One GPUBox to rule all GPUs

You can connect all available GPUs, no matter if they are on Linux or Windows, and force  them to work on single render. Moreover you can use the GPUs on demand on any computer connected to the same network, even more, you can share the GPUs with others.

In the following video we are going to show you how to connect additional GPU to already existing GPUBox infrastructure. For the sake of demonstration we used Amazon EC2 instances however you can use any computers with GPUs. We choose EC2 to show you they are on demand and they are really cheap. If you start instances g2.2xlarge as spot, for a dollar you can have about 15 GRID K520 GPUs with the power of GTX 670 each for entire hour.

We connected Windows to already running 8 instances with Linux. Details of configuration are on the picture which speaks for itself.

Connect GPUBox to 9 GPUs You can find more information in our documentation

… so enjoy the watch

Soon we are going to show you

  • side by side how to install, configure and manage GPUs on two computers,
  • and how to start GPUBox on Amazon EC2 infrastructure with ease,

… and of course you can always share your thougts and write to us

GPUBox 1.6.58 released


We are pleased to announce the fresh release of GPUBox. The new version 1.6.58 supports Linux and the most awaited, GPUBox fully supports Windows with super easy to use Setup Wizard.

Along with the new version for Windows operating system we also released Open GPUBox which allows to use up to 4 GPUs for free for any purposes.

Now it is childishly easy to connect at least two computers, install GPUBox and rule all of the available GPUs. You can even create peer-to-peer alike network of GPUs and share the power among the team’s members.

In the new version we fixed a few bugs and improved the stability and the performance.


Major GPUBox 1.6.58 features:

  • Supports Windows and Linux.
  • Open GPUBox, a free version to use for any purposes.
  • Easy to configure and manage.
  • Supports native InfiniBand
  • Intuitive Web Console
  • Sharing the same device among many users
  • Dynamic provisioning of any number of GPU devices
  • Scalability and flexibility
  • Reduction of total cost and better GPU utilization

Download GPUBox:
Documentation online:


Major update is coming…


As we have announced a while ago, within the next update of GPUBox Arion will join the portfolio of officially supported GPU-based renderers. Our cooperation with RandomControl developers already resulted in exceeding the limit of 8 GPUs (that some of you might have noticed while watching our previous video). We are thrilled about that because the GPUBox-Arion combination exhibits astonishing potential in terms of scalability and performance. Just take a look at the video:

Stay tuned for the next updates. Soon we are going to reveal some more information about supported scientific applications and further GPU-based renderers (yes, V-Ray RT as well…)

GPUBox is going to support Arion


We can now officially announce that GPUBox is going to support RandomControl Arion.

Arion is a hybrid, GPU-accelerated, and physically-based production render engine capable of generating hyper-real images. RandomControl offers it in stand-alone version as well as plugins for 3ds Max and Rhinoceros. First tests of this renderer running on GPUBox proved that it exhibits great scalability and nearly-native performance while using multiple GPUs virtualized with GPUBox.

We launched the Arion benchmark on 4 x GTX 690 (8 GPUs) running within our local testing environment as well as on 6xGRID K520 within GPUBox Web Service:


Official support for this outstanding renderer will be introduced to GPUBox within the next major update. Besides Arion, GPUBox is also going to open for scientific applications such as LAMMPS, CUDASW++, HOOMD-blue, Gromacs or BarraCUDA to name just a few. And last but not least – GPUBox GPUServer will no longer require Linux. Windows version is currently in the beta-tests stage.

Short video tutorials for OServer and GPUServer

Some of our users claimed that the installation process of GPUBox components – especially OServer and GPUServer – is sometimes a bit problematic. As we understand that for lots of GPUBox users it is something completely new, we have prepared two short video tutorials presenting an exemplary installation of OServer and then GPUServer. We believe that those videos will be a good supplement for the documentation. In the video we used the demo version that can be downloaded for free in the Products section of our website. However the installation process for the full version looks nearly the same. Continue reading

Introducing GPUBox Web Service!


We are proud to announce that GPUBox Web Service is now available!

It is the first web service that combines the capabilities of GPUBox technology with computing power of Amazon EC2. GPUBox Web Service automatically configures a multi-GPU cloud computing environment that can be accessed from a regular web browser. In other words – it makes you a few clicks away from using an extremely powerful, GPUBox-enabled infrastructure to edit your scenes and render them in real time.

