SYLIN

AMD

Maker · 6 entries

Everything from AMD

NameWhat to knowPrice or memoryRanks
AMD Radeon AI PRO R9700HardwareEight R9700 cards at home run local Qwen and GLM models for debugging, kernel and driver work, per one user
AMD InstinctHardwareMI355X running vLLM with open FlyDSL kernels beats Blackwell on MiniMax M3.1, a more than 3x improvement
Instella-MoEModel16B MoE (2.8B active) trained from scratch on AMD Instinct GPUs with ROCm; checkpoints, recipes and code published~32 GB
FarSkip-CollectiveTechniqueOverlaps expert-parallel GPU data exchange with compute; AMD reports 12.7% shorter pre-training and up to 39.2% lower time to first token
Gated MLATechniqueLearned gate after multi-head latent attention that turns down weak signals, credited for strong results with few active parameters
AMD Ryzen AI MAX+ 395HardwareSingle APU (Radeon 8060S, gfx1151, ~125GB visible unified memory) is enough to run a 284B MoE model locally