Redshift vs Octane on RTX 5090: RT Cores, VRAM and Multi-GPU Settings (2026)
Both engines use the RTX 5090's RT cores, 32 GB of VRAM and every GPU in the machine, but through different settings. Redshift has a Hardware Ray-Tracing switch and a memory share; Octane has Use RTX acceleration, an out-of-core RAM limit and a license that decides how many GPUs it may use.
- RT cores: speed up ray traversal only; gains depend on how much of the frame is ray tracing.
- VRAM: each GPU holds its own copy of the scene; 32 GB per card is the per-frame limit.
- Multi-GPU: Redshift renders one bucket per GPU; Octane Studio+ has no GPU limit per machine.
- Compare: same scene, same server, same time budget; our 4x RTX 5090 test is below.
On paper the RTX 5090 is the same card for every renderer: 21,760 CUDA cores, RT cores for ray tracing and 32 GB of GDDR7. How much of that a render actually uses depends on settings that Redshift and Octane expose in different places, and on what kind of work the scene spends its time on.
This guide shows where each engine turns on RT core acceleration, how each handles VRAM and out-of-core memory, how each uses several GPUs, and how to compare them fairly on one server. It is based on Maxon and OTOY documentation and one recorded iRender test; it does not quote speed-up percentages, because those depend on the scene. For the difference in how the two engines estimate light, see Biased vs Unbiased Rendering.
Checked against Maxon Redshift and OTOY OctaneRender for Cinema 4D documentation and NVIDIA RTX 5090 specifications, November 2026.
1. How do Redshift and Octane use the RTX 5090's RT cores?
RT cores accelerate one job: finding which triangle each ray hits. In Redshift it is the Hardware Ray-Tracing option in the Globals settings, which uses RT hardware on cards such as the RTX series. Maxon notes that non-ray-traced effects such as Irradiance Point Cloud and Irradiance Cache GI are not accelerated by hardware RT, and that the impact varies from scene to scene, with scenes that spend more time on ray tracing gaining more.
In Octane for Cinema 4D the switch is Use RTX acceleration in the Device Settings dialog, and it is on by default. OTOY says Octane does not require RTX but renders some scenes much faster when RT core hardware is present.
So the same GPU upgrade helps a glass-and-chrome product shot with many bounces more than a scene dominated by shading, point-based GI or CPU-side scene export. Leave the RT option on in both engines for final renders; turn it off only to test whether a specific problem comes from it.
2. How do they use 32 GB of VRAM, and what happens when it runs out?
Each RTX 5090 has 32 GB of GDDR7, according to NVIDIA. Neither engine pools memory across GPUs: OTOY puts it plainly, VRAM is not compounded when you add cards, and Redshift also loads the scene on every GPU. With four cards, each frame must still fit in 32 GB.
When a scene does not fit, both engines can move data to system RAM, at a speed cost:
- Redshift reserves a share of free VRAM per frame (90% by default, per the Memory documentation) and splits it into geometry memory and a texture cache. Geometry and textures that do not fit go out of core.
- Octane uses out-of-core only when Out of Core is enabled, with a RAM usage limit you set. OTOY notes that out-of-core data must sit in non-swappable memory, so leave enough RAM for the system.
A 32 GB card moves the point where paging starts compared with a 24 GB RTX 4090. If a scene pages heavily even on 32 GB, trim textures or geometry before adding more of the same GPU.
3. How do they split work across several GPUs?
How each engine uses a 2x, 4x or 8x RTX 5090 server
From Maxon and OTOY documentation.
| Redshift | Octane | |
|---|---|---|
| How a frame is split | Buckets; each bucket is rendered by a single GPU | The frame is rendered on all enabled GPUs |
| Where GPUs are chosen | Redshift preferences, GPU devices | Octane Settings, Devices tab |
| RT core switch | Globals: Hardware Ray-Tracing | Device Settings: Use RTX acceleration (default on) |
| Out-of-core control | Memory share of free VRAM, automatic | Out of Core tab: Enabled, RAM usage limit |
| License and GPUs | Maxon license per machine | Studio+: no limit on GPUs per machine |
| VRAM across GPUs | Not pooled | Not pooled |
In Redshift’s bucket mode, each bucket is rendered by a single GPU, so a 4x or 8x server renders more buckets at the same time. Bucket size can matter for GPU performance; Redshift uses 128 × 128 buckets by default. Octane renders each frame on every GPU enabled in the Devices tab, and the Studio+ subscription places no limit on the number of GPUs. On any machine, disable a GPU that also drives a monitor if it slows interactive work.