You are now able to deploy your private GPU cloud computing cluster in a few minutes without worrying about hardware or power supply. You do not need anything more than an Amazon AWS account and some GPUHours that you can purchase on our website. Besides, there are a few facts about GPUBox Web Service that you will find interesting:

It is great for heavy tasks

Engaging the processing power of dozens of high-end GPUs can decrease the time needed to complete your work by days or even months.

It allows you to keep your work secure

On the contrary to renderfarms, you can store your files on a private virtual machines hosted by Amazon AWS. You can exchange the files between your system and a virtual machine over secure transfer protocol. It gives you full control on who can access your precious files.

You can bring your own application

You can treat the virtual machine on Amazon AWS as your own system. It means that you can enhance your favorite application supported by GPUBox Artist with massive number of GPUs by simply installing it on your private virtual system within GPUBox Web Service. And if you are a Blender user – good news, it is already installed on the machine image and ready to work. At the moment GPUBox is working within CUDA environment, however we are going to introduce support for OpenCL and later also OpenGL. Continue reading

GPUBox Web Service teaser


We have published a brief teaser for the possibilities that will be brought to you with the upcoming GPUBox Web Service.

Lots of things are happening in this short video, but the most important is that we combined 50 GPUs from 50 Amazon EC2 instances (we used the g2.2xlarge instance type) with GPUBox and we used Blender as an exemplary application to show how it works.

GPUBox Web Service automatically configures a multi-GPU cloud computing environment that within a few minutes can be accessed from a regular web browser. In other words, you will be a few clicks away from using an extremely powerful infrastructure to render your scenes. Continue reading

Next steps of Renegatt Software


The time has come to reveal some of our future plans.

So, let us make it brief and official – Renegatt Software is going to introduce:

Dedicated versions of GPUBox Artist

GPUBox Artist licenses restricted only to a particular application will be available to be purchased at a lower price. At the beginning, we will release GPUBox Artist separately for Blender and Octane. Each version will cost €59 per GPU.

GPUBox Web Service

Thanks to GPUBox Web Service you will be a few clicks away from using more GPUs than ever. The first edition of GPUBox Web Service is going to use Amazon EC2 instances.

Windows Server and Mac OS X Client

At the moment the GPUs can be served to the infrastructure only from Linux servers. Some of you would like to virtualize GPUs under Windows, and this is what we are going to do. Also, we are planning to develop GPUBox Client for Mac OS X.

Support for new applications

More CUDA-based renderers are going to join Blender and Octane.


If you want to be always up-to-date with us, follow Renegatt Software on Twitter/Facebook or subscribe to our newsletter.

Blender Cycles benchmark on GPUBox Artist


In the previous post we delivered you some information about scalability and performance in Octane Render. Now it is time to take a closer look at Blender and see how the situation looks here. We recorded another video showing rendering widely-used benchmark scene BMW 1M by MikePan, but instead of default settings we applied 8000×4200 resolution and 1000 samples.

When it comes to using multiple GPUs, Blender is rather a long-runner than a sprinter: the heavier is the scene and the longer is the rendering, the better is the scalability. Adding more and more GPUs will not be very efficient if simple, low-quality scene is being rendered. Preparing and distributing the scene to GPUs is not always made up with the performance boost during the rendering itself.  But the magic happens while using multiple virtualized GPUs to render complex scenes on high settings. Continue reading

Performance and scalability in Octane Render


One of the most popular questions regarding GPUBox Artist refers to performance and scalability. We prepared a short but substantial clip that presents you those two aspects of using Octane Render within the GPUBox-powered infrastructure.

For the purpose of this video we used Octane Render 1.55 Standalone running on CentOS 6.4. The scene that was used for the test was well-known OctaneBenchmark scene on default settings (except samples, which were set to 6000) and Path Tracing kernel.

In the first part of the video we compared two rendering sequences. The left side presents rendering the scene on native 4 GPUs (2 x GeForce GTX 690). Simultaneously, on the right side of the screen you can see rendering the very same scene on the same settings, on identical GPUs, but mounted in a different PC and virtualized with GPUBox Artist.

Native rendering was finished after 4:05, while rendering using the GPUs remotely with GPUBox took 4:07.

In the second part we virtualized and engaged another six GeForce GTX 690 cards (12 GPUs) and launched again the same scene on a total number of 16 GPUs comparing it to native rendering on 4 GPUs. Theoretically, rendering on 4 times as much GPUs should result in 4 times faster rendering and this is what happened here – the rendering took 1 minute and 1 second.

During the test we were using a 20Gb/s InfiniBand network, but additionally we launched the same scene on 1 Gb Ethernet which is not presented in the video. The results are shown in the following table: Continue reading

Subscribe to our newsletter to receive latest
news and updates.