Unlike traditional hypervisors, Hyperborea also includes GPU virtualization for most enterprise and consumer GPUs. This enables you to spend less money on your software, increase your margins, and be more competitive on pricing for your customers.
Thanks to some of Hyperborea’s exclusive features like “Simultaneous Multi-Virtual GPU”, your customers will see over 65% increases in performance versus traditional market hypervisors. In addition to this, Hyperborea allows for up to 100% utilization of GPU resources by allowing production resources to utilize the execution capabilities and GPU cores of under-utilized and idle resources. This ensures that your customer is always getting what they’re allocated but usually much more.
For graphical applications, your customers will forget that they’re working in a virtual machine. Thanks to Hyperborea’s exclusive technology, applications regularly run upwards of 200 frames per second in their virtual machines.
To provide additional revenue capabilities, Hyperborea allows for up to 64 virtual GPUs per physical GPU (most models) with support coming for 256+. Why waste resources when your customers workloads don’t always utilize the full memory of your GPU?
Arc Compute’s hypervisor stack eliminates multiple costly virtualization fees, so you can spend less and charge more. Our cloud providers save 40-50 percent on average, vastly increasing their margins.
Thanks to Hyperborea’s architecture, applications can achieve speeds of up to 200 frames per second, worldwide.
With up to 64 virtual machines per GPU, and support coming for 256+, you can support greater user densities in your cloud deployments.
In addition to enterprise GPUs, Hyperborea supports a wide range of consumer NVIDIA, Intel, and AMD GPUs. Who wants to pay more for the same thing?
Hyperborea supports a wide range of operating systems including Ubuntu, Centos, PopOS, and Windows.
LibVF.IO is Arc Compute’s open source IOMMU parser and VFIO allocation library. Visit libvf.io for more details.
Thule is Arc Compute’s open mediated device driver, which we call the “nouveau of mdev drivers”. Thule supports AMD, NVIDIA and Intel hardware.
DDMAM performs host mediated vtg_balloon like API calls, allowing VRAM to be allocated at run-time, using hardware enforced IOMMU to maintain address separation.