© 2022 Images et Technologie | All rights reserved | Privacy Policy (GDPR)
Graphcore
IPU-POD
Second generation IPU systems for
AI infrastructure at scale
IPU-MACHINE: M2000
A core, new building block for AI infrastructure, the IPU-M2000 is powered by 4 x Colossus Mk2 GC200, Graphcore’s second generation 7nm IPU. It packs 1 PetaFlop of AI compute, up to 450GB Exchange Memory and 2.8Tbps IPU-Fabric for super low latency communication, in a slim 1U blade to handle the most demanding of machine intelligence workloads.
IPU-M2000
Key Features
Compute
• 4 x Colossus™ Mk2 GC200 IPU
• 1 PetaFlop AI compute
• 5888 independent processor cores
WORLD-CLASS
PERFORMANCE
IPU-POD systems support industry-standard software tools. Developers can work with frameworks such as TensorFlow, PyTorch, PyTorch Lightning and Keras, as well as open standards like ONNX and model libraries like Hugging Face
Graphcore IPU Lets Innovators Make New Breakthroughs in Machine Intelligence
Memory
• Up to 450GB Exchange Memory™
• 180TB/s Exchange Memory™
bandwidth
Communications
• 2.8Tbps ultra-low latency IPU-Fabric™
• Direct connect or via Ethernet switches
• Collectives and all-reduce operations
support
Software
• Poplar SDK
• PopVision visualization and analysis tools
Converged Infrastructure Support
• Virtual-IPU comprehensive virtualization and workload manager support
• Support for SLURM workload manager
• Support for Kubernetes orchestration
• OpenBMC management built-in
• Grafana system monitoring tool interface
IPU Gateway SoC
• Arm Cortex quad-core A-series SoC
Form Factpr
• Industry standard 1U
16
IPU-POD16 DA (Direct Attach) is the ideal platform
for exploration, innovation and development. This
lets AI teams make new breakthroughs in machine
intelligence. Four IPU-M2000s, supported by a host
server, deliver a powerful 4 petaFlops of AI compute for
both training and inference workloads in an affordable,
compact 5U system.
For Innovators making new AI Breakthroughs
IPU-POD16 opens up a new world of machine intelligence innovation.
IPU-POD16 DA is designed to get you up and running in no time. A turnkey system, featuring IPU-M2000s directly attached to an approved host server ready for installation in your datacenter. Extensive documentation and support is provided both by AI experts at Graphcore and our elite partner network.
IPU-POD16 DA is a powerful, standalone AI compute resource. However, it also offers the opportunity for growth, on your terms. Your IPU-POD16 DA system investment can be expanded later into a larger IPU-POD system
Designed specifically for the communication requirements of AI workloads at scale, IPU-Fabric is Graphcore’s innovative low-latency, jitter-free interconnect using industry standard IT equipment. It supports highly efficient, deterministic, all-to-all IPU interconnect across your system regardless of size.
Fast & efficient for dense matmul models
Excels at sparse & fine-grained computation
Expert support to get you up and running quickly
Graphcore
IPU-POD
64
The IPU-POD64 is a single rack configuration featuring 16 IPU-M2000™ compute blades, based on the innovative GC200 Intelligence Processing Unit (IPU). The IPU-POD64 can deliver up to 16 petaFLOPS of AI compute.
The Building Block for Machine Intelligence Scale-Out
The whole system, hardware and software, has been architected together. IPU-POD64 supports standard frameworks and protocols to enable smooth integration into existing data center environments. Innovators can focus on deploying their AI workloads at scale, using familiar tools while benefitting from cutting-edge performance.
Built for AI developers
For deeper control and maximum performance, the Poplar framework enables direct IPU programming in Python and C++. Poplar allows effortless scaling of models across many IPUs without adding development complexity, so developers can focus on the accuracy and performance of their application.
Plug-and Play with Direct Attach
AI infrastructure built to scale
Start small, scale big
EXTENSIVE ECOSYSTEM
Software tools and integrations to support every step of the AI lifecycle from development to deployment to improve productivity and AI infrastructure efficiency. And just make it easier to use.
Designed from the ground up for high performance training and inference workloads, the IPU-M2000 unifies your AI infrastructure for maximum datacentre utilization. Get started with development and experimentation then ramp to full scale production.
