Intel Partners to Enter the Generative AI Race—Aurora genAI
By: Craig Kennedy
Intel Partners to Enter the Generative AI Race—Aurora genAI
On Monday May 21st, at the ISC High Performance Computing Conference 2023 in Hamburg, Germany, Intel announced Aurora genAI, a series of science focused generative AI models that will run on the Intel Aurora supercomputer.
Aurora GenAI Focuses on Science
Aurora GenAI is a collaborative project between Intel, HPE, and Argonne National Laboratory to develop a suite of generative AI models for the scientific research community. These models, containing up to 1 trillion parameters, will be used across various scientific disciplines from cancer and disease research to the design of molecules and materials and much more.
It will be trained on a wealth of data from general text, code, scientific texts, as well as structured scientific data across the fields of chemistry, biology, material science, physics, and medicine.
GenAI Powered by Aurora
At the same conference, Argonne National Laboratory and Intel brought us up to speed on the status of Aurora, announcing the full specifications and some early performance results. This 2+ exaflop supercomputer contains 21,248 CPUs, 63,744 GPUs and 1,024 DAOS nodes, all connected using HPE Slingshot high-performance Ethernet interconnects.
Early performance results are showing up to a 2x performance boost compared to AMD M1250 GPUs and a 20% performance boost compared to NVIDIA H100 GPUs. Testing is also showing near linear scaling up to hundreds of nodes. These are promising results however Aurora contains 10,624 nodes so scalability testing still has a way to go.
Generative models tuned for science and trained on an extensive data source of scientific data have the potential to accelerate discoveries in healthcare, material science, and a wealth of other scientific research areas. This project has the potential to make a major impact on the scientific community, providing insights into many areas that could directly benefit humanity.
Bottom Line
Generative models tuned for science and trained on an extensive data source of scientific data have the potential to accelerate discoveries in healthcare, material science, and a wealth of other scientific research areas. This project has the potential to make a major impact on the scientific community, providing insights into many areas that could directly benefit humanity.
2 WEEKS TO GO—JOIN US FOR TRANSFORM TOUR 2023!
Thursday, June 15th | 10 AM PT \ 1 PM ET
Kickoff the Summer with Aragon Research’s next #TransformTour23 stop!
Since the launch of OpenAI’s GPT service, organizations have been scrambling to put out their own copilot service. In our June Transform Tour, taking place on Thursday, June 15, 2023, we will discuss how Copilots work and how you can define a strategy to support digital labor.
This blog is a part of the Digital Operations blog series by Aragon Research’s Sr. Director of Research, Craig Kennedy.
Missed an installment? Catch up here!
Have a Comment on this?