November 15, 2026 Linh Nguyen

MASH Instancing and Redshift Proxies in Maya: Millions of Objects Without Running Out of VRAM

InstancingHeavy Maya scenes with MASH and Redshift Proxies
From Maxon documentation

How you repeat objects decides GPU memory and scene start-up time. Check these before a heavy scene:

  • MASH: Redshift renders both Instancer and Repro modes; Instancer instances, Repro makes unique geometry
  • Instances: store a transform per copy and use very little render memory
  • Proxies: .rs files load at render time and cost almost nothing during scene translation
  • Many copies: place one proxy and instance it, rather than many separate proxies
  • Limits: proxies cannot be deformed after export; instanced proxies use fixed tessellation

Forests, crowds of props, city blocks and abstract motion graphics all repeat the same objects thousands of times. In Maya you can build them with MASH, plain instances or Redshift Proxies, and the choice decides two things that matter on any render machine: how much GPU memory the scene takes and how long each frame spends before Redshift starts rendering.

This guide explains how Redshift handles MASH Instancer and Repro modes, Maya instances and Redshift Proxies, how to set up a heavy scene step by step, and how to check memory and start-up time before a long batch.

Checked against Maxon’s Redshift for Maya documentation for Maya Particles and FX (MASH), Instances and Redshift Proxies, November 2026.

1. Copies, instances or proxies: what changes in memory?

Maxon describes instances as lightweight copies that keep a link to the original object, store little more than a transform matrix, and use very little memory when rendering with Redshift; in Maya a mesh is instanced when its shape node appears several times in the DAG, for example after Duplicate Special (Maxon: Instances (Maya)). Hair is the other common memory load; see XGen and Yeti hair with Redshift for Maya.

iRender
Copies, Instances and Proxies in GPU Memory
Schematic
MASH Repro / copiesunique geometryGPU memoryevery copy stored as its own meshMASH Instancer / instancesshared geometryGPU memory1 meshtransformsone mesh + a transform per copyInstanced Redshift Proxyshared, stored on diskGPU memory1 proxy(.rs)geometry read from disk at render;not extracted from Maya each frameinstanced proxy: light scene, fast start
unique meshshared meshper-copy transform
Schematic · based on Maxon Instances, Proxies and Maya Particles docs

The three panels compare the options. Copies store every mesh again. Instances share one mesh. An instanced Redshift Proxy shares one mesh too, and its geometry comes from the .rs file at render time; Maxon notes proxies have almost no time cost during scene translation, because the slow geometry extraction is paid once at export (Maxon: Redshift Proxies (Maya)).

2. How Redshift renders MASH

Maxon states that Redshift renders MASH output as a mesh in both Repro and Instancer modes, including Redshift Proxies in Instancer mode. Instancer mode uses Maya’s native instancer and instances the objects; Repro mode creates unique geometry. Translation and object-based deformation motion blur work in both modes, and the only setup needed is a Redshift material on the base mesh (Maxon: Maya Particles and FX).

So the rule of thumb is simple: use Instancer for anything that repeats unchanged, and Repro only when each copy must be its own editable geometry. A thousand identical rocks in Repro mode are a thousand meshes in GPU memory.

iRender
Count instanced and unique meshes
python
1# Script Editor (Python): how many mesh shapes are instanced vs unique?
2import maya.cmds as cmds
3
4shapes = cmds.ls(type="mesh", noIntermediate=True)
5shared = [s for s in shapes if len(cmds.listRelatives(s, allParents=True) or []) > 1]
6print("mesh shapes:", len(shapes))
7print("shapes used by more than one transform (instanced):", len(shared))
A shape with several parents is a Maya instanceMaxon: Instances (Maya)

The script counts mesh shapes and how many of them are shared by more than one transform. A heavy scene with very few shared shapes is a candidate for Instancer mode or proxies.

