Picking a V-Ray render farm gets hard when the scene renders fine on your workstation, but the animation has 600 frames and the deadline is Friday. Which farm can take the job without changing the look you approved? A ranking will not answer that, but four checks will: which renderer mode the scene uses, which V-Ray build the farm runs, how much memory the node has, and how the meter is billed. This guide walks through each check and then looks at what Fox, GarageFarm, RebusFarm and iRender publish about V-Ray as of November 2026.

1. V-Ray Render Farm Basics: What Actually Runs on the Node

A V-Ray render farm is a pool of rented machines that run V-Ray on your scene, on CPU, on NVIDIA GPUs, or on both together. SaaS farms accept an uploaded scene and schedule it on their own nodes. IaaS services give you a remote server where you install the host application, V-Ray and the license yourself, and iRender works this way.

First, the renderer mode matters before any price comparison. Chaos explains that V-Ray Hybrid, which runs CUDA code on GPUs and CPUs together, is a separate renderer from V-Ray Production, the CPU engine, and said in December 2025 that the two engines will remain separate. Puget Systems also adds that hybrid mode uses CUDA and therefore does not use RT cores. As a result, a scene tuned on one engine can behave differently on the other, which is why the mode belongs in your first message to any farm. Our Arnold farm guide covers the same CPU and GPU question for a different renderer.

1.1 SaaS Upload or Remote Server: What Changes for V-Ray

For example, with a SaaS farm the service controls the V-Ray build, the plugins and the node hardware, and you choose from what it lists. With a remote server you control the build and the plugins, but you handle the installation, the license activation and the session yourself. However, neither model is better by default. SaaS suits a standard scene that fits the farm’s supported list, and a remote server suits a scene that depends on an unusual build or a plugin the farm does not offer.

2. Which V-Ray Version Should a V-Ray Render Farm Match in 2026?

First, match the exact build in your scene file, which is now V-Ray 7 Update 4 or Update 5. Chaos published Update 5 on 5 November 2026, five days before this article, and rolled Update 4 out host by host in July and August, so the hotfix level differs between 3ds Max, Maya and Houdini. Therefore, a farm that lists only “V-Ray 7” may not run the same build as your workstation.

Motion Media reports V-Ray 7 Update 4 Hotfix 2 for 3ds Max on 13 August 2026 and Update 4 Hotfix 1 for Maya on 29 July 2026. Architosh describes the Update 4 feature set, including the Veras AI engine in the frame buffer, VRayParallaxTex, quick caustics and 3D Gaussian Splats, and says the Houdini build was due in mid-July 2026. In addition, a scene that uses one of those features needs a node with a build that includes it. Chaos’s Update 5 adds an AI Assistant, MCP server support and AMD GPU rendering via HIP. We found no V-Ray 8 announced on Chaos’s site at the time of writing.

However, the farm pages we read are less specific. RebusFarm lists V-Ray up to 7.0, GarageFarm lists 3.4 to “7+”, and Fox lists “1.5 or later”. None of them names Update 4 or Update 5 as an installed build. Because the base release and the latest hotfix can differ in shader behavior, ask each farm which point update runs on its nodes before you upload.

3. V-Ray CPU or GPU: Which Hardware Fits Your Scene

Choose V-Ray CPU when the scene needs more memory than a graphics card holds, when it relies on older plugins, or when the same look must repeat across hundreds of frames. On the other hand, choose V-Ray GPU when the scene fits in video memory and you want fast iteration. Test one frame on both before committing a farm budget.

3.1 Memory and Scaling on a V-Ray Render Farm

Puget Systems says V-Ray CPU performance scales with clock speed and core count, and that V-Ray GPU needs more PCIe lanes than cores, with Chaos suggesting about six cores per card. Also, its rule of thumb for GPU work is system RAM equal to twice the total VRAM. In addition, multi-GPU scaling is strong but not perfect, so the second card never doubles the speed. Puget’s January 2025 launch review measured the RTX 5090 about 38% faster than the RTX 4090 in V-Ray RTX on V-Ray 6, and newer builds may change that.

