, and it is also where most memory confusion starts. Maxon documents how it works, which settings control it and what to avoid. This guide summarizes that documentation. The Cinema 4D edition of the Memory page did not load for this article, so figures come from the Redshift for Maya edition; confirm menu names in your host application.
Contents
What is Redshift out of core rendering?
Redshift out of core rendering means textures and geometry that do not fit in VRAM are handled from system memory. Maxon documents out-of-core texturing and out-of-core geometry, with texture streaming that uploads only what is needed and lower-resolution MIP maps for distant textures, and recycled geometry memory and texture cache so scenes with gigabytes of textures can render.
Source: Redshift memory documentation (Redshift for Maya edition).
How does Redshift use GPU memory by default?
By default Redshift reserves 90% of free GPU memory, leaving 10% for other apps and the operating system. Maxon suggests lowering it if other GPU-heavy applications have problems, and says raising it to 100% is only for machines where no other app uses the GPU. Going above 90% is discouraged and may cause instability or driver crashes.
- Inactivity timeout: memory is released after 10 seconds without rendering activity.
- Automatic Memory Management: recommended, and should stay enabled unless advanced users see wrong decisions.
- Warning: the page states that incorrect settings can cause poor rendering performance and crashes.
Which Redshift memory values matter for out-of-core scenes?
These are the values the page lists. Treat them as reference points, not tuning targets.
| Setting | Documented value | What Maxon says |
|---|---|---|
| GPU memory percentage | 90% default | Lower to 80% or 70% for other GPU apps; above 90% not recommended |
| Texture cache percentage | 15% usual | Capped by the maximum texture cache size |
| Maximum texture cache size | 128 MB default | If the Uploaded figure climbs into hundreds of MB or GB, try 256 MB first |
| Maximum CPU texture cache size | 4 GB default | Raise to 8 GB or more for texture-heavy scenes with performance issues |
| Ray reserved memory | 0 default | Redshift picks a size from shader configuration |
Geometry memory is roughly 1 million triangles per 60 MB by the same page, so a 20 million triangle scene uses about 1.2 GB for geometry (our arithmetic from that ratio, not a measurement).
What are the limits of Redshift out of core rendering?
Redshift still needs a minimum amount of free VRAM. Maxon documents errors such as “Redshift cannot operate with less than 256MB of free VRAM,” which mostly appear on GPUs with 4 to 6 GB of VRAM or when other GPU apps are running. Cards with 11 GB or more are far less likely to hit them.
- Irradiance buffers: point cloud and cache working memory default to 128 MB each, and the manual says to increase them to 256 or 512 MB if a computation fails. Messages such as “doesn’t fit in VRAM. Frame aborted” point to these budgets.
- Other VRAM users: SLI mode, Windows Aero, Chrome and apps such as Photoshop can take memory away from the render.
Source: Dealing with Out-Of-VRAM situations.
Five settings to check when a Redshift scene is bigger than VRAM
Run these in order on your heaviest frame. The order is our recommendation; the behavior is from Maxon.
- Free the GPU. Close Chrome and other GPU apps; disable SLI.
- Keep the 90% default. Lower it if other apps need the card; avoid going above 90%.
- Watch the texture cache. If the Uploaded figure grows very fast, raise the maximum cache to 256 MB, then the percentage.
- Check irradiance budgets. If the error mentions the frame not fitting in VRAM, raise the working memory to 256 or 512 MB.
- Leave Automatic Memory Management on. Change one memory value at a time and re-render the same frame.
Related reading: Octane GPU memory error VRAM fixes, Redshift vs Octane animation stability and motion blur wrong render engine fix.
Conclusion
Redshift out of core rendering is a safety net with a floor: it needs free VRAM, sensible cache sizes and untouched defaults. Check the heaviest frame, adjust one value at a time, and move to a larger card when 4 to 6 GB is the limit. For cards with more memory, see the top render farms of 2026.
In short, redshift out of core rendering comes down to the documented limits and one change at a time. Test redshift out of core rendering on a short frame range first, and note what changed whenever redshift out of core rendering behaves differently on the full sequence.
FAQ
Does Redshift support out-of-core rendering?
Yes. Maxon documents out-of-core texturing and out-of-core geometry. Textures are streamed so only needed parts are uploaded, and geometry memory and texture cache are recycled.
What is the default GPU memory percentage in Redshift?
Redshift reserves 90% of free GPU memory by default. Maxon says going above 90% is discouraged by Maxon and may cause system instability or driver crashes.
Why does Redshift say it cannot operate with less than 256MB of free VRAM?
Redshift supports out-of-core rendering but still needs a minimum of free VRAM. Maxon says the error mostly occurs on 4 to 6 GB GPUs or when other GPU applications are running.
How much memory does geometry use in Redshift?
Maxon gives a ratio of roughly 1 million triangles per 60 MB of geometry memory. A 20 million triangle scene would therefore use about 1.2 GB, as an estimate.
COMMENTS