May 22 (Reuters) – Intel (INTC.O) said on Monday that it would shift its strategy to take on Nvidia (NVDA.O), revealing a number of artificial intelligence (AI) computing chips it plans to introduce in 2025. revealed some new details. ) and Advanced Micro Devices, Inc. (AMD.O).
Intel announced at its supercomputing conference in Germany on Monday that its upcoming Falcon Shores chips will have 288 gigabytes of memory and support 8-bit floating point arithmetic. These technical specs are important as artificial intelligence models like services like ChatGPT are exploding in size and companies want more powerful chips to run them.
The details are quickly emerging as Intel implements a strategy shift to catch up with AI chip market leader Nvidia and AMD, which is expected to challenge Nvidia’s position with a chip called the MI300.
In contrast, Intel has virtually no market share since a chip called the Ponte Vecchio that would be Nvidia’s competitor suffered years of delays.
Intel announced on Monday that the Argonne National Laboratory’s Aurora supercomputer, based on Ponte Vecchio, is nearly complete. Intel claims the supercomputer will outperform Nvidia’s latest AI chip, the H100.
However, Intel’s Falcon Shores successor chip won’t hit the market until 2025, by which time Nvidia will likely launch another chip of its own.
Geoff McVay, interim head of Intel’s accelerated computing systems and graphics group, said the company had a traditional strategy of combining graphics processing units (GPUs) and central processing units (CPUs). said it has been spending time reworking chips since abandoning it.
“We aim to offer the best CPU and best GPU on the market, but it’s hard to say that one vendor at a time can offer the best combination of these,” McVeigh told Reuters. Told. “If there are separate products, the platform will allow him to choose both ratios and vendors at the level.”
Reported by Stephen Nellis of San Francisco.Editing: Bill Barclot
Our standards: Thomson Reuters Trust Principles.
