# UC Berkeley

Published articles for UC Berkeley.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Chip Industry Technical Paper Roundup: Oct. 6

DevFeed: [Chip Industry Technical Paper Roundup: Oct. 6](<https://devfeed.tech/articles/chip-industry-technical-paper-roundup-oct-6-65548.md>)

Original publisher: [Read original article](<https://semiengineering.com/chip-industry-technical-paper-roundup-oct-6/>)

Author: Linda Christensen

Published: 2026-10-06T07:01:55Z

Content type: article

Language: en

Sources: [Semiconductor Engineering](<https://devfeed.tech/sources/semiconductor-engineering.md>)

Topics: [Chip design](<https://devfeed.tech/topics/chip-design.md>), [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [LLM Techniques](<https://devfeed.tech/topics/llm-techniques.md>)

Tags: [fau-erlangen-nurnberg](<https://devfeed.tech/tags/fau-erlangen-nurnberg.md>), [furiosaai](<https://devfeed.tech/tags/furiosaai.md>), [georgia-institute-of-technology](<https://devfeed.tech/tags/georgia-institute-of-technology.md>), [hong-kong-polytechnic-university](<https://devfeed.tech/tags/hong-kong-polytechnic-university.md>), [huazhong-university-of-science-and-technology](<https://devfeed.tech/tags/huazhong-university-of-science-and-technology.md>), [infineon-technologies](<https://devfeed.tech/tags/infineon-technologies.md>), [national-university-of-singapore](<https://devfeed.tech/tags/national-university-of-singapore.md>), [paper](<https://devfeed.tech/tags/paper.md>), [purdue-university](<https://devfeed.tech/tags/purdue-university.md>), [qatar-computing-research-institute](<https://devfeed.tech/tags/qatar-computing-research-institute.md>), [research](<https://devfeed.tech/tags/research.md>), [research-lphp](<https://devfeed.tech/tags/research-lphp.md>), [semiconductor-research](<https://devfeed.tech/tags/semiconductor-research.md>), [syracuse-university](<https://devfeed.tech/tags/syracuse-university.md>), [technical](<https://devfeed.tech/tags/technical.md>), [tu-dresden](<https://devfeed.tech/tags/tu-dresden.md>), [tu-munich](<https://devfeed.tech/tags/tu-munich.md>), [uc-berkeley](<https://devfeed.tech/tags/uc-berkeley.md>), [uc-santa-barbara](<https://devfeed.tech/tags/uc-santa-barbara.md>), [uc-santa-cruz](<https://devfeed.tech/tags/uc-santa-cruz.md>)

### AI overview

This roundup lists newly added semiconductor research papers on WSe₂ transistors, clock meshes for 2nm GAAFETs, DRAM processing, high-throughput LLM serving, CAN XL security analysis, and abstraction and validation from physical devices to RTL.

### Source excerpt

Low-contact-resistance WSe₂ transistors; backside clock meshes for 2nm GAAFETs; row-parallel processing in DRAM; HBF for high-throughput LLM serving; formal security analysis of CAN XL; abstraction and validation from physical devices to RTL. The post Chip Industry Technical Paper Roundup: Oct. 6 appeared first on Semiconductor Engineering.

## UC Berkeley and FuriosaAI study high-bandwidth flash for LLM serving

DevFeed: [UC Berkeley and FuriosaAI study high-bandwidth flash for LLM serving](<https://devfeed.tech/articles/hbf-for-high-throughput-llm-serving-uc-berkeley-furiosaai-64274.md>)

Original publisher: [Read original article](<https://semiengineering.com/hbf-for-high-throughput-llm-serving-uc-berkeley-furiosaai/>)

Author: Technical Paper Link

Published: 2026-10-02T20:15:40Z

Content type: article

Language: en

Sources: [Semiconductor Engineering](<https://devfeed.tech/sources/semiconductor-engineering.md>)

Topics: [Large Language Model](<https://devfeed.tech/topics/llm.md>), [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [Multi-GPU](<https://devfeed.tech/topics/multi-gpu.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai-accelerators](<https://devfeed.tech/tags/ai-accelerators.md>), [ai-ml-dl](<https://devfeed.tech/tags/ai-ml-dl.md>), [architectures](<https://devfeed.tech/tags/architectures.md>), [data-movement](<https://devfeed.tech/tags/data-movement.md>), [dram](<https://devfeed.tech/tags/dram.md>), [flash-memory](<https://devfeed.tech/tags/flash-memory.md>), [furiosaai](<https://devfeed.tech/tags/furiosaai.md>), [hbf](<https://devfeed.tech/tags/hbf.md>), [hbm](<https://devfeed.tech/tags/hbm.md>), [high-bandwidth-flash](<https://devfeed.tech/tags/high-bandwidth-flash.md>), [high-bandwidth-memory](<https://devfeed.tech/tags/high-bandwidth-memory.md>), [kv-cache](<https://devfeed.tech/tags/kv-cache.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [llm](<https://devfeed.tech/tags/llm.md>), [llm-inference](<https://devfeed.tech/tags/llm-inference.md>), [llm-serving](<https://devfeed.tech/tags/llm-serving.md>), [memory](<https://devfeed.tech/tags/memory.md>), [memory-bandwidth](<https://devfeed.tech/tags/memory-bandwidth.md>), [memory-hierarchy](<https://devfeed.tech/tags/memory-hierarchy.md>), [memory-management](<https://devfeed.tech/tags/memory-management.md>), [nonvolatile-memory](<https://devfeed.tech/tags/nonvolatile-memory.md>), [performance](<https://devfeed.tech/tags/performance.md>), [power-performance](<https://devfeed.tech/tags/power-performance.md>), [simulations](<https://devfeed.tech/tags/simulations.md>), [storage](<https://devfeed.tech/tags/storage.md>), [system-design](<https://devfeed.tech/tags/system-design.md>), [technical-papers](<https://devfeed.tech/tags/technical-papers.md>), [trace](<https://devfeed.tech/tags/trace.md>), [uc-berkeley](<https://devfeed.tech/tags/uc-berkeley.md>), [university-of-california-berkeley](<https://devfeed.tech/tags/university-of-california-berkeley.md>)

### AI overview

A UC Berkeley and FuriosaAI research team evaluates high-bandwidth flash (HBF) as additional memory for high-throughput LLM serving. The study examines an HBM-HBF-host storage hierarchy and buffered cache-aware scheduling using trace-driven simulations. In evaluated workloads, the fastest HBF-augmented systems reduced completion time by 36.1-87.0% versus HBM-only systems, with modeled energy savings up to 55.8%; results varied for light workloads. Buffered scheduling extended estimated HBF write lifetime from 4.77 to 14.82 years in the evaluated configuration.

### Source excerpt

Researchers at the UC Berkeley and FuriosaAI published a technical paper titled "Characterizing High Bandwidth Flash for LLM Serving." Abstract "Large language model (LLM) serving requires substantial memory to store model weights and KV caches. As models grow larger and contexts become longer, memory capacity and bandwidth increasingly become bottlenecks for serving performance. Agentic workloads... " read more The post HBF for High-Throughput LLM Serving (UC Berkeley, FuriosaAI) appeared first on Semiconductor Engineering.