#MI355
The latest vLLM release, version 0.30.1rc0, brings significant enhancements for #AMD #GPU users. The update adds kernel mirrors for the MI355 accelerator, supporting both dense and Mixture of Experts (MoE) model architectures

#ai

www.minaxlab.com/blog/vllm-mi...
vLLM v0.30.1rc0 Adds MI355 GPU Support for Dense and MoE Models
vLLM 0.30.1rc0 introduces ROCm kernel mirrors for AMD MI355 GPUs, enabling optimized inference for dense and Mixture of Experts models.
www.minaxlab.com
September 25, 2026 at 3:28 AM
AgentX - InferenceXv3: Does CUDA Moat Hold up in Agentic Inferencing?
AgentX - InferenceXv3: Does CUDA Moat Hold up in Agentic Inferencing?
$3 Million USD dataset open sourced, 1 Mil+ Context Length, Multiturn, Sub Agents 95%+ KVCache HitRate, GB300 NVL72, MI355, B200
newsletter.semianalysis.com
September 2, 2026 at 7:19 PM
SemiAnalysis Open-Sources $3M Agentic Inference Dataset

SemiAnalysis open-sourced a $3M agentic inference dataset targeting GB300 NVL72, MI355, B200. It tests whether CUDA's moat holds under 1M+ context, multiturn workloads.

https://gentic.news/article/semianalysis-open-sources-3m
SemiAnalysis Open-Sources $3M Agentic Inference Dataset
SemiAnalysis open-sources $3M agentic inference dataset targeting GB300 NVL72, MI355, B200, testing CUDA's moat with 1M+ context and 95%+ KVCache hit rates.
gentic.news
August 25, 2026 at 4:26 PM
$3 Million USD dataset open sourced, 1 Mil+ Context Length, Multiturn, Sub Agents 95%+ KVCache HitRate, GB300 NVL72, MI355, B200
AgentX - InferenceXv3: Does CUDA Moat Hold up in Agentic Inferencing?
newsletter.semianalysis.com
August 25, 2026 at 10:01 AM
用4个Mi355 运行 Kimi-K2.5-MXFP4(MXFP4 量化版本), 4 Bits, 160GB, 无需软件解压, 计算时保持低精度, 节省带宽和算力, 启动参数里还有针对 AMD 的优化
--attention-backend ROCM_AITER_MLA
--kernel-config.enable_flashinfer_autotune false
--enforce-eager
用Kimi Code做简单工作目前感觉ok
August 22, 2026 at 11:05 PM
MyPOV: @AnthropicAI to be deploying 2GW's of @amd #helios says Tom Brown to Lisa Su #AMDAdvancingAI

Anthropic likes the AMD Instinct MI355
July 23, 2026 at 4:54 PM
AMD and Meta contributors detail the ROCm port that brings PyTorch Monarch to AMD Instinct GPUs.

Validation includes Llama 3 8B on 128 GPU MI300 and 256 GPU MI355 clusters.

https://pytorch.org/blog/bringing-pytorch-monarch-to-amd-gpus-single-controller-distributed-training-on-rocm/
July 7, 2026 at 5:00 PM
Feed: "Yahoo Finance"
Published on Tuesday, March 24, 2026
AMD Is Rapidly Expanding into South Korea. Can That Move the Needle for AMD Stock?
AMD accelerates its expansion process into South Korea with negotiations underway for 10,000 MI355 AI accelerators.
finance.yahoo.com
March 26, 2026 at 12:06 AM
Feed: "DCD"
Published on Monday, March 23, 2026
Upstage in talks with AMD to deploy 10,000 MI355 GPUs across South Korea
Companies currently negotiating deal following visit to country by AMD CEO, Lisa Su
www.datacenterdynamics.com
March 24, 2026 at 7:47 AM
韓国のAIスタートアップUpstageがAMDと提携し、今後1年でInstinct MI355アクセラレータを導入。韓国のAI主権強化と自社モデル開発を加速させる戦略的動きです。詳細は記事で。
https://biggo.jp/news/202603231044_Upstage_AMD_MI355_AI_Accelerator_Deployment

#AMDInstinctshtag/AMDInstinctMI355" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#AMDInstinctMI355 #AMDInstinct
March 23, 2026 at 11:06 AM
韓國AI新創Upstage聯手AMD,部署MI355加速器打造自主AI主權,減少對外依賴!
https://biggo.com.tw/news/202603231044_Upstage_AMD_MI355_AI_Accelerator_Deployment

#AMDInstinctMI355 #AMDMI355
March 23, 2026 at 11:06 AM
Korean AI startup Upstage says it is in talks to acquire 10K of AMD's MI355 chips, in a bid to "diversify to other chips" as "we have a lot of Nvidia chips" (Bloomberg)

Main Link | Techmeme Permalink
March 23, 2026 at 10:05 AM
AMD and Upstage are strengthening their partnership to accelerate the development of Korea's sovereign AI models using next-generation Instinct MI355 accelerators.

🔗 Read more on NewDecoded: https://nwd.plus/JZb1x4g
AMD and Upstage Expand Strategic Alliance to Power South Korea Sovereign AI Infrastructure | NewDecoded
AMD and Upstage are strengthening their partnership to accelerate the development of Korea
nwd.plus
March 20, 2026 at 9:01 PM
又是雷聲大雨點小!AMD何時才能硬氣?
又是雷聲大雨點小!AMD何時才能硬氣?
MI355量產不達預期,指望MI450的“機架級”方案。
www.headline01.com
February 4, 2026 at 5:33 AM
$AMD +1.5% [KeyBanc upgrades AMD to Overweight with a $270 price target, citing strong momentum in the company's AI accelerators, especially the MI355 and upcoming MI455 chips.] notreload.xyz/keybanc-upgr...
KeyBanc Upgrades AMD on Strong AI and Server Demand
KeyBanc upgrades AMD to Overweight with a $270 price target, citing strong momentum in the company's AI accelerators, especially the MI355 and upcoming MI455 chips.
notreload.xyz
January 13, 2026 at 11:21 AM