However, one misunderstanding costs real money here. The RTX 4090 has no NVLink, and as far as we know the RTX 5090 has none either, so VRAM does not pool. For instance, an eight-card server renders a fitting scene faster, but it cannot render a scene that exceeds a single card’s 24GB or 32GB. The NovEdge V-Ray tip makes the matching point from the other side: do not assume a CPU scene looks identical when switched to GPU, and check displacement, proxies, render elements and denoiser output first.

4. What Each V-Ray Render Farm Publishes About V-Ray Support

Farm pages show how each service describes its V-Ray support, not how fast it renders. Table 2 lists what each page states, in alphabetical order, as read on 10 November 2026.

Two points stand out. First, version ranges are not the same as installed builds, so the written confirmation in the table matters. Second, because no page we read gave a V-Ray benchmark on its own hardware, so we make no speed comparison between farms. RANCH’s V-Ray page would not load during our research, and we leave RANCH out rather than guess. For a 3ds Max scene, our 3ds Max render farm guide goes further on host-specific checks.

5. iRender for V-Ray: Remote GPU Servers and Their Prices

iRender rents dedicated GPU servers by the hour, so a V-Ray job runs on a Windows or Ubuntu machine that you configure yourself. That suits V-Ray GPU scenes that fit within 24GB or 32GB of video memory per card. It is a smaller fit for V-Ray CPU work that needs very many cores, because the price page lists 16-core and 32-core Threadripper PRO processors, while GarageFarm’s page cites thousands of CPU cores across its farm.

5.1 V-Ray Render Farm Packages and Price per Card

Per card, the 2-card and 4-card RTX 4090 packages work out to $7.50 per card hour and the 8-card package to $6.50 (our arithmetic from the listed rates). However, that gap only pays off when your scene scales across cards, which Puget describes as strong but not perfect.

iRender publishes a walkthrough of its 4x RTX 4090 server running Maya and V-Ray on its older YouTube channel. Any timings in it come from the provider, and we have not reproduced them, so we cite none.

5.2 Video, Data Handling and Billing

In practice, iRender saves the machine after every session, so V-Ray, plugins and files on the C drive are still there the next time you boot it. Instead, what you can lose is work that never reached the disk, such as an unsaved scene or output written straight to the network Z drive during a dropped connection. iRender’s documentation recommends rendering to the local C drive and moving the finished output to the Z sync drive afterwards. Boot takes roughly 7 to 15 minutes according to the docs, which say the boot step is not charged, although an iRender blog post describes the clock as starting at boot, so check your first invoice.

Also, the drawback to plan for is the meter. According to iRender’s docs, billing starts at the Connect step and stops when you press Shutdown in the app or web, or when the Scheduler shuts the machine down. The docs list no other way to stop it, so do not rely on closing the Remote Desktop window. On data protection, iRender states that its data center is rated Uptime Tier III and holds ISO 27001 (BS7799), ISO 20000 (ITIL) and ISO 9001:2000 certification. Its terms and conditions say client projects are kept confidential, an NDA can be signed on request, and complete render projects are not stored after a job is finished, while the machine environment itself is kept for a limited period after your last session. We have not verified these claims independently.

6. Pre-Flight Checklist Before You Send a V-Ray Job to a Farm

Five steps catch most failed V-Ray farm jobs before they cost money. Run them on every new scene, not just the first one.

6.1 Step 1: Pin the V-Ray Build and Renderer Mode

First, record the exact V-Ray build from the V-Ray About dialog or your plugin manager, and the production engine selected in Render Setup (CPU, or the GPU options). Compare both with the farm’s written list, and ask about your exact update by name.

6.2 Step 2: Gather Every Asset

Collect proxies, bitmaps, IES files, HDRIs and any cache files with the host’s archive or collect-assets function, and check that no path points to a local drive. For example, a missing proxy fails silently in some scenes and shows up as an empty frame.

6.3 Step 3: Render One Test Frame

Then pick a middle frame at final resolution and note the render time and peak memory. On a remote server, watch GPU memory in Task Manager or nvidia-smi so you know whether the scene sits near the 24GB or 32GB limit.

6.4 Step 4: Compare Against Your Local Frame

Next, open both frames in an image editor and check render elements, denoiser output and any displacement. NovEdge lists these as the places where GPU and CPU results most often differ.