Graphcore IPU Lets Innovators Make New Breakthroughs in Machine Intelligence
WORLD-CLASS
PERFORMANCE
Software tools and integrations to support every step of the AI lifecycle from development to deployment to improve productivity and AI infrastructure efficiency. And just make it easier to use.
EXTENSIVE ECOSYSTEM
IPU-POD systems support industry-standard software tools. Developers can work with frameworks such as TensorFlow, PyTorch, PyTorch Lightning and Keras, as well as open standards like ONNX and model libraries like Hugging Face
Built for AI developers
For deeper control and maximum performance, the Poplar framework enables direct IPU programming in Python and C++. Poplar allows effortless scaling of models across many IPUs without adding development complexity, so developers can focus on the accuracy and performance of their application.
Second generation IPU systems for
AI infrastructure at scale
IPU-MACHINE: M2000
A core, new building block for AI infrastructure, the IPU-M2000 is powered by 4 x Colossus Mk2 GC200, Graphcore’s second generation 7nm IPU. It packs 1 PetaFlop of AI compute, up to 450GB Exchange Memory and 2.8Tbps IPU-Fabric for super low latency communication, in a slim 1U blade to handle the most demanding of machine intelligence workloads.
IPU-M2000
Key Features
Compute
• 4 x Colossus™ Mk2 GC200 IPU
• 1 PetaFlop AI compute
• 5888 independent processor cores
Memory
• Up to 450GB Exchange Memory™
• 180TB/s Exchange Memory™
bandwidth
Communications
• 2.8Tbps ultra-low latency IPU-Fabric™
• Direct connect or via Ethernet switches
• Collectives and all-reduce operations
support
Software
• Poplar SDK
• PopVision visualization and analysis tools
Converged Infrastructure Support
• Virtual-IPU comprehensive virtualization and workload manager support
• Support for SLURM workload manager
• Support for Kubernetes orchestration
• OpenBMC management built-in
• Grafana system monitoring tool interface
IPU Gateway SoC
• Arm Cortex quad-core A-series SoC
Form Factor
• Industry standard 1U
Designed from the ground up for high performance training and inference workloads, the IPU-M2000 unifies your AI infrastructure for maximum datacentre utilization. Get started with development and experimentation then ramp to full scale production.
Designed from the ground up for high performance training and inference workloads, the IPU-M2000 unifies your AI infrastructure for maximum datacentre utilization. Get started with development and experimentation then ramp to full scale production.
Graphcore
IPU-POD
16
IPU-POD16 DA (Direct Attach) is the ideal platform for exploration, innovation and development. This lets AI teams make new breakthroughs in machine intelligence. Four IPU-M2000s, supported by a host server, deliver a powerful 4 petaFlops of AI compute for both training and inference workloads in an affordable, compact 5U system.
For Innovators making new AI Breakthroughs
Fast & efficient for dense matmul models
Excels at sparse & fine-grained computation
Expert support to get you up and running quickly
IPU-POD16 opens up a new world of machine intelligence innovation.
IPU-POD16 DA is designed to get you up and running in no time. A turnkey system, featuring IPU-M2000s directly attached to an approved host server ready for installation in your datacenter. Extensive documentation and support is provided both by AI experts at Graphcore and our elite partner network.
IPU-POD16 DA is a powerful, standalone AI compute resource. However, it also offers the opportunity for growth, on your terms. Your IPU-POD16 DA system investment can be expanded later into a larger IPU-POD system
Designed specifically for the communication requirements of AI workloads at scale, IPU-Fabric is Graphcore’s innovative low-latency, jitter-free interconnect using industry standard IT equipment. It supports highly efficient, deterministic, all-to-all IPU interconnect across your system regardless of size.
Plug-and Play with Direct Attach
AI infrastructure built to scale
Start small, scale big
Graphcore
IPU-POD
64
The IPU-POD64 is a single rack configuration featuring 16 IPU-M2000™ compute blades, based on the innovative GC200 Intelligence Processing Unit (IPU). The IPU-POD64 can deliver up to 16 petaFLOPS of AI compute.
The Building Block for Machine Intelligence Scale-Out
The whole system, hardware and software, has been architected together. IPU-POD64 supports standard frameworks and protocols to enable smooth integration into existing data center environments. Innovators can focus on deploying their AI workloads at scale, using familiar tools while benefitting from cutting-edge performance.