This letter presents a 16nm 496-core RISC-V network-on-chip (NoC). The mesh achieves 1.4GHz at 0.98V, yielding a peak throughput of 695 Giga RISC-V instructions/s (GRVIS), a peak energy efficiency of 314.89 GRVIS/W, and a record 825,320 CoreMark benchmark score. Unlike previously reported rovinski2019, this new score was obtained without modifying the core benchmark code. The main feature is the NoC architecture, which uses only 1881 per router node, enables highly scalable and dense compute, and provides up to 361 Tb/s of aggregate bandwidth.
1 January 2020 in IEEE Journal of Solid-State Circuits
A sparse matrix-matrix multiplication (SpMM) accelerator with 48 heterogeneous cores and a reconfigurable memory hierarchy is fabricated in 40-nm CMOS. The compute fabric consists of dedicated floating-point multiplication units, and general-purpose Arm Cortex-M0 and Cortex-M4 cores. The on-chip memory reconfigures scratchpad or cache, depending on the phase of the algorithm. The memory and compute units are interconnected with synthesizable coalescing crossbars for efficient memory access. The 2.0-mm x 2.6-mm chip exhibits 12.6x (8.4x) energy efficiency gain, 11.7x (77.6x) off-chip bandwidth efficiency gain, and 17.1x (36.9x) compute density gains against a high-end CPU (GPU) across a diverse set of synthetic and real-world power-law graph-based sparse matrices.
Large language models (LLMs) have shown significant potential in serving as domain-specific chatbots. Recently, these models have emerged as powerful tools for chip design, providing both natural language responses and script generation for domain-specific inquiries. Previous work has demonstrated the effectiveness of LLMs in assisting with physical design automation; however, these approaches often rely on proprietary tools, APIs, technologies, and designs. As a result, access to these models is extremely limited, particularly for new chip designers who could greatly benefit from a design assistant. This paper introduces OpenROAD-Assistant, an open-source chatbot for OpenROAD that relies only on public data and responds to queries in either prose or Python script using the OpenROAD APIs. OpenROAD-Assistant leverages the Llama3-8B foundation model and employs retrieval-aware fine-tuning (RAFT) to respond to physical design-specific questions for OpenROAD. Notably, OpenROAD-Assistant outperforms other foundational models such as ChatGPT3.5, ChatGPT4, Code Llama, Claude3, and other ablation study baselines on the measured metrics (pass@k for scripting and BERTScore/BARTScore for question-answering). OpenROAD-Assistant achieves a 77% pass@1 score, 80% pass@3 score for scripting, and it achieves a 98% BERTScore and 96% BARTScore on question-answering.
18 February 2025 in IEEE Computer Architecture Letters
Large language models (LLMs) have grown exponentially in size, presenting significant challenges to traditional memory architectures. Current high bandwidth memory (HBM) systems are constrained by chiplet I/O bandwidth and the limited number of HBM stacks that can be integrated due to packaging constraints. In this letter, we propose a novel memory system architecture that leverages silicon photonic interconnects to increase memory capacity and bandwidth for compute devices. By introducing optically connected multi-stack HBM modules, we extend the HBM memory system off the compute chip, significantly increasing the number of HBM stacks. Our evaluations show that this architecture can improve training efficiency for a trillion-parameter model by 1.4× compared to a modeled A100 baseline, while also enhancing inference performance by 4.2× if the L2 is modified to provide sufficient bandwidth.