3. How to set up a heavy Maya scene for Redshift (step by step)

Settings checklist (copy or follow line by line):

Settings checklist
MASH network
Instancer → for repeated objects // instances, low memory
Repro → only when each copy must be unique geometry
Material → Redshift material on the base mesh
Redshift Proxy
Export → final materials assigned; motion blur on at export if it moves
Place → one proxy, then instance it (MASH Instancer or duplicates)
Tessellation → fixed when the proxy is instanced
File → .rs inside the project folder
Check → heaviest frame: geometry memory and extraction time in the log

Step 1: Choose Instancer for repeats

Build the MASH network in Instancer mode and assign the Redshift material to the base mesh.

Step 2: Export heavy assets as proxies

Export detailed assets such as trees or vehicles as Redshift Proxy (.rs) files. Assign the final materials first, and turn on motion blur in the render settings before export if the proxy will move; Maxon notes this is required for proxies to render with motion blur (Maxon: Intro to Proxies (Maya)).

iRender
The Redshift Proxy workflow in Maya
Schematic
1 · Maya assetfinal materials assignedmotion blur on if it moves2 · Export .rsgeometry extracted oncestore it inside the project3 · One proxy nodeinstanced by MASH Instanceror by duplicates4 · Render timegeometry read from .rsalmost no translation costAfter exportthe proxy cannot be deformed or edited, except for tessellation; keep the original Maya asset.Instancedinstanced proxies use fixed tessellation only, not adaptive.Pay the geometry cost once at export instead of at the start of every frameMaxon: Intro to Proxies (Redshift for Maya)
Setup stepLimits to plan for
Schematic, not to scale

Step 3: Instance one proxy

Place one Redshift Proxy and instance it, through MASH Instancer or duplicates; Maxon calls this much more efficient for memory and performance than many separate proxies (Maxon: Intro to Proxies).

Step 4: Keep files with the project

Store .rs files inside the project so the render machine finds them, and keep the original Maya assets, since a proxy cannot be turned back into a Maya mesh.

Step 5: Check the heaviest frame

Render the frame with the most objects and read geometry memory and extraction time in the Redshift log.

4. Proxy limits and common problems

After export, proxy geometry cannot be deformed or modified except for tessellation, and embedded shaders from another application may not render in Maya (Maxon: Intro to Proxies (Maya)). Instanced proxies can only use fixed tessellation, not adaptive (Maxon: Intro to Proxies). Animated proxies are per-frame caches and can be heavy.

TroubleshootSymptom to fix

When heavy Maya scenes misbehave in Redshift

Based on Maxon documentation

SymptomLikely causeFix
Long wait before each frame startsGeometry extracted from Maya every frameMove heavy static assets into proxies
Out of VRAM with many repeated objectsMASH in Repro mode or plain copiesUse Instancer mode or instances
Proxy has no motion blurMotion blur off at exportEnable motion blur, then export again
Displacement looks wrong on instanced proxiesAdaptive tessellation on instancesUse fixed tessellation for instanced proxies
Proxy renders without materials or greyEmbedded shaders from another DCC, or material overrideUse scene materials or re-export from Maya
Proxies missing on the render machine.rs files outside the projectStore .rs files in the project folder

When frames are slow but the GPUs look idle, the time is often spent extracting geometry rather than rendering; our low GPU usage guide shows how to read that in the log. For memory beyond VRAM, see our out-of-core guide.

5. SaaS or IaaS render farm for heavy Maya scenes?

CompareNeutral guide

Which farm type fits a heavy instanced scene

Both farm types render MASH and proxies; the difference is how you check memory and start-up time.

SituationUsually a better fitWhy
Scene already optimized, proxies in the projectSaaSUpload the project and collect frames
Memory and extraction time still being tunedIaaSYou test frames and change MASH or proxy setup on the render machine
Scenes that need 32 GB per GPUEitherChoose RTX 5090 cards wherever you render
Studio pipeline with its own submit toolsEitherPick the farm that supports your submit workflow

Rule of thumb: heavy scenes are won or lost in setup; tune memory on a machine you can open, then render where it suits you.

