Latest
Quantum-2 Versus Ethernet: Tradeoffs For Large-...
Summary AI buildouts are reallocating spend from general-purpose servers to accelerators and, next, to networking. Mapping which fabrics hyperscalers deploy is now strategic. Within this shift, two contests are unfolding:...
Quantum-2 Versus Ethernet: Tradeoffs For Large-...
Summary AI buildouts are reallocating spend from general-purpose servers to accelerators and, next, to networking. Mapping which fabrics hyperscalers deploy is now strategic. Within this shift, two contests are unfolding:...
Revisiting Semi-Custom: Meta’s Path To AI And V...
Overview Meta announced facility shifts toward higher-power, liquid-cooled datacenters alongside updates on internal silicon programs for video processing and AI acceleration. As with other hyperscalers, public disclosures emphasize earlier generations...
Revisiting Semi-Custom: Meta’s Path To AI And V...
Overview Meta announced facility shifts toward higher-power, liquid-cooled datacenters alongside updates on internal silicon programs for video processing and AI acceleration. As with other hyperscalers, public disclosures emphasize earlier generations...
Evaluating AmpereOne 192-Core Chiplet Design An...
Overview Ampere Computing targets cloud datacenters with high-core-count Arm servers and has landed deployments across major cloud providers and automotive. Industry estimates placed 2022 shipments above 300,000 CPUs. The firm’s...
Evaluating AmpereOne 192-Core Chiplet Design An...
Overview Ampere Computing targets cloud datacenters with high-core-count Arm servers and has landed deployments across major cloud providers and automotive. Industry estimates placed 2022 shipments above 300,000 CPUs. The firm’s...
Sapphire Rapids Redesign: Architecture and Plat...
Overview Emerald Rapids (EMR), the 5th-Gen Xeon Scalable, reverts from four chiplets to two large dies in its highest core-count configuration. Counter to recent disaggregation trends, this reaggregation targets performance...
Sapphire Rapids Redesign: Architecture and Plat...
Overview Emerald Rapids (EMR), the 5th-Gen Xeon Scalable, reverts from four chiplets to two large dies in its highest core-count configuration. Counter to recent disaggregation trends, this reaggregation targets performance...
Automotive Demand Weakness and TSMC Capacity Pl...
Earnings Overview Revenue for Q1 2023 was $16.72 billion, down 4.8% year on year. Q2 guidance implies $15.6 billion at the midpoint, a 14.6% year-on-year decline. Full-year guidance points to...
Automotive Demand Weakness and TSMC Capacity Pl...
Earnings Overview Revenue for Q1 2023 was $16.72 billion, down 4.8% year on year. Q2 guidance implies $15.6 billion at the midpoint, a 14.6% year-on-year decline. Full-year guidance points to...
Systems Matter More Than Microarchitecture in G...
Executive Summary AI-era cost structures push capital and operating spend toward infrastructure. Microarchitecture matters, but system architecture, deployment flexibility, and networking dominate total cost of ownership (TCO). Google’s TPU platforms...
Systems Matter More Than Microarchitecture in G...
Executive Summary AI-era cost structures push capital and operating spend toward infrastructure. Microarchitecture matters, but system architecture, deployment flexibility, and networking dominate total cost of ownership (TCO). Google’s TPU platforms...