siliconandsystems.com/en/order
ARTICLE ORDER · NEWEST FIRST

Article order

An operator view arranged from the most recently added article.

#001[memory]HBM Saves Clock Power by Rebuilding Four Phases at the Data GroupsDIGEST · 24m#002[silicon]What Commercial AI EDA Has Actually Put Through TapeoutORIGINAL · 21m#003[silicon]When AI Placement Becomes Production SiliconORIGINAL · 21m#004[arch]RTL Automation Starts After the First DraftORIGINAL · 15m#005[arch]AI Can Search a Microarchitecture, but It Still Needs an Executable ContractORIGINAL · 15m#006[silicon]Why AI Reached Industrial Physical Design Before ArchitectureORIGINAL · 15m#007[silicon]Place and Route Is Becoming an Objective NegotiationORIGINAL · 14m#008[silicon]The Next EDA Advantage Is a Recipe That TransfersORIGINAL · 15m#009[fabric]Pooling PCIe Devices Without a PCIe SwitchDIGEST · 14m#010[ai-infra]Alibaba Rebuilt Cloud RDMA Across Host, PCIe, and FabricDIGEST · 15m#011[fabric]An RDMA NIC Needs a Scheduler, Not Just Faster QueuesDIGEST · 14m#012[ai-infra]Why Elastic Tensors Help PCIe GPU Servers MoreDIGEST · 13m#013[memory]RDMA Remote Memory Without a Memory-Node CPUDIGEST · 15m#014[fabric]Routable PCIe Is a Fabric, Not a Longer BusDIGEST · 12m#015[arch]Making a SmartNIC Fast Across a Slow PCIe BoundaryDIGEST · 13m#016[fabric]The Protocol Debt Inside Hyperscale RoCEDIGEST · 14m#017[silicon]The Soft Materials Set the Shape of the Advanced PackageDIGEST · 16m#018[silicon]Evatec Uses Film Stress to Flatten the Packaging WaferDIGEST · 16m#019[silicon]IBM Moves the Heat Spreader into the Organic SubstrateDIGEST · 16m#020[fabric]Count the Laser: Co-Designing a 224 Gb/s Coherent CPO LinkDIGEST · 18m#021[fabric]CPO Has to Survive Reflow: Intel's Fiber-to-EMIB PackageDIGEST · 18m#022[fabric]The 400 µm Modulator That Pushes CPO to 212 Gb/s per WavelengthDIGEST · 19m#023[memory]No Giant Switch: How Octopus Builds a 96-Server CXL Memory PodDIGEST · 13m#024[memory]A Real CXL Memory Box Under SAP HANA: What Can Move, and What Must Stay LocalDIGEST · 13m#025[fabric]Pool the Buffer, Not the Device: A CXL Shortcut Around PCIe SwitchesDIGEST · 13m#026[memory]Eighty Percent Far, Eighty-Five Percent Fast: A Database Placement Rule for CXLDIGEST · 12m#027[fabric]One Photonic Switch, Two Routing DimensionsDIGEST · 14m#028[silicon]At 100 GHz, the Laser Package Becomes Part of the CircuitDIGEST · 14m#029[fabric]Let the Optical Topology Follow the GPU AllocationDIGEST · 14m#030[ai-infra]The GPU Did Not Crash, Yet 1,024 Workers Slowed DownDIGEST · 19m#031[ai-infra]When Idle Models Leave the GPU: Paying Only for Active InferenceDIGEST · 19m#032[fabric]One DPU, Eight RNICs: Tencent's Pegasus Network for Bare-Metal AI CloudDIGEST · 15m#033[silicon]Three Nanosheets over Three: Samsung Builds Logic Upward at 42 nmDIGEST · 14m#034[fabric]Compile the Fabric before Deploying It: Meta's Matryoshka Network Design SystemDIGEST · 15m#035[silicon]CFET Leaves the Device Lab: TSMC Runs an Oscillator and SRAM below 48 nmDIGEST · 14m#036[fabric]One RoCE Fabric, Four Distance Classes: Meta's Communication Stack for 100K+ GPUsDIGEST · 15m#037[fabric]UCIe Crosses the