Rendering a heavy Maya scene on an iRender server

iRender
How iRender works: your computer controls your own multi-GPU server
IaaS render farm
Your PC or MacNo powerful GPU neededRemote desktopLive screen1:1 like your PCYOUR OWN GPU SERVERbare-metal · not sharedGPUGPUGPUGPUGPUGPUGPUGPU1, 2, 4 or 8 × RTX 5090 (32 GB) or RTX 4090 (24 GB)Threadripper PRO · 256 GB ECC RAM · 2 TB NVMe SSDFull admin access · your setup is saved between sessions1Register, top upFirst $50+ top-up doubled2Upload your projectiRender Drive, free3Start and connectRemote desktop, any OS4Render, shut downBilling stops, per minute✓ Your workstation, just more GPUs: no queue to wait, no farm plugin, no job submission, no settings changed
Your mouse and keyboardServer screen streamed back
Schematic, not to scale

An iRender server is a full Windows machine with 32 GB RTX 5090 or 24 GB RTX 4090 GPUs:

  1. Upload the project, including .rs proxy files, to iRender Drive before you start a server; uploading is free and does not use rental time.
  2. Start a server, install your Maya version and the matching Redshift build once; they stay in your image. Setup details are in Redshift for Maya on a multi-GPU render server.
  3. Render the heaviest frame and read geometry memory and extraction time in the Redshift log.
  4. Render the full range, then shut down. Multi-GPU servers bill per minute; 1-GPU servers have a one-hour minimum.

Heavy Maya scene checklist

Tick each item before you start the full range.

ChecklistHeavy Maya scene checklist
0 / 10 done
Tip: tick each item as you set up the job; the ticks reset when the page reloads.

Frequently asked questions

1. Does Redshift render MASH in Maya?

Yes. Maxon states Redshift renders MASH output as a mesh in both Instancer and Repro modes, including Redshift Proxies in Instancer mode, with translation and deformation motion blur supported.

2. Should I use MASH Instancer or Repro with Redshift?

Use Instancer for objects that repeat unchanged; it instances them and uses little memory. Use Repro only when every copy must be its own geometry, since it creates unique meshes.

3. Why use Redshift Proxies in Maya?

Proxies load their geometry from .rs files at render time and cost almost nothing during scene translation, so heavy assets no longer slow the start of every frame.

4. How do I place many copies of a proxy efficiently?

Place one Redshift Proxy and instance it. Maxon notes this is much more efficient for memory and performance than many separate proxies.

5. Why does my proxy have no motion blur?

Motion blur must be enabled in the render settings when the proxy is exported. Enable it and export the proxy again.

6. Can I deform or edit a Redshift Proxy after export?

No. After export the geometry cannot be deformed or modified except for tessellation settings, so keep the original Maya asset.

7. Why does displacement look wrong on instanced proxies?

Instanced proxies can only use fixed tessellation; adaptive tessellation is not available when instancing.

8. Can I render heavy Maya scenes on a multi-GPU iRender server?

Yes. iRender servers have 1, 2, 4 or 8 RTX 5090 (32 GB) or RTX 4090 (24 GB) GPUs; you install Maya and Redshift once and check memory on the heaviest frame before the batch.

Render your Maya and Redshift projects on iRender

iRender is an IaaS GPU render farm: you rent a bare-metal server with 1, 2, 4 or 8 NVIDIA RTX 4090 (24 GB) or RTX 5090 (32 GB) 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. Your server image is saved between sessions, so your plugins and settings stay installed. Maya is not pre-installed: you install the Maya version your scene uses with your own Autodesk license, and it stays in your image. Prime accounts get Redshift licenses pre-activated on every server.

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. Elite support is available 7 days a week by hotline, live chat, Skype, WhatsApp and email.

This quick start shows how to start a server and connect to it:

A full Maya and Redshift session we recorded on an 8x RTX 4090 server:

New users get a 100% bonus on the first top-up (minimum $50). See plans on the Maya render farm for Redshift page or download the iRender GPU app.

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:
, , , , , , ,

Linh Nguyen

Linh Nguyen works in customer support at iRender, helping artists render Blender and Houdini projects on multi-GPU servers. She has written more than 300 articles for the iRender blog, covering Blender, Houdini and practical GPU rendering.
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 →