Virtual machines (VMs) are a huge benefit to the computing world. However, setting up a VM, and ensuring that the bulk of the software that you need is installed and updated can be a time consuming and complex process. What’s more, pay for use VMs are often complex and less up to date, and so they’ve outlived their usefulness. Enter Paperspace. Last week the company announced that they would begin offering a new VM built on a Linux platform with some impressive stats designed for the niche market of data scientists, but suitable for most complex computing including cloud gaming.
Paperspace starts by offering the familiar Linux desktop setup which can then be accessed as a shell or directly via web browser. They’ve even built their own terminal inside Chrome. “During the beta, we quickly learned that while many people are proficient in a terminal, they tend to prefer working in a desktop environment. To that end, we built a full desktop experience that runs directly in a web browser. Because the browser is so powerful, we also took a stab at building our own terminal, directly inside Chrome.” With this new system they’re offering some surprisingly powerful computing options at prices that look to compete with some of the bigger players.
The hardware upgrade that’s drawing attention is the use of NVIDIA Pascal GPUs on the machines. This system allows them to optimize for ultra high-end simulation and machine learning (ML). They’ve also built in a preloaded collection of really useful ML tools. Their current version has fully updated versions of CUDA, cuDNN, TensorFlow, Python, OpenBLAS, Caffe, Theno, Torch, and Keras as well as all the latest GPU drivers. The helpful part of the VM world is that this template is built once and is then offered through Paperspace as a built machine. Their goal is to make a machine that is useful, workable, and without the kinks, flaws, and time-wasters of user built machines on other cloud services like Amazon Web Services (AWS).
What’s more, this very powerful and helpful platform is available at some really remarkable rates. Right now, they are offering access to the Pascal machines with 16 GB of memory and 2560 CUDA cores for a mere $.65 per hour. This is a huge undercut to the current market pricing. While the MLaaS (machine learning as a service) market space has been under some scrutiny of late (because of the apparent contradiction between technical users seeking simplified solutions), Paperspace is offering something that is particularly unique, and should have some traction in the right fields. Already Y Combinator is working on using the streamlined VMs in some of its newest AI startups. NYU and Insight Data Science have also partnered with Paperspace for using the GPU machines.
The future is unclear, but the access to such powerful machines at such an attractive rate will likely inspire some really serious looks from companies using less attractive cloud computing models.