Viettel IDC

Cloud NPU (vCNPU)

A unique and optimized infrastructure solution for new era AI.

Cloud NPU is a Public Cloud service providing to customers Virtual Machines attached with one or many NPU Qualcomm Cloud AI 100. It is aimed at businesses and organizations that need to deploy and provide artificial intelligence (AI) applications including computer vision (CV), natural language processing (NLP), generative AI (GenAI), and more. Viettel Cloud NPU helps businesses and organizations solve the price-performance (P/P) challenge of deploying AI applications.

NPU stands for Neural Processing Unit. It is a specialized integrated circuit (ASIC) designed specifically for AI processing.
Support a wide range of deployment models, from small to large scale

Support a wide range of deployment models, from small to large scale

Up to 06 NPU Qualcomm Cloud AI 100 can be attached to one single Virtual Machines, those machines then can easily and privately connect through a virtual Switch

Organize and manage AI core in a flexible manner

Organize and manage AI core in a flexible manner

Each Qualcomm NPU device has multiple AI cores. Users can assign one or many AI cores to each AI model running concurrently on the NPU. They can also use the Scheduler feature to automate the allocation and scheduling of idle AI cores to each AI model.

Fully support popular AI Frameworks and Models

Fully support popular AI Frameworks and Models

NPU Qualcomm works perfectly with common AI Frameworks including but not limited to TensorFlow, PyTorch, ONNX, and with common AI Models like ResNet, YOLO, BERT, GPT, and more.

Powerful SDK enabling effortless code conversion

Powerful SDK enabling effortless code conversion

Qualcomm provides two powerful SDKs including Cloud AI Apps SDK and Cloud AI Platform SDK, helping AI Developers and Engineers performing the work of model optimization, compilation and deployment in an easy and effortless manner.

Powerful Management Toolkit

Powerful Management Toolkit

Qualcomm provides free management tools in both command line and graphical user interface (GUI) formats, supporting monitoring NPU parameters during operation as well as fine-tuning parameters, workflow, and compiling AI models.

Easy Integration with Kubernetes & Containers

Easy Integration with Kubernetes & Containers

Qualcomm provides a device plugin for Kubernetes to use in conjunction with their NPU, without the additional license costs like some of other GPU vendors. It supports resource-mapping down to the level of individual containers with individual AI cores on the card.

Well-suited for a wide range of AI, excel in Computer Vision

Well-suited for a wide range of AI, excel in Computer Vision

Qualcomm NPU works well with a variety of AI Inference workloads, from Computer Vision (CV), Natural Language pProcessing (NLP) to newer types of AI such as generative AI (GenAI). In many use cases, Qualcomm NPU demonstrates superior Perf/Watt compared to many equivalent GPU devices from other vendors.Qualcomm NPU works well with a variety of AI Inference workloads, from Computer Vision (CV), Natural Language pProcessing (NLP) to newer types of AI such as generative AI (GenAI). In many use cases, Qualcomm NPU demonstrates superior Perf/Watt compared to many equivalent GPU devices from other vendors.

Optimizing Cost with High Performance

Optimizing Cost with High Performance

Using Qualcomm NPU to deploy AI applications offers customers many cost benefits: all features and accompanying tools are free; Cloud NPU service rental costs are more economical due to Qualcomm NPU's Total Cost of Ownership (TCO) optimization.

​Easy code conversion

​Easy code conversion

The process of converting code running on GPU of other vendors to Qualcomm NPU is simple and fast, thanks to Qualcomm ‘s powerful SDKs, comprehensive conversion guide, plus in-depth technical support and hands-on training from the Viettel IDC-Qualcomm team of experts

Image & video processing

Popular AI models for Face Recognition, Object Detection, and Image Processing all run perfectly and are optimized on Qualcomm NPU. Qualcomm NPU are suitable for businesses and organizations that need to deploy systems and applications for FaceID logging, eKYC, AI Surveillance Cameras, etc.

Virtual Assistant & Chatbot

Qualcomm NPU is suitable for NLP AI models such as BERT, GPT, and CNN-Text, meeting the needs for applications such as Text-to-Speech, Speech-to-Text, Virtual Assistant Customer Care, etc.

Generative AI (GenAI)

Generative AI models for text, images, videos, and code can all benefit from using Qualcomm NPU.

Robotics AI, AR/VR

Qualcomm NPU is a good fit for AI Robotics tasks such as navigation and environmental awareness, human-robot interaction, and learning & adaptation. It can also be used for AR applications in training and shopping.

Qualcomm NPU Cloud AI 100 is an ASIC (Application-Specific Integrated Circuit), which differs from GPGPU (General Purpose Graphics Processing Unit) of NVIDIA or AMD in the sense that: NPU is designed with hardware that only includes components for one particular purpose - AI Inference, while a GPGPU consists of many different components (different types of Cores) to meet the needs of several workloads such as video decoding, post-production film processing, cryptocurrency mining, and of course AI Training & Inference.

Cloud NPU provides customers with Virtual Machines attached with one or several (up to 06 per VM) NPU Qualcomm Cloud AI 100. You will access Cloud NPU service through SSH/Remote Desktop protocol; or through a Private Connection with other services of Vcloudia such as Cloud Server, Virtual Private Cloud, Dedicated Private Cloud.

Cloud NPU supports all popular OS such as Windows Server, Ubuntu, etc. with new versions. We also constantly update our OS supported list.

Cloud NPU can meet the needs of almost every type of popular AI that’s on the market right now, ranging from Computer Vision to NLP and GenAI.

You should NOT. NPU Qualcomm Cloud AI 100 is specifically designed and optimized or AI Inference only, therefore it might not be a wise choice for AI Training.

With Cloud NPU, we guarantee the service uptime SLA at 99.5% and higher, which helps minimize your worries about service disruptions affecting end-user experience.