Simple & clear.
Easy to setup.

GPU accelerated compute instances

Our GPU compute servers are available as bare metal or as virtual machine (VPS). The virtual servers are powered by:

  • Up to 8 NVidia Tesla V100 GPUs, each containing 5120 CUDA cores and 640 Tensor Cores
  • Second generation Xeon Scalable 4214R CPUs [2020], up to 48 threads in total topping out at 3.5GHz
  • NVLink for high bandwidth P2P communicaton

 

Instance GPUs vCPU Mem [GiB] GPU Mem [GiB] GPU P2P Storage [GB] Price [$/h]
1V100.6V 1 6 21 16 - 225 0.60
2V100.10V 2 10 43 32 NVLink 450 1.20
4V100.20V 4 20 90 64 NVLink 890 2.40
8V100.48V 8 48 180 128 NVLink 1750 4.80



The GPU's are passed through to the dedicated virtual machines, which delivers native performance. They are interlinked via NVLink up to 100GB/s.
The servers are dedicated: the hardware is allocated to your machine and your machine only.

The name 4V100.20V is composed as follows; 4x V100, 20 CPU cores & virtualised.

Our GPU compute servers are available as bare metal or as virtual machine. The 8V100.48M bare metal instance is powered by:

  • 8 NVidia Tesla V100 GPUs, each containing 5120 CUDA cores and 640 Tensor Cores
  • 2 Second generation Xeon Scalable 4214R CPUs [2020], 48 threads in total topping out at 3.5GHz
  • NVLink for high bandwidth P2P communication

 

Instance GPUs vCPU Mem [GiB] GPU Mem [GiB] GPU P2P Storage [GB] Price [$/h]
8V100.48M 8 48 192 128 NVLink 1800 5.00



A $5 setup fee is charged when ordering a bare metal instance. The 8V100.48M gives you access to the bare metal; the complete server. The GPU's are interlinked via NVLink up to 100GB/s. As all instances, this option is a dedicated server. This is the go-to option for training large models. We can interlink multiple servers or add additional storage/RAM/CPU on demand.

The name 8V100.48M is composed as follows; 8x V100, 48 CPU cores & bare metal.

What's next?

The next hardware we are taking to the datacenter is a storage cluster. This will enable backups, allow you to stop your server without losing data and could allow network storage similar to AWS' elastic block storage.

Besides storage, we will continue to add different flavours of high-end GPU servers tailored for AI & machine learning.

Arrow-up