Chinese-first Ramulator2 ECC/EDC reliability research prototype for AI and HPC memory studies
-
Updated
Aug 25, 2026 - C++
#
hbm3
Here are 3 public repositories matching this topic...
Software-Defined PIM-HBM Bypass Infrastructure. 0ns JAX/XLA memory view fusion with real-time asynchronous NCCL collective fault-scanning & no-recompile Mux hot-swapping.
zero-copy cuda high-performance-computing gpu-acceleration pim compiler-optimization fault-tolerant distributed-training nccl jax system-software xla double-buffering branchless hardware-software-codesign llama3 hot-plugging hbm3 memory-bypass matrix-sharding
-
Updated
Jul 16, 2026 - Python
Professional benchmarking guide for evaluating NVIDIA H200 GPU memory bandwidth using the STREAM Triad (FP32) benchmark with CUDA 12.x on Linux. Includes setup, execution, performance analysis, and multi-GPU benchmarking results.
performance high-performance-computing stream-benchmarks gpu-benchmarking nvidia-h200 memory-bandwidth-benchmarking cuda124 hbm3
-
Updated
Jul 25, 2026
Add this topic to your repo
To associate your repository with the hbm3 topic, visit your repo's landing page and select "manage topics."