The race for extra highly effective and environment friendly AI {hardware} surged forward this week with Intel and Google asserting new chips to assist them turn into much less reliant on NVIDIA tech.
It looks like each week new AI fashions are being launched. Behind every launch are weeks of coaching on cloud computing knowledge facilities, most of that are powered by NVIDIA GPUs.
Intel and Google each introduced new in-house AI chips that may prepare and deploy massive AI fashions sooner whereas utilizing much less energy.
Intel’s Gaudi 3 AI accelerator chip
Intel might be higher recognized for the chips that energy your PC however on Tuesday the corporate introduced its new AI chip known as Gaudi 3.
NVIDIA’s H100 GPUs have shaped the majority of AI knowledge facilities, however Intel says Gaudi 3 delivers “50% on average better inference and 40% on average better power efficiency than Nvidia H100 – at a fraction of the cost.”
An enormous contributor to the facility effectivity of Gaudi 3 is that Intel used Taiwan Semiconductor Manufacturing Co’s 5nm course of to make the chips.
Intel didn’t give any pricing info however when requested the way it compares with NVIDIA’s merchandise, Das Kamhout, VP of Xeon software program at Intel stated, “We do expect it to be highly competitive.”
Dell Applied sciences, Hewlett Packard Enterprise, Lenovo, and Supermicro would be the first to deploy Gaudi 3 of their AI knowledge facilities.
Intel CEO Pat Gelsinger summarized the corporate’s AI ambitions saying “Intel is bringing AI everywhere across the enterprise, from the PC to the data center to the edge.”
The #IntelGaudi 3 #AI accelerator affords a extremely aggressive different to NVIDIA’s H100 with increased efficiency, elevated scalability, and PyTorch integration. Discover extra key product advantages. https://t.co/sXdQKjYFw0 pic.twitter.com/iJSndBQkvT
— Intel AI (@IntelAI) April 9, 2024
Google’s Arm and TPU upgrades
On Tuesday Google introduced its first customized Arm-based CPUs it plans to make use of to energy its knowledge facilities. The brand new chip, dubbed Axion, is a direct competitor to Intel and AMD’s CPUs.
Google claims Axion delivers “30% better performance than the fastest general-purpose Arm-based instances available in the cloud today, up to 50% better performance and up to 60% better energy-efficiency than comparable current-generation x86-based instances.”
Google has been shifting a number of of its companies like YouTube and Google Earth to its present technology Arm-based servers which can quickly be upgraded with the Axion chips.
Having a robust Arm-based choice makes it simpler for purchasers emigrate their CPU-based AI coaching, inferencing, and different purposes to Google’s cloud platform with out having to revamp them.
For giant-scale mannequin coaching, Google has largely relied on its TPU chips as an alternative choice to NVIDIA’s {hardware}. These will even be upgraded with a single new TPU v5p now containing greater than double the variety of chips within the present TPU v4 pod.
TPU v5p, our strongest and scalable TPU, is now typically obtainable! #GoogleCloudNext pic.twitter.com/mmfWlzHeqs
— Google Cloud Tech (@GoogleCloudTech) April 9, 2024
Google isn’t seeking to promote both its new Arm chips or TPUs. The corporate is seeking to drive its cloud computing companies quite than changing into a direct {hardware} competitor of NVIDIA.
The upgraded TPUs will present a lift to Google’s AI Hypercomputer which permits large-scale AI mannequin coaching. The AI Hypercomputer additionally makes use of NVIDIA H100 GPUs that Google says will quickly get replaced with NVIDIA’s new Blackwell GPUs.
The demand for AI chips isn’t prone to decelerate anytime quickly, and it’s wanting much less just like the one-horse NVIDIA race than earlier than.
