The AMD Infinity Platform with eight Instinct MI300X GPUs.

Here's how the designs are different across fourth-gen EPYC chips for general purpose (Genoa), cloud computing (Bergamo) and technical computing (Genoa-X).

Brad McCredie, AMD's vice president of data center and accelerated processing, explains the building blocks of the Instinct MI300X GPU in response to a question from @CDemerjian.

My take on AMD's biggest announcements from today:

AMD Eyes AI, Cloud Expansion With Instinct MI300X, EPYC 97X4 Chips via @CRN

AMD revealed the GPU-only Instinct MI300X, which has 192GB of HBM3 memory, beating the 80GB capacity of Nvidia's H100. This will allow users to run large language models on fewer GPUs, giving the MI300X leadership total cost of ownership, Su said.

AMD announced Genoa-X, the fourth-gen EPYC processors that use 3D V-Cache tech to up the L3 cache to 1.1GB. Microsoft Azure is supporting the chips with new HBv4 and HX cloud instances for technical computing.

Here it is: fourth-gen EPYC Bergamo for cloud-native applications. Up to 128 Zen 4c cores, each of which are 35% smaller than Zen 4 cores. Aimed at leadership performance-per-watt versus the performance-per-core focus of general-purpose EPYC Genoa.

AWS EC2 Vice President Dave Brown said new cloud instances coming with EPYC Genoa and its Nitro system. AMD is trying to compete with AWS Nitro with its Pensando DPUs.

AMD EPYC is now powering more than 640 cloud instances globally, Su said. Another 200 on track to launch by the end of the year.

Lisa Su is here, confirming news is coming about new EPYC processors and Instinct chips.

Here at AMD's Data Center and AI Technology Premiere, where the company is expected to talk about new data center and AI chips.

It should come as no surprise, but all the solution providers I talked to for this @CRN cover story about AI opportunities for channel partners — AKA the companies selling IT services to much of the corporate world — said everything revolves around Nvidia.

Hey, it's me (not @dylan522p).

First in-person semiconductor event in more than three years for me!

Intel Earnings Get Worse Before ‘Modest Recovery’ Expected In Late 2023 via @CRN

Look at where the former head of Intel Foundry Services landed.

Three weeks after AMD lost one channel executive, the chip designer is losing his boss, North America Channel Chief Terry Richardson, who gave AMD a lot of cachet with the value-added reseller community when he joined two years ago: #amd via @CRN

Eric McLaughlin, general manager of Intel's Wireless Solutions Group, told me that his division is "profitably selling software" now and that it has bigger ambitions in the commercial software space.

Intel Arms Partners With Rich Network Insight By Turning PCs Into Sensors via @CRN

ICYMI: The head of AMD's research group is leaving and will be replaced by former Xilinx CTO Ivo Bolsens: I highlighted a few other exec moves that happened recently at AMD, including Jack Huynh becoming GM of the Computing and Graphics group. via @CRN

Late Friday AMD news: Head Of AMD’s Influential Research Group Leaves Amid Other Exec Changes #amd via @CRN