Flexible Engine Elastic Cloud Server introduces new PI2 instances for GPU-accelerated use cases
Flexible Engine
Release Notes
September 1, 2020
Elastic Cloud Server (ECS) is Flexible Engine’s virtual machine service based on Openstack which offers a wide choice of flavors designed to run your workloads on highly performant instances at the best price.
ECS has receive the following updates :
Elastic Cloud Server is updated with a new New PI2 instances for GPU-accelerated use cases. They are equipped with NVIDIA Tesla T4 GPUs dedicated for real-time AI inference. These ECSs use the T4 INT8 calculator for up to 130 TOPS of INT8 computing. The PI2 ECSs can also be used for light-load training.
The P2s ECS features are:
- CPU: 2nd Generation Intel® Xeon® Scalable processors (Cascade Lake Xeo 6278 processors at 2.6 GHz of base frequency and 3.5 GHz of turbo frequency)
- Up to 4 NVIDIA Tesla T4 GPUs on an ECS
- GPU hardware passthrough
- Up to 8.1 TFLOPS of single-precision computing on a single GPU
- Up to 130 TOPS of INT8 computing on a single GPU
- 16 GiB of GDDR6 GPU memory with a bandwidth of 320 GiB/s on a single GPU
- One built-in NVENC and two NVDEC GPUs
This new ECS type is available in 3 different flavors, with a vCPU number/RAM GB ratio of 1:4.
pi2.2xlarge.4 |
8 |
32 |
10/4 |
50 |
4 |
1 x T4 |
1×16 |
pi2.4xlarge.4 |
16 |
64 |
15/8 |
100 |
8 |
2 x T4 |
2×16 |
pi2.8xlarge.4 |
32 |
128 |
25/15 |
200 |
16 |
4 x T4 |
4×16 |
Use Cases
PI2 flavors are suited for in GPU-based inference computing scenarios, such as image recognition, voice recognition, and natural language processing. The PI2 ECSs can also be used for light-load training.
Limitations
PI2 has a maximum number of 4 Nvidia Tesla T4 GPU with 64 GiB of GPU memory, 32 vCPU and 128 GiB of RAM.
PI2 ECSs support the following commonly used software:
- Deep learning frameworks, such as TensorFlow, Caffe, PyTorch, and MXNet
Console / API and Resources
As all other instance flavors, the new PI2 series is available through the console and API.
More information on ECS GPU-accelerated is available here