Search Results: AMD

Google has developed BPF-CCX, an innovative Linux scheduler leveraging eBPF for optimized thread placement across AMD Zen core complexes. Benchmarks presented at Linux Plumbers Conference 2025 show significant gains over the default EEVDF scheduler, particularly for virtualized workloads. This approach marks a strategic shift in CPU resource management for EPYC servers and Ryzen systems.

A new study details the first large-scale mixture-of-experts (MoE) pretraining entirely on AMD hardware, introducing ZAYA1—a 760M active parameter model that rivals Qwen3-4B and Gemma3-12B. Packed with microbenchmarks for Pollara networking and MI300X optimizations, it offers actionable guidance for developers eyeing non-NVIDIA AI training stacks. This work underscores AMD's readiness for competitive foundation model development.

While AMD's mainstream AMDGPU driver lacks HDMI 2.1 support due to HDMI Forum licensing restrictions, the company's AMD-Xilinx driver for ZynqMP/Versal SoCs shows partial implementation. This disparity highlights hardware-firmware differences that sidestep open-source challenges faced by discrete GPUs. Developers and Linux users await potential shifts in licensing or hardware design to bridge the gap.

As AMD GPUs match and even surpass NVIDIA's performance in raw specs, a critical software gap remains. Researchers from Stanford's Hazy Research group introduce HipKittens, a collection of programming primitives that unlocks the true potential of AMD's CDNA architecture, paving the way for a more diverse and competitive AI hardware landscape.

When AMD mistakenly pushed the FSR 4 source code to its public repo, it exposed more than an unreleased build—it revealed a sharp, disciplined hybrid of hand-tuned graphics engineering and tiny neural networks. This is how AMD is closing the gap with DLSS without surrendering to bloated AI, and what it means for engine developers and GPU vendors racing to own the frame.

Oracle announces an unprecedented 18 zettaFLOPS AI infrastructure rollout using 800,000 Nvidia Blackwell and 50,000 AMD MI450X GPUs, marking one of history's largest compute deployments. The clusters will leverage Nvidia's Spectrum-X networking and AMD's open UALink architecture, with OpenAI positioned as a key beneficiary. This massive scaling raises critical questions about practical FP4 utilization and the staggering energy demands of next-gen AI.