English
Русский
Hosting provider FirstVDS has expanded its lineup of servers with graphics accelerators. Alongside traditional VDS with GPU passthrough, the company launched vGPU-powered plans. Now anyone looking to tinker with neural networks or render 3D hedgehogs can pick a less headache-inducing way to get graphical muscle.
What vGPU Means and How It Differs from Passthrough
vGPU is a technology that slices one physical graphics accelerator into several virtual pieces and hands them out to different servers. Each piece thinks it's the real deal. In passthrough mode, an entire GPU gets strapped to a single virtual server, leaving nothing for the neighbors.
The advantage of vGPU over passthrough is savings and flexibility. You can stuff several small tasks onto one physical GPU and avoid paying for a whole accelerator by yourself. Especially useful for projects that don't need an entire graphics card — just half will do. Plus, vGPU doesn't force the client to wrestle with device passthrough and driver compatibility — the host system already did the partitioning.
Under the Hood and Out of Your Wallet
The new vGPU-based VDS run on an NVIDIA L40S accelerator with 48 GB of memory — a beast, sure, but a virtual beast. The CPU side is an AMD EPYC clocked up to 3.9 GHz, with KVM virtualization. Launch takes a couple of clicks: the company prepared ready-made configurations — just pick one and go. No setup headache: OS templates already come with NVIDIA drivers and CUDA preinstalled. All that's left is to run your code and pretend it was always this easy.
Pricing starts at 299 rubles per day (about $3.50). Daily billing means no month-long commitments — use exactly as much as you need and don't pay for weekends when your server sits in the corner gathering digital dust. You can order this joy from the FirstVDS customer panel.