Latest Intel AI Platforms in Intel® Developer Cloud beta

Aug. 14, 2023—To help developers, customers, and companies accelerate multiarchitecture programming with access to the latest advanced Intel architectures, we announced the Intel® Developer Cloud at Intel Innovation last year and invited select customers to participate in a private beta trial. We have been hard at work extending exciting new capabilities in this environment and shifted to a public beta. I’d like to share insight into the progress we’ve made.

Our goal remains clear. We want to make new and future hardware platforms available for pre-launch development and testing to give developers, customers, and researchers early and efficient access to Intel® technologies from a few months up to a full year ahead of product availability. This developer cloud will be a one-stop shop for developers and ecosystem collaborators from around the world to get virtual access to our latest and greatest silicon. Production is progressing and we look forward to enabling general availability to our developer cloud soon.

New capabilities include:

Making new architectures available quicker & easier: Last year we unveiled early access to the 4^th Gen Intel® Xeon® Scalable processor with its advanced AI capabilities for our early-innovator customers and partners. Since 4^th Gen Xeon’s general availability earlier this year, we now make it available for both virtual and bare metal access. And I am particularly excited to make Intel’s highest-performing, highest-density, general-purpose discrete GPU - Intel® Data Center GPU Max Series 1100 and 1550, Intel® Xeon® CPU Max Series - the only x86-based processor with high-bandwidth memory (HBM), and Habana® Gaudi2® Processor for Deep Learning, available to test. We are also preparing access to our next generation CPU codenamed Emerald Rapids soon. Over the next months, we’ll ramp up capacity for all these platforms to support a long waiting list of customers.

Access to the 4^th gen Xeon processors is available with virtual machine for developers to test drive and develop new AI and oneAPI-optimized applications. The Max Series 1100 GPU is also available using batch access for various test use cases, as well as for oneAPI training. For select customers, we provide access to Habana Gaudi2 processors for multi-billion parameter large language models (LLMs) such as Falcon, Dolly, WizardLM, and LLaMa2 for training, fine-tuning and inference performance testing.

We are also working with cutting-edge tech startups in the GenAI space through the Intel® Liftoff program to enable new features and help optimize their AI-powered applications through development made possible on our dev cloud

Sneak previews at hackathons, oneAPI DevSummit for AI & easy-to-consume training: While the developer cloud is still in beta, we’re doing live testing at events to ensure we’re getting things right before opening the doors widely. Be a part of our previews in developer hackathons, workshops, and the oneAPI DevSummit for AI on Aug. 21-22. We made oneAPI onboarding training available with just a click on the button in the cloud. This allows developers to focus on important things such as learning and building applications and not worry about how to connect hardware and software pieces together. I am immensely proud of the work by our engineers to make this super simple.

So far, we’ve worked with 35 AI innovator startups and AI researchers at 20 universities, along with many enterprise customers for testing. Students gained their first experience building real-world high-performance computing (HPC) and AI applications using the latest and greatest hardware. We had an enormous response to this training. Within 1 week of a limited launch, we trained more than 1,250 developers. We recently completed our first-ever distributed LLM fine-tuning and inference hackathon with our collaborators and customers.

We are working closely to develop patterns and capabilities using Intel® Neural Compressor and Intel® Extension for Transformers to deploy and accelerate AI using 4^th gen Xeon key features like Intel® Advanced Matrix Instructions (Intel® AMX).

What to expect next:

A new & improved user interface (UI): We listened to feedback from early customers to improve consistency and accessibility of features and streamlined workflows, and moved toward a more modern interface with an improved user experience. We are making continuous progress as we get closer to offering production-level general availability. I cannot wait to get this in your hands, and I welcome your comments and feedback.

Open ecosystem focus: The oneAPI open ecosystem provides a solid foundation with a single programming environment for platforms including CPUs and accelerators across multiple vendors’ products. This model provides significant long-term productivity benefits for developers through the use of a single codebase and reduced code maintenance without compromising performance. Intel’s oneAPI tools contain optimized compilers, oneAPI libraries, DL frameworks, and analysis tools purpose-built for developers who need accelerated AI and compute. Intel Developer Cloud now provides easier access to launch the Intel® oneAPI Base Toolkit and libraries so users can test and build high-performance, multiarchitecture applications on Intel Xeon and Intel Max GPUs using best-in-class compilers.

LLM Trainings & Workshops coming soon: We’ll be enabling capabilities for running parallel-distributed LLM fine-tuning models. Stay tuned for details on how to register and participate in these.

In addition to the previews at the oneAPI DevSummit for AI, I look forward to seeing you at Intel Innovation on Sept. 19-20 where we’ll share more updates. Here is the link to register for Intel Innovation if you have not already reserved your seat at this exciting tech preview event.

Harmen