» Archived Articles | Login | Activate My Subscription | Subscribe | Free Sample Articles | Purchase an Article

Current MPR Articles

November 23, 2020

• Cisco Debuts 12.8Tbps Switch Chips
  Just 10 months after entering the merchant switch-chip market, Cisco announced its Silicon One 7nm family for data-center and service-provider networks.
  

• Intel Takes First Step in Discrete GPUs
  The Xe Max, code-named DG1, is Intel’s first discrete-graphics product in two decades. It will debut later this quarter and targets laptop PCs using Tiger Lake processors.
  

• Editorial: Qualcomm Bounces Back
  After a bad 2019, Qualcomm has recently seen a run of good news, including legal victories, a market-share uptick, approval to sell to Huawei, and booming RF revenue.
  

November 16, 2020

• SiFive VIU7-256 Takes Vector Lead
  The VIU7-256 is SiFive’s first vector CPU. The new 64-bit core, which processes 512-bit vectors, is built on the newest RISC-V standard. Preliminary RTL is shipping now.
  

• Imagination Series4 Tiles Tensors
  Imagination Technologies’ new IMG Series4 deep-learning accelerators reduce DRAM bandwidth by splitting tensor operations into multilayer groups, keeping most intermediate data on chip.
  

• Flex Logix Shows First AI Chip
  At the recent Linley Fall Processor Conference, the startup revealed that its InferX X1 chip is running in its lab, and it announced the X1P1 PCI Express accelerator card.
  

November 9, 2020

• Ryzen 5000 Rides Zen 3 to the Top
  AMD’s second-generation 7nm CPU drives leadership desktop-PC performance. The Zen 3 microarchitecture increases both single-thread performance and power efficiency.
  

• Cortex-A78C Improves Security
  Arm’s newest CPU for the PC market boosts the core count, expands the L3 cache, and improves security. These system-level changes target the CPU for the midrange PCs and mobile gaming.
  

• Groq Gets Connected
  Advancing toward production, the startup disclosed additional details about its TSP chip, focusing on the high-speed-I/O subsystem and the initial products, a PCIe accelerator card and a complete server.
  

November 2, 2020

• Untether Delivers At-Memory AI
  Using what it calls an at-memory architecture, Untether AI has created a highly power-efficient PCIe card that can achieve a stunning two petaop/s for neural-network inferencing using nearly 800MB of SRAM.
  

• Deep AI Retrains Networks at the Edge
  Startup Deep AI is enabling infrastructure-edge hardware to perform both retraining and inference. The company licenses its software and RTL for use with FPGA-based accelerator cards from Xilinx.
  

• Ambient’s Analog Cuts AI Power
  Startup Ambient Scientific uses analog technology to reduce power for AI inference and training in IoT devices. Its GPX-10 SoC delivers 510 billion operations per second (GOPS) while consuming only 120mW.
  

October 26, 2020

• Arm’s SVE2 Enables Shift From Neon
  Arm’s second-generation Scalable Vector Extension (SVE2) will enable a transition from the decade-old Neon technology, lifting SIMD performance in smartphones.
  

• Think Silicon Spins AI Accelerator
  Think Silicon’s licensable Neox cores combine RISC-V with nonstandard extensions that accelerate graphics and neural-network inference. Customers can add their own instructions.
  

• Editorial: Shopping for Chip Companies
  AMD is reportedly negotiating a $30 billion bid for Xilinx, the leading FPGA vendor. The two companies have little product or end-market synergy; the deal seems driven more by financial concerns.
  

October 19, 2020

• Neoverse Advances SIMD for HPC
  Arm’s Neoverse V1 core adopts the Scalable Vector Extension, replacing the venerable Neon core. It’s the company’s first CPU purpose built for high-performance computing.
  

• SiPearl Develops Arm HPC Chip
  SiPearl is a little company with a big goal: helping Europe wean itself off of x86-based supercomputers and reduce its dependence of foreign silicon. Its first chip, Rhea, uses Arm’s Neoverse V1 CPU.
  

• Coherent Logix Configures Edge AI
  The niche embedded-processor supplier is expanding into the commercial edge-AI market. By the end of the year, it plans to tape out its fourth-generation HyperX processor, delivering 6.6 TOPS.
  

October 12, 2020

• Elkhart Lake Has Industrial Aims
  Intel’s Atom x6000E moves to 10nm technology and adds embedded features, serving applications beyond those available to its PC-derived predecessors.
  

• Kneron KL720 Boosts Efficiency
  Kneron has quickly rolled out its second product, offering a big performance boost to 1.5 TOPS with a power of about 1W, making it well suited to smart cameras and other vision-based edge systems.
  

