INTRODUCING IP‌‍U-MA‍C‍HINE A‌ND‌ IP‌U-P‌OD‌ SYSTEMS

Second generation IPU systems for AI infrastructure at scale

IPU-Machine: M2000

Our second generation, core building block for AI infrastructure, for all your AI workloads. Available standalone to get started and with built-in networking for supercomputing scale, the IPU-M2000 packs 1 PetaFlop of AI compute and 450GB Exchange-Memory. Available to pre-order today.

Learn More

IPU-POD₆₄

Building blocks and reference designs for AI supercomputing scale that every enterprise can benefit from. Graphcore IPU-POD₆₄ incorporates transformative innovations in compute and networking. Available to test today in our private Graphcloud™ preview.

Speak to an expert

IPU-Server systems

Industry standard OEM server systems for training and inference, based on Graphcore dual-IPU C2 PCIe Cards with MK1 Colossus™ IPUs. Get started with development and experimentation then ramp to full scale production. Available on Microsoft Azure and to buy direct from Dell, Inspur and our partners.

Buy now

BENCHMARKS

State of the Art with today’s large, complex models

Natural Language Processing - BERT

The IPU delivers over 25% faster time-to-train with the BERT language model, training BERT-base in 36.3 hours with seven C2 IPU-Processor PCIe cards, each with two IPUs, in an IPU Server system. For BERT inference, we see more than 2x higher throughput at the lowest latency resulting in unprecedented speedups.

Computer Vision - EfficientNet

The Graphcore C2 IPU-Processor PCIe card achieves 15x higher throughput and 14x lower latency compared to a leading alternative processor. High throughput at the lowest possible latency is key in many of the important use cases today such as visual search engines and medical imaging.

More benchmarks

Our Customers

“Graphcore’s technology is being evaluated by J.P. Morgan to see if its solutions can accelerate the bank’s advances in AI, specifically in Natural Language Processing and speech recognition.”