4. How to compare Redshift and Octane fairly on one server
A fair comparison needs the same scene, the same server and a goal both engines must meet, such as the same noise level at the same resolution. Two tests from iRender’s channel show the method on a 4x RTX 5090 server, each scene rendered in both Octane and Redshift: a short car scene, where the two runs finish in 18 and 28 minutes, and a long sailing-boat scene, where they finish in 2 h 01 min and 3 h 19 min. Comparing a short and a long render shows whether the gap holds as the job grows. The numbers belong to those scenes and settings, so repeat the method on your own shot.
Common comparison mistakes:
- Different noise targets: a faster engine at a higher noise level has not won. Match the look first, then time it.
- RT off in one engine: check Hardware Ray-Tracing in Redshift and Use RTX acceleration in Octane before timing.
- First-frame only: the first frame includes scene export and loading. Time a short sequence, not a single frame.
- Paging in one engine: if one engine goes out of core and the other does not, you are measuring memory, not the renderer.
- Mismatched denoising: denoise both or neither.
A test workflow on an iRender server
- Upload the project to iRender Drive before you start the server; uploading is free and does not use rental time.
- Start a server with 2, 4 or 8 RTX 5090 (32 GB) GPUs. Cinema 4D, Redshift and Octane are pre-installed; Octane runs with your own OTOY license.
- Confirm RT and devices: Hardware Ray-Tracing on in Redshift, Use RTX acceleration on and all GPUs enabled in Octane.
- Render a short sequence in each engine to the same noise target and compare time per frame.
- Render the job in the faster engine for that scene, then shut down; billing is per minute.
Frequently asked questions
1. Does Redshift use the RT cores on the RTX 5090?
Yes, when Hardware Ray-Tracing is enabled in the Globals settings. Maxon notes that point-based GI such as Irradiance Point Cloud and Irradiance Cache is not accelerated by hardware RT, and that gains vary by scene.
2. Does Octane use RTX acceleration by default?
Yes. In Octane for Cinema 4D, Use RTX acceleration is in the Device Settings dialog and is enabled by default.
3. Is the RTX 5090 faster than the RTX 4090 in Redshift and Octane?
It has more CUDA cores, newer RT cores and more memory bandwidth, so frames are generally faster, but the gain depends on how much of the scene is ray tracing. Measure it on your own scene.
4. Do four RTX 5090s give 128 GB of VRAM?
No. In both Redshift and Octane each GPU holds its own copy of the scene, so each frame must fit in 32 GB. More GPUs make rendering faster, not larger.
5. How many GPUs can Octane use?
With a Studio+ subscription there is no limit on the number of GPUs, according to OTOY. Free Prime licenses are limited.
6. How does Redshift split a frame across GPUs?
In bucket mode each bucket is rendered by a single GPU, so more GPUs render more buckets at the same time. The default bucket size is 128 × 128.
7. Which is faster on RTX 5090, Redshift or Octane?
It depends on the scene and settings. Compare both on the same server with the same noise target and a short sequence rather than a single frame.
8. Can I test Redshift and Octane on the same iRender server?
Yes. Cinema 4D, Redshift and Octane are pre-installed. Redshift uses your Maxon license or iRender’s on Prime accounts, and Octane uses your own OTOY license.
Render your Cinema 4D, Redshift and Octane 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. Cinema 4D, Redshift and Octane are pre-installed on every server; Octane runs with your own OTOY license.
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.
These quick starts show how to start a server and render with each engine:
Two runs we recorded on 4x RTX 5090 servers: Redshift, 150 frames at 1080p at 40.84 seconds per frame (about 1h 42m), and Octane, 300 frames in 1h 18m. They use different scenes, so read each on its own:
New users get a 100% bonus on the first top-up (minimum $50). See plans on the Cinema 4D and Redshift render farm and Octane render farm pages, or download the iRender GPU app.
Sources: NVIDIA: GeForce RTX 5090 specifications; Maxon: Redshift Globals; Maxon: Redshift Globals (Hardware Ray-Tracing); Maxon: Redshift Memory; Maxon: Redshift System (buckets); Maxon: Redshift Sampling – Production; OTOY: Devices tab (Cinema 4D); OTOY: Hardware requirements; OTOY: Hardware requirements (multi-GPU); OTOY: Out-of-core settings; OTOY: OctaneRender subscriptions.