• Qualcomm Samples First AI Chip
  The Cloud AI 100 accelerator provides more than 4x the ResNet-50 inference throughput of an Nvidia T4 card at about the same power. Most initial customers plan to deploy it at the network edge.
  

October 5, 2020

• Ampere Delivers Big Graphics Boost
  The new Ampere graphics cards deliver one of Nvidia’s largest generational performance increases ever and offer a second-generation ray-tracing engine that doubles throughput and improves motion blur.
  

• Qualcomm Rolls More Snapdragon 7s
  Qualcomm’s new Snapdragon 750G reduces the entry price for mmWave 5G phones. Although it offers greater performance than all 7-series chips other than the 768G, it targets phones that sell for just $400.
  

• Xilinx T1 Enables Open RAN for 5G
  The T1 Telco Accelerator card uses a pair of UltraScale+ FPGAs to handle fronthaul termination and Layer 1 acceleration in virtualized 5G baseband units.
  

September 28, 2020

• Cortex-R82 Brings Linux to SSDs
  Arm’s first 64-bit Cortex-R adds a memory-management unit for high-level operating systems, enabling computational-storage drives to run Linux. It inherits classic real-time features from the predecessor Cortex-R8.
  

• IBM z15: A Modern Mainframe Chip
  The z15 is IBM’s next-generation Z-series processor for mainframes, operating at a 5.2GHz base frequency. Relative to the z14, the company added two CPU cores and a handful of on-die accelerators.
  

• Editorial: Nvidia Grabs Arm for AI
  Nvidia aims to become the world’s leading computing company with its $40 billion acquisition of Arm, but the mismatch between the two businesses will make it a challenge to pull off.
  

September 21, 2020

• Quantum Computing Seeks Supremacy
  Google, IBM, Intel, Microsoft and other companies are pushing the frontiers of quantum computing, but researchers must reduce the qubit error rate before quantum processors can perform useful tasks.
  

• Lightmatter Realizes Photonic AI
  Lightmatter emerged from stealth by presenting details of its Mars test chip. The 3D chip stacks photonic and digital die, and it shows that a complete product could quickly follow.
  

• Ambiq Doubles CPU Frequency
  The Apollo 4 microcontroller delivers industry-leading energy efficiency while adding a graphics subsystem. Ambiq’s new part also adds location support to its low-power Bluetooth subsystem.
  

September 14, 2020

• Jericho2c+ Brings 7nm to Routers
  Broadcom is sampling its first 7nm router chip, the Jericho2c+ packet processor. The new product serves in line cards for chassis-based routers, pairing with the shipping Ramon fabric chip.
  

• SuperFin Rescues Intel’s 10nm Node
  After last year’s disappointing debut, Intel’s 10nm technology received a much needed performance boost from a new improvement branded SuperFin, now shipping in the company’s Tiger Lake processors.
  

• Baidu Debuts First AI Accelerator
  Joining other cloud-service providers (CSPs), Baidu is taking a shot at AI mastery by offering the Kunlun K200 SoC. The Chinese company designed the 256-TOPS accelerator for its internal deep-learning workloads.
  

September 7, 2020

• Tiger Lake Debuts 10nm SuperFin
  Intel’s 11th Generation Core processors for laptop PCs offer significant improvements in performance and battery life, thanks to the powerful Xe-LP GPU and a more advanced manufacturing process.
  

• Ice Lake-SP Boosts Memory and I/O
  The new Whitley platform boosts memory bandwidth and brings PCI Express Gen4, but Ice Lake’s 10nm process will hold back compute performance relative to Cascade Lake.
  

• Blaize Ignites Edge-AI Performance
  Edge-AI startup Blaize is sampling modules that integrate its El Cano inference processor. The chip can run Yolo v3 on five separate HD-video streams, delivering results at 10fps.
  

• Xbox Series X Doubles GPU Speed
  The semicustom Xbox Series X processor from Microsoft outpaces its predecessor Xbox One X design by offering a 3x CPU-performance boost and 2x GPU-performance boost.
  

August 31, 2020

• Intel Xe Targets Multiple Tiers
  Intel’s new GPU architecture promises to double graphics performance compared with the previous Gen11 design. It will appear in a number of products from laptop processors to discrete graphics cards.
  

• Google Details TPUv3 Architecture
  Google’s TPU chips implement a flexible VLIW microarchitecture comprising scalar, vector, and matrix units to efficiently accelerate neural-network training and inference workloads.
  

• Alibaba Designs Fastest RISC-V CPU
  The Xuantie-910 is Alibaba’s first RISC-V CPU core, targeting the network edge. It includes custom arithmetic and memory instruction to provide a new alternative to midrange Arm cores.
  

More Articles

Purchase an Article