skip to Main Content

Graphcore IPU Lets Innovators Make New Breakthroughs in Machine Intelligence

State of the art performance for natural language processing, computer vision and much more

Best for Natural Language Processing

The IPU delivers impressive performance for NLP. IPU-POD64 trains BERT-Large over 2.5 times faster than comparable DGX A100 platforms, cutting hours from AI development cycles.


Best for Computer Vision

The IPU-M2000 delivers a significant performance advantage compared with the Nvidia A100 GPU. Running EfficientNet on the IPU is straightforward and doesn’t require extra INT8 quantisation effort which can also affect accuracy.


Graphcore Colossus MK2 GC200 IPU

Introducing our second-generation Colossus™ MK2 IPU processor – the GC200. The IPU is a completely new kind of massively parallel processor, co-designed from the ground up with the Poplar® SDK, to accelerate machine intelligence. Since our first generation Colossus IPU, we have developed groundbreaking advances in compute, communication and memory in our silicon and systems architecture, achieving an 8x step up in real-world performance compared to the MK1 IPU. The GC200 is the world’s most complex processor made easy to use thanks to Poplar software, so innovators can make AI breakthroughs.

With 59.4B transistors, and built using the very latest TSMC 7nm process, the Colossus MK2 GC200 IPU is the world’s most sophisticated processor. Each MK2 IPU has 1472 powerful processor cores, running nearly 9,000 independent parallel program threads. Each IPU holds an unprecedented 900MB In-Processor-Memory™ with 250 teraFLOPS of AI compute at FP16.16 and FP16.SR (stochastic rounding). The GC200 supports much more FP32 compute than any other processor

Screenshot 2021-05-25 123833

IPU Platforms

The IPU-M2000 is the fundamental compute engine for IPU-based machine intelligence, built with the powerful Colossus MK2 IPU designed from the ground up for AI. It packs 1 petaFLOP of AI compute with up to 526GB Exchange-Memory™ in a slim 1U blade. The IPU-M2000 has a flexible, modular design, so you can start with one and scale out to many in our IPU-POD platforms.

Directly connect one IPU-M2000 to an AgilityEdge host server, to build the entry level IPU-POD4 to start your IPU journey.

IPU-POD16 opens up a new world of machine intelligence innovation. Ideal for exploration and experimentation, the IPU-POD16 is the perfect new tool to develop concepts and pilots consolidating both training and inference in one affordable system. With a powerful 4 petaFLOPS of AI Compute in a compact 5U footprint, it offers unprecedented performance and flexibility.

IPU-POD16 is your easy-to-use starting point for building better, more innovative AI solutions with IPUs.

IPU-POD64 delivers ultimate flexibility to maximize all available space and power in your datacenter, no matter how it is provisioned. 16 petaFLOPS of AI-compute for both training and inference to develop and deploy on the same powerful system.



Back To Top