Houdini: what’s installed, what you bring.
Already installed
- Blender
Bring your own licence
- SideFX Houdini
- Karma (included with Houdini)
- Redshift, if you use it
Houdini signs in with your SideFX licence; Karma ships with Houdini. Each session starts from a clean machine, so software you install is removed when it ends, unless you choose a persistent workstation (available on selected GPUs), which keeps your own Windows disk between sessions. Everything on the machine.
The GPU we recommend for Houdini
RTX A5000, 24 GB VRAM
Karma XPU and GPU renderers keep the scene in VRAM; most simulation runs on the CPU and system RAM, listed with each card.
What Houdini costs per hour here.
One price per GPU-hour, in US dollars. The recommended card is picked; tap another to change it.
RTX A5000 chosen. You sign up when you press Start: your email and a WhatsApp code. You pay before any GPU time, then download Connect.
Blender, Unreal Engine, Twinmotion and D5 Render are preinstalled. Commercial software runs on your own licence.
How to sign out of common appsYou sign in to Houdini with your own licence; the price is for the machine. Billed per minute from when you connect, not per whole hour. Preparing the workstation and uploads are free. If you don't connect within 10 minutes of your workstation being ready, it shuts down and those 10 minutes are charged. If a connection fails on our side, nothing is charged. Full pricing.
How Houdini runs here: measured, not quoted.
We publish only figures we have measured on our own workstation, and we have not published Houdini timings yet. Each row says so. Ask and we will run your scene, or a reference one, and send the numbers.
| What was measured | Result |
|---|---|
| Karma XPU 4K still, our reference scene | Measured on request |
| Peak VRAM used by that reference scene | Measured on request |
| Workstation ready after Launch (sign-in to desktop) | Measured on request |
Want a number before you top up? Ask for a measured run. We will tell you the GPU, the scene and the settings with every figure.
The details.
Where Houdini uses the GPU, and where it doesn't.
Most of Houdini's simulation work (FLIP, pyro, Vellum and RBD) runs on the CPU and needs system RAM for its caches. The GPU draws the viewport and runs Karma XPU, Redshift and the OpenCL parts of some solvers.
So pick the card for your renderer's VRAM and check the workstation's vCPU and RAM, listed with each card, for your simulations. Write caches to your project drive so they are there next session.
Licensing. You install it and sign in with your own licence, exactly as you would on a new PC. We don't supply licences for commercial software, because the licence belongs to you and not to the machine. Houdini signs in with your SideFX account or your own licence server; Karma ships with Houdini.
See what is preinstalled →Questions.
Can I run Houdini on a cloud workstation?
Yes. A Coreframe workstation is a full Windows desktop (Windows 11 or Windows Server) with a dedicated NVIDIA RTX GPU, and Houdini runs on it with your own SideFX licence. You reach it through the Coreframe Connect app and pay per minute.
Does Karma XPU work on a Coreframe workstation?
Yes. Karma XPU uses the NVIDIA RTX GPU and the CPU together, and Karma comes with Houdini, so there is nothing extra to license.
Will my Houdini simulations run faster?
Only if your own machine is short of CPU cores or RAM. Most Houdini simulation runs on the CPU, so compare the vCPU and RAM listed with each card against what you have. We don't quote simulation times because they depend on the setup.
Does Coreframe provide a Houdini licence?
No. You sign in with your own SideFX account or point Houdini at your own licence server. Coreframe supplies the machine.
What GPU do I need for Houdini?
For Houdini we recommend the RTX A5000 with 24 GB of VRAM. Karma XPU and GPU renderers keep the scene in VRAM; most simulation runs on the CPU and system RAM, listed with each card. Every card on the price list runs it; you pick the card when you start.
How much does a Coreframe workstation cost?
Ad-hoc sessions are $1.49 per GPU-hour, billed per minute from when you connect, not per whole hour. Preparing the workstation and uploads are free. If you don't connect within 10 minutes of your workstation being ready, it shuts down and those 10 minutes are charged. If a connection fails on our side, nothing is charged. Committed monthly plans start at $70.99/month and bill extra hours below the ad-hoc rate at every tier: $1.99/hr on Small Team, $0.99/hr on Medium Team. Prices in US dollars carry no GST (export of services), and the invoice is in US dollars.
Are prices the same worldwide?
Each billing country has one price list. Accounts billed outside India pay in US dollars, from $1.49 an hour with no GST (export of services), billed per minute. Accounts billed in India pay a rupee price list with GST included. The workstations are the same everywhere, hosted in a Tier III data centre in India.
Is there an offer for new customers?
Yes. Your first top-up of $20 (no GST) buys 8 hours on an RTX 5080 (or about 18 hours on the RTX A4000, 12 on the RTX A5000), to use within 30 days, one per customer. You sign up with an email and a WhatsApp code to your mobile, with no ID documents, and international cards work. We won't say what you can finish in that time, because it depends entirely on your scene.
No account until you press Start. New to it? How a session works.
Related
