Skip to main content

NPU Instance

NPU stands for Neural Processing Unit, a hardware specialized in performing artificial neural network tasks. NPU-based instances refer to instances that use these NPUs to accelerate tasks.
NPUs are optimized to efficiently process deep learning tasks differently from traditional Central Processing Units (CPUs), capable of handling vector and matrix operations in parallel, and quickly performing large-scale matrix multiplication and convolution operations. Such NPU-based instances are particularly useful when handling large-scale deep learning tasks. They can train and infer deep learning models faster, significantly improving the performance and throughput of deep learning applications.

  • Applicable type: gf1i

gf1i​

gf1i instance is powered by 4th Gen Intel Xeon Scalable Processors and is equipped with FuriosaAI WARBOY NPU (Neural Processing Unit) cards, making it highly suitable for deep learning workloads.

Hardware specifications​
  • Up to 3.4GHz 4th generation Intel Xeon Scalable Processor (Sapphire Rapids 6430)
  • Up to 50Gbps network bandwidth
  • Instance sizes supporting up to 96 vCPUs and 768GB memory
  • Up to 4 FuriosaAI WARBOY NPUs
  • Support for Intel instruction sets (AVX, AVX2, AVX-512)
  • Support for enabling/disabling CPU multithreading
Use cases​
  • Acceleration tasks for high-performance deep learning model inference such as object detection/face recognition
  • Acceleration tasks for inference of deep learning models in the CNN(Convolutional Neural Network) family of computer vision areas
Detailed information​
Instance sizeNPUvCPUMemory (GiB)Network Bandwidth(Gbps)
gf1i.6xlarge12419212.5
gf1i.12xlarge24838425
gf1i.24xlarge  4    96    768     50