Tag
AMD is pouring substantial engineering and compute resources into vLLM to bypass NVIDIA’s CUDA moat, drive the standardization of inference software, and break hardware lock-in—benefiting the entire ecosystem while yielding asymmetric returns for the two companies.
AMD’s China Developer Platform has launched free API access to 15 large language models, including DeepSeek V4.1 Flash, GLM-5.3 Flash, Qwen3.8, and MiMo. No payment card is required, and quotas reset daily at midnight. Users can sign in with a phone number or GitHub account to obtain an API key and start making calls directly.
据报导，AMD 计划在 11 月再次上调 Radeon GPU 价格，这可能进一步影响显卡市场的定价动态。
AMD released the ROCm HRX Loom SDK v0.1, a specialized compiler and IR stack that targets AMD CPUs, GPUs, and NPUs, with support for AMDGPU HSA, SPIR-V/Vulkan, and emerging x86/WebAssembly backends.
AMD 发布了针对其 Ryzen AI NPU 优化的 Stable Diffusion 3.5 Medium，可在 Hugging Face 上获取，实现高效的本地文本到图像生成。
Meta表示Nvidia、AMD和Broadcom对其普通股表达了合计最多50亿美元的非约束性购买意向，同时Meta将整体融资目标从950亿美元上调至9750亿美元，创下其史上最大规模融资。
A llama.cpp pull request discussion about improving strata decode performance on an AMD R9700 GPU, referencing the ggml/llama.cpp C/C++ LLM inference project.
Intel and AMD jointly announced they will collaborate on x86 CPU software optimizations for PyTorch via oneDNN, with AMD contributing its ZenDNN improvements upstream and joining the UXL Foundation as a steering member.
Rumors suggest AMD's upcoming Zen 6 desktop processors may replace the integrated GPU with a dedicated NPU, raising questions about its practical usefulness for running AI models without hybrid NPU-GPU workloads.
AnythingLLM Desktop is replacing its bundled Ollama engine with AMD's open-source Lemonade server running llama.cpp, aiming for faster updates, direct GGUF downloads from Hugging Face, and broader hardware support including Vulkan, CUDA, Metal, and NPU acceleration.
A post floats the idea of using AI to reverse-engineer NVIDIA's CUDA into an open-source implementation, inspired by AI-assisted game modding, and highlights ongoing frustration among AMD Radeon users over proprietary CUDA lock-in.
Developer helped speed up ROCmFPX, a llama.cpp fork, on AMD Strix Halo (Ryzen AI Max+ 395), achieving major prefill and decode speedups with 16 merged PRs and improved quantization accuracy.
zzpanic shares the final revision of his vllm-radiance build tuned for a single AMD R9700 GPU running qwen3.8-27b, adding startup caching and a KV-offload feature (vram-sysram-disk) with memory thinning to enable multi-agent workloads on one card. He notes the project is winding down as users migrate to StillDeadcode's newer radiance inference engine.
AMD is reportedly increasing GDDR6 memory prices for its board partners, potentially impacting GPU pricing and availability downstream.
MTP 投机解码的草稿 LM head 原本每步都要读取 1.27 GB 的 BF16 权重，现改为 Q4_0 量化草稿头（验证与正常解码仍用 BF16），实现约 10.6% 的 token/s 提升，且 greedy 输出 md5 完全一致。相关改动已在 ROCmFPX 的 PR #49 中提交。
A user shares a 4x AMD Radeon R9700 AI Pro workstation build (128GB total VRAM, Threadripper 9970x) serving Qwen 3.8 Next Flash/27B and DSV4 Flash for a 10-person startup, reporting strong aggregate prefill/decode throughput across 16 concurrent sessions.
ROCmFPX 的融合 HC norm 内核通过在寄存器中直接写出 Q8_1 副本，省去了 MMQ 重复量化 335MB 数据的开销，在 2.3%–3.6% 的 pp8192 性能提升下保持位级一致。
TensorFold 1.0.5 发布首个 AMD GPU 运行时（支持 gfx1150/gfx1151/gfx1201），在 Strix Halo 上跑 Qwen3.6-35B-A3B 解码速度比 llama.cpp 快 36-54%，预填充快 40%，目标是让单一模型跨 NVIDIA、Mac 和 AMD 运行且远快于 llama.cpp。
Due to soaring RAM prices driven by the DDR5 shortage, Intel and AMD are reviving support for decade-old DDR4 memory, with new LGA 1700 and AM4 motherboards expected as DDR4 prices remain far lower than DDR5.
AMD consumer chip chief Jack Huynh confirmed that FSR 4 will expand to APUs, gaming laptops, and handheld devices by the end of 2026, with both new chips and a lightweight FSR 4 model reportedly in the works, though it remains unclear whether existing handhelds like the Steam Deck will be supported.