Rack in Light: Ayar Labs' 8.192 Tb/s Optical RetimerDIGEST · 15m#038[fabric]Optics Moves Inside the Silicon: Celestial AI's Photonic Fabric ModuleDIGEST · 15m#039[fabric]A 4,000 mm² Optical Backplane: Reading Lightmatter's Passage M1000DIGEST · 16m#040[silicon]A 600 mm² Mainframe Die Becomes a System: Inside IBM Telum IIDIGEST · 16m#041[ai-infra]The Same H100 Is Not the Same Rental: Measuring the GPU Cloud LotteryDIGEST · 18m#042[memory]KV Reuse Breaks When the Text Moves, Even If the Text Does NotDIGEST · 17m#043[fabric]A Flow-Size Distribution Cannot Describe a BurstDIGEST · 20m#044[fabric]A 100K-GPU Network Starts as a Compiler ProblemDIGEST · 22m#045[fabric]Spend Delay Before the First Hop to Buy Back BandwidthDIGEST · 21m#046[memory]The Fastest Model Loader Changed the Page Cache, Not the FrameworkDIGEST · 16m#047[memory]Five Regions Changed the Economics of Apple's Object StoreDIGEST · 16m#048[memory]Compression Improves When the Filesystem Sorts Before It PacksDIGEST · 14m#049[memory]The Boundary Learned to Move: 52 Industry Memory and Storage Papers in 2026DIGEST · 28m#050[fabric]Beyond NVLink and InfiniBand: A One-Chip-Like CXL DatacenterDIGEST · 33m#051[silicon]Arrays, Timing, and Structured Compute: Industry Circuits in 2026DIGEST · 18m#052[silicon]AmpereOne Hides Memory Tags in the Bits a Server Already Pays ForDIGEST · 19m#053[silicon]Li Auto Lets the Compiler Drive Data Instead of Rebuilding a GPU CacheDIGEST · 20m#054[silicon]Microsoft's Rowhammer Defense Sleeps Until One Sub-Bank Looks DangerousDIGEST · 20m#055[ai-infra]Reasoning Models Hit the Memory-Capacity Wall Before the Compute CeilingDIGEST · 21m#056[arch]SPEC CPU 2026 Stops Pretending Every Core Runs the Same ProgramDIGEST · 19m#057[memory]The SSD Stayed Local, but Its Control Path Moved Three TimesDIGEST · 17m#058[memory]A Thousand Cartridges, Four Drives: Why Tape Cloud Storage Must Wait on PurposeDIGEST · 19m#059[ai-infra]Meta Gives Kernel Optimization a Search Tree and a MemoryDIGEST · 20m#060[memory]The Hierarchy Became the Product: 34 Industry Memory and Storage Papers from 2025DIGEST · 29m#061[fabric]Ask the Switch Which Path the Probe Actually TookDIGEST · 18m#062[silicon]Precision Moves Into the Architecture: Industry Circuits in 2025DIGEST · 19m#063[fabric]A Local Queue Cannot See the Congestion Two Switches AwayDIGEST · 18m#064[fabric]The Training Job Already Knows Which Network Paths MatterDIGEST · 18m#065[arch]A CPU Benchmark That Meta Was Willing to Buy Servers WithDIGEST · 19m#066[ai-infra]Why Llama 3 Needed Four Kinds of Parallelism at OnceDIGEST · 20m#067[ai-infra]Alibaba Makes Collective Communication Diagnose the ClusterDIGEST · 21m#068[ai-infra]The Largest GPU Jobs Fail Most, but Small Jobs Still Set Fleet PolicyDIGEST · 21m#069[ai-infra]AI Energy Needs a Meter That Survives Nine Orders of MagnitudeDIGEST · 21m#070[silicon]What Industry Put on Silicon in 2024DIGEST · 18m#071[memory]Memory Stopped Being a Component: 35 Industry Papers from 2024DIGEST · 29m#072[silicon]MI300A Turns the Accelerator Package into the Compute NodeDIGEST · 19m#073[silicon]FuriosaAI Replaces the Fixed Matrix with a