TSMC
TSMC (Taiwan Semiconductor Manufacturing Company) is the world's largest dedicated semiconductor foundry, headquartered in Hsinchu, Taiwan. The company focuses on providing integrated circuit manufacturing services for customers and does not independently design or sell chip products. TSMC commands leading-edge process technology and serves major chip design companies globally, occupying a central position in the global semiconductor supply chain.
Core Facts
AMD's 3D V-Cache technology vertically stacks additional SRAM cache chips above the CCD, expanding the L3 cache from 32MB to 96MB or more.
3D V-Cache is based on TSMC's SoIC advanced packaging technology, using a Hybrid Bonding process to stack additional SRAM cache chips on top of the processor's CCD.
Timeline (last 90 days)
The chiplet architecture, which splits an SoC into multiple small chips, manufactures each using the most suitable process, and then interconnects them through advanced packaging (such as TSMC CoWoS and Intel EMIB), is the current mainstream design paradigm for high-performance processors.
The main data center semiconductor players listed by DMR include Intel, NVIDIA, AMD, Broadcom, Qualcomm, Samsung Electronics, Micron, SK hynix, TSMC, and Marvell Technology.
The GB202 is manufactured using TSMC's 4N process and contains 92 billion transistors.
The B70 is based on the XE2 Battlemage architecture (TSMC N5 process), featuring 32 XE cores, 256 XMX engines, 608 GB/s bandwidth, and approximately 23 TFLOPS FP32.
NVIDIA GPU efficiency improvements are driven jointly by more advanced process nodes from TSMC/Samsung (from 7nm to 4nm) and chip architecture optimizations.
The Snapdragon X2 Elite is manufactured on TSMC's 4nm process and adopts a 12-core Oryon CPU architecture
TSMC's process roadmap advances from N5 (5nm) and N3 (3nm) to N2 (2nm), with A14 being the next generation of advanced process after N2.
TSMC's A14 process node is a 1.4-nanometer-class process, with the '14' in its name corresponding to 14 Angstroms, approximately equal to 1.4 nanometers.
TSMC is gradually transitioning from the traditional 'nanometer' naming convention to a naming system using 'A' (Angstrom) as the unit.
TSMC began introducing nanosheet technology at the N2 node, referring to it as 'Nanosheet FET' (Nanosheet field-effect transistor).
9 more timeline events
All Facts (20)
AMD's 3D V-Cache technology vertically stacks additional SRAM cache chips above the CCD, expanding the L3 cache from 32MB to 96MB or more.
70%Verified3D V-Cache is based on TSMC's SoIC advanced packaging technology, using a Hybrid Bonding process to stack additional SRAM cache chips on top of the processor's CCD.
65%UnverifiedThe Snapdragon 865 is manufactured on TSMC's 7nm process, featuring one 2.84GHz Cortex-A77 prime core, three 2.42GHz mid cores, and four 1.8GHz efficiency cores
90%UnverifiedArrow Lake是Intel第一代桌面分离式模块化架构，采用Intel 20A和台积电N3B等多种工艺节点混合封装
90%UnverifiedArrow Lake处理器的计算核心采用台积电3nm工艺制造
80%UnverifiedTSMC has raised foundry prices multiple times, with cumulative increases exceeding 20%
75%Unverified台积电先进制程（3nm/5nm）每片晶圆成本已超过2万美元，较28nm时代翻了数倍
75%UnverifiedGaming console core processors (APUs) are typically designed by AMD and manufactured by TSMC
60%UnverifiedThe Xbox Series X uses a custom AMD Zen 2 + RDNA 2 APU based on TSMC's 7nm process
60%UnverifiedThe chiplet architecture, which splits an SoC into multiple small chips, manufactures each using the most suitable process, and then interconnects them through advanced packaging (such as TSMC CoWoS and Intel EMIB), is the current mainstream design paradigm for high-performance processors.
50%UnverifiedThe main data center semiconductor players listed by DMR include Intel, NVIDIA, AMD, Broadcom, Qualcomm, Samsung Electronics, Micron, SK hynix, TSMC, and Marvell Technology.
50%UnverifiedThe GB202 is manufactured using TSMC's 4N process and contains 92 billion transistors.
50%UnverifiedThe B70 is based on the XE2 Battlemage architecture (TSMC N5 process), featuring 32 XE cores, 256 XMX engines, 608 GB/s bandwidth, and approximately 23 TFLOPS FP32.
50%UnverifiedNVIDIA GPU efficiency improvements are driven jointly by more advanced process nodes from TSMC/Samsung (from 7nm to 4nm) and chip architecture optimizations.
50%UnverifiedThe Snapdragon X2 Elite is manufactured on TSMC's 4nm process and adopts a 12-core Oryon CPU architecture
50%UnverifiedTSMC began introducing nanosheet technology at the N2 node, referring to it as 'Nanosheet FET' (Nanosheet field-effect transistor).
50%UnverifiedTSMC is gradually transitioning from the traditional 'nanometer' naming convention to a naming system using 'A' (Angstrom) as the unit.
50%UnverifiedTSMC's A14 process node is a 1.4-nanometer-class process, with the '14' in its name corresponding to 14 Angstroms, approximately equal to 1.4 nanometers.
50%UnverifiedTSMC has already procured High-NA EUV equipment for research and development.
50%UnverifiedTSMC's process roadmap advances from N5 (5nm) and N3 (3nm) to N2 (2nm), with A14 being the next generation of advanced process after N2.
50%