June 14 (Bloomberg) -- According to Reuters, theAMD is working closely with a number of AI startups with the intention of enhancing its own software capabilities and designing more advancedchip.. As more and more AI companies seek toNvidiaAn alternative to chips, AMD began expanding its footprint with plans to build more competitive hardware and the acquisition of server maker ZT Systems.
But a prerequisite for a successful chip product launch is a strong software ecosystem to run AI developers' programs more efficiently.AMD has acquired several small software companies in recent weeks to bolster its technical talent pipeline, andAccelerated optimization of a software platform called ROCm. said Vamsi Bopanna, senior vice president of AMD AI, "This is a well-planned, long-term, cross-generation program for us."
AMD's commitment to continually optimizing software like ROCm is especially beneficial to customers like AI startup Cohere, who get software updates and new features faster.
Cohere focuses on building customized AI models for large enterprises, rather than basic models like those OpenAI focuses on. The company's CEO, Aidan Gomez, said in an interview thatAMD's Software Improvements Dramatically Increase EfficiencyCohere's software migration to the AMD chip used to take weeks. Whereas it used to take weeks for Cohere to migrate software to AMD chips, it now takes just a few days.
Gomez didn't disclose exactly what percentage of Cohere software uses AMD chips, but he said it's part of theSignificant share of the global computing base.
AMD executive vice president Forrest Norrod says OpenAI has had a significant impact on the design of the upcoming MI450 series of AI chips. According to IT House, AMD's MI400 series chips will be used in the core architecture of the new server Helios, which will be released in 2024.
NVIDIA has launched a similar server product, as AI computing requires hundreds or even thousands of chips to work together.
On Thursday, OpenAI CEO Aultman appeared at an AMD event in San Jose to briefly discuss the partnership.
Norrod said OpenAI's needs profoundly influenced the design of the memory architecture of the MI450 series of chips and how to meet the need for large-scale chip scaling for high-performance AI applications. "The feedback we received from OpenAI largely determined the final design of the chips."