AMD is getting personal with chips as it sets sail to make products more to its customers’ tastes.
The chipmaker has detailed a modular chip future where customers can mix and match non-AMD processors in a custom chip package.
“We’re working to make it easier to implement chips with more flexibility,” AMD chief technology officer Mark Papermaster said at the analyst day meeting late last week.
AMD will allow customers to implement multiple dies – also known as chiplets or compute tiles – in a tight chip package. AMD already uses tiles, but now welcomes third parties to make accelerators or other chips to include in 2D or 3D packages alongside its x86 CPUs and GPUs.
AMD has been customizing chips for many years, especially for game consoles made by Sony and Microsoft, but mostly using in-house technology.
Customers want flexibility to meet specific organizational requirements and are looking to integrate accelerators into chips for applications that include AI and automotive, AMD executives said at the show.
“We’re going to make it easier to add third-party IPs as well as customer IPs to this chip platform,” AMD CEO Lisa Su said at the conference.
AMD was to expand its chip platform for technologies acquired from Xilinx, which makes FPGAs and AI accelerators, and Pensando, which makes networking chips.
“So far, we’ve had a lot of positive customer engagement when you think of hyperscalers, when you think of 5G in automobiles,” Su said, adding, “These are great opportunities where people want personalize and we want to be their partner. of choice.”
Most of AMD’s chips, including its flagship PC and server chips, are built with the x86 instruction set architecture, while Pensando and Xilinx are primarily Arm-based. Xilinx and AMD are also members of the RISC-V open source foundation, which has created an open source chip architecture to compete with x86 and ARM.
“We are happy to be independent from the ISA here,” Su said.
Intel and Nvidia also shared their chip plans. Intel has opened up its x86 architecture for licensing and has a chip strategy that allows customers to bundle Arm and RISC-V cores into a single package.
Intel’s goal is to attract more customers to its Foveros packaging, which will bring more business to its factories. Intel is investing billions to build new factories using its latest nodes.
“I think for a company the size of AMD it’s really about designing and optimizing their own chips to give them a manufacturing edge, but I’m not sure they really need to. to support those other standards or use third-party chips, they can get better things by designing themselves,” said Linley Gwennap, principal analyst at TechInsights.
AMD bases its chip strategy on CoWoS (Chip-on-Wafer-on-Substrate) packaging technologies from TSMC, also backed by Nvidia and Apple. Nvidia invites third parties to develop cores that connect to its CPUs and GPUs using the proprietary NVLink interconnect.
Chip strategy is more relevant to manufacturing, with Intel offering foundry services around its chip strategy. AMD and Nvidia are representatives of TSMC in its manufacturing competition with Intel, Gwennap said.
AMD’s custom chip strategy will revolve around the new Infinity 4.0 architecture, which is the interconnection of dies within the chip package. The exclusive Infinity fabric will be compatible with the CXL 2.0 interconnect.
The Infinity Interconnect will also support UCIe (Universal Chiplet Interconnect Express) to link chiplets in a package. UCIe is already supported by Intel, AMD, Arm, Google, Meta and others.
“The implementation of the fourth-generation AMD Infinity Architecture also enables unified, consistent and shared memory between host and external devices,” Papermaster said.
AMD’s server chips are already used by meta and cloud providers including Amazon, Microsoft and Google. Meta builds its metaverse infrastructure with AMD Epyc chips, and all major cloud providers offer virtual machines with AMD chips.