Shape-Shifting Tensor EngineDIGEST · 20m#074[arch]Vera Rebalances the CPU: 88 Cores Fed by 1.2 TB/sDIGEST · 19m#075[ai-infra]The Agent Graph Becomes the Cloud Scheduler: Murakkab Optimizes the Whole WorkflowDIGEST · 19m#076[memory]A Cache Hit That Arrives Late Is Still a Miss: Strata Schedules Long-Context MemoryDIGEST · 19m#077[ai-infra]Fill the Idle Half of On-Policy RL: Weave Co-Schedules Rollout and TrainingDIGEST · 20m#078[ai-infra]When Storage Understands the Job: AITURBO Turns AI I/O into a Group OperationDIGEST · 19m#079[ai-infra]The Datacenter Became the Runtime: Nine Industry Papers That Redefined AI InfrastructureDIGEST · 23m#080[ai-infra]Idle Is Not Available: The Accounting Problem Inside Alibaba's 155,410-GPU FleetDIGEST · 19m#081[ai-infra]A Silent GPU Error Needs Three Different QuestionsORIGINAL · 19m#082[ai-infra]Trace the Request That Missed Its SLO, Not Every Request Around ItDIGEST · 19m#083[ai-infra]The Pipeline Is No Longer Made of Identical BricksDIGEST · 19m#084[ai-infra]Move the Training Job Before It Learns That a Machine LeftDIGEST · 18m#085[ai-infra]The Weather File Expires Before the Datacenter: Google Reprices Cooling for 2044DIGEST · 24m#086[ai-infra]Power Is the Cluster: What 83,000 GB200s Teach About 150 MWDIGEST · 28m#087[ai-infra]What a Neocloud Number Proves, and What It Leaves UnmeasuredORIGINAL · 20m#088[fabric]An AI Fabric Needs Three Views of the Same FailureORIGINAL · 19m#089[ai-infra]When Thousands of GPUs Wait for Storage: Reading MLPerf Storage 2.0DIGEST · 19m#090[arch]One Scheduler for Ten Accelerators: XSched's Three Levels of PreemptionDIGEST · 19m#091[ai-infra]Three Percent Globally, a Grid Problem Locally: Reading the IEA's AI ForecastDIGEST · 20m#092[memory]Mobile Memory Takes the Server Socket: LPDDR6 Silicon, SOCAMM2 and DRAM That ComputesDIGEST · 19m#093[memory]NAND Moves Up a Tier: Reading the First High Bandwidth Flash SpecificationDIGEST · 20m#094[silicon]The PCIe Tax on CXL, and the 4 nm Silicon Built to Repeal ItDIGEST · 22m#095[silicon]Dropping the PHY: d-Matrix Bonds Compute Directly onto DRAMDIGEST · 19m#096[fabric]Who Wires the Rack: Reading UALink Against NVLink and EthernetORIGINAL · 21m#097[ai-infra]Seven Models, One GPU: What the Long Tail Costs and How Aegaeon Reprices ItDIGEST · 20m#098[ai-infra]Evict First, Ask Questions Later: How ByteDance Keeps 9,600 GPUs Worth Training OnDIGEST · 19m#099[fabric]The FLOPS You Cannot Buy, You Wire: Inside Tencent's Half-Million-GPU FabricDIGEST · 20m#100[fabric]Firing the Switch: A Scale-Up Domain Built from TransceiversDIGEST · 19m#101[ai-infra]The Hardware Wishlist DeepSeek Wrote on a Halved NVLinkDIGEST · 21m#102[silicon]No HBM, No Apologies: MTIA 2i and Meta's Productionization LedgerDIGEST · 19m#103[ai-infra]One Controller Above, Many Below: HybridFlow, the Engine Inside verlDIGEST · 19m#104[ai-infra]Buying Back FLOPs with DRAM: Mooncake, the Cache That Serves KimiDIGEST · 21m#105[memory]CXL Leaves the Lab: How Meta Turned Retired DDR4 into a Production Memory TierDIGEST · 21m#106[memory]KVCache Without the Network: Alibaba Runs LLM Serving on a Real CXL 2.0 SwitchDIGEST · 19m