6.5 Step 5: Budget the Job and Set a Stop

Finally, multiply minutes per frame by frame count, divide by the number of nodes, add time for file transfer, and multiply by the hourly rate. On a remote server, set the Scheduler or press Shutdown the moment the last frame finishes.

7. Cost Maths for a V-Ray Render Farm Job

On a remote server, cost equals connected hours multiplied by the package rate. In contrast, on a SaaS farm, cost equals the unit price multiplied by the work units, and only the farm’s own estimator can turn your scene into units. RebusFarm’s unit prices, for example, are 1.20 US cent per GHz hour for CPU and 0.45 US cent per OB hour for GPU, before VAT, which cannot be converted to an hourly figure without a benchmark of your scene.

Here is a worked case with illustrative inputs, not measurements. For example, suppose a 120-frame sequence takes 4 minutes per frame on one RTX 4090. That is 480 minutes, or 8 hours, on a 3S node: 8 x $7.38 = $59.04 at the 3-hour-plus rate. On a 4S node with two cards, perfect scaling would give 4 hours and 4 x $13.50 = $54.00. However, real scaling is not perfect, and at 4.5 hours the 4S job costs $60.75, already above the single card. Measure your own scene before choosing the bigger package.

Maximum Speed, Absolute Freedom. Take control of your VFX workflow with iRender’s dedicated GPU servers. Render with iRender →

See more: Top 5 Cinema 4D Redshift render farm comparison

Technical FAQ: V-Ray Render Farm

What is a V-Ray render farm?

A V-Ray render farm is a service that runs V-Ray on rented machines instead of your workstation. SaaS farms take an uploaded scene and schedule it on their own CPU or GPU nodes. IaaS services such as iRender give you a remote server where you install V-Ray and the host app yourself.

Is V-Ray GPU or V-Ray CPU better on a render farm?

Neither wins everywhere. V-Ray GPU is fast when the scene fits in card VRAM, 24GB on an RTX 4090 and 32GB on an RTX 5090. V-Ray CPU renders from system RAM and suits memory-heavy scenes and long animations. Test one frame on both engines before you choose.

Which V-Ray version is current in 2026?

V-Ray 7 Update 5, published by Chaos on 5 November 2026, is the newest. Update 4 shipped host by host before it, including Maya 7.4 Hotfix 1 on 29 July 2026 and 3ds Max Hotfix 2 on 13 August 2026. We found no V-Ray 8 announced on Chaos’s site.

Do render farms include a V-Ray license?

Check each farm. On iRender remote servers you install V-Ray yourself, and the price page we read does not state that a renderer license is included, so plan to bring your own. For SaaS farms, ask support in writing before you upload.

How much does a V-Ray GPU render farm server cost per hour?

On iRender, RTX 4090 packages run from $8.20 to $52.00 per hour and RTX 5090 packages from $10.80 to $62.00, with a 10% lower rate for 3 hours or more. Per iRender’s docs, billing runs from the Connect step until you press Shutdown.

Can eight RTX 4090 cards pool their VRAM for V-Ray GPU?

No. RTX 4090 cards cannot pool memory, so each card must hold the scene in its own 24GB. Eight cards render faster when the scene fits, but they do not let a larger scene fit. A scene that passes 24GB needs V-Ray CPU or a lighter scene.

Sources

• Chaos, “Understanding V-Ray Hybrid Rendering”, blog.chaos.com, read 10 Nov 2026
• Puget Systems, V-Ray hardware recommendations, page dated 21 Oct 2026, read 10 Nov 2026
• Motion Media, “V-Ray 7.40.04 for 3ds Max and Maya”, read 10 Nov 2026; Chaos, “V-Ray 7 Update 5”, 5 Nov 2026; Puget Systems, GeForce RTX 5090 content creation review, 23 Jan 2025
• Architosh, “Looking at Chaos V-Ray 7 Update 4”, 15 Jul 2026
• NovEdge, “V-Ray CPU vs. GPU: Choosing the Right Renderer for Your Workflow”, read 10 Nov 2026
• Fox Renderfarm V-Ray page, GarageFarm V-Ray page, RebusFarm V-Ray page, read 10 Nov 2026; RebusFarm unit prices read 6 Nov 2026
• iRender price page and terms and conditions, read 10 Nov 2026; iRender documentation (docs.gpuhub.net), read 10 Nov 2026