#fastflowlm — Public Fediverse posts

Купил мини‑ПК с приставкой «AI» ради локальной LLM. Что может NPU на самом деле

TL;DR. Я купил мини‑ПК с Ryzen AI 9 365, чтобы гонять большую локальную модель на NPU. Начал с Windows ради гибрида NPU+iGPU, который так и не заработал, прошёл квест с драйвером NPU, выяснил, что NPU видит только половину оперативки, переехал на Proxmox и в итоге запустил Qwen3.6–35B‑A3B через FastFlowLM прямо на хосте. Модель работает 24/7: prefill около 200 ток/с, decode 13–17 ток/с. По дороге выяснилось, что NPU обслуживает один запрос за раз, падает с double free , если клиент не дождался ответа, и умеет выгнать из памяти большую модель ради маленькой embedding‑модели. Ниже вся дорога по порядку, мини‑гайд и цифры.

https://habr.com/ru/articles/1089040/

#npu #amd_ryzen_ai #локальные_llm #fastflowlm #qwen #proxmox #xdna #litellm #selfhosted #инференс

Купил мини‑ПК с приставкой «AI» ради локальной LLM. Что может NPU на самом деле

TL;DR. Я купил мини‑ПК с Ryzen AI 9 365, чтобы гонять большую локальную модель на NPU. Начал с Windows ради гибрида NPU+iGPU, который так и не заработал, прошёл квест с драйвером NPU, выяснил, что NPU видит только половину оперативки, переехал на Proxmox и в итоге запустил Qwen3.6–35B‑A3B через FastFlowLM прямо на хосте. Модель работает 24/7: prefill около 200 ток/с, decode 13–17 ток/с. По дороге выяснилось, что NPU обслуживает один запрос за раз, падает с double free , если клиент не дождался ответа, и умеет выгнать из памяти большую модель ради маленькой embedding‑модели. Ниже вся дорога по порядку, мини‑гайд и цифры.

https://habr.com/ru/articles/1089040/

#npu #amd_ryzen_ai #локальные_llm #fastflowlm #qwen #proxmox #xdna #litellm #selfhosted #инференс

(more Linux and FOSS news in previous posts of thread)

Lemonade 11.6 Integrates Muse-Glimmer 30B, Experimental TheNoise ROCm Image Generation:
https://www.phoronix.com/news/Lemonade-SDK-11.6

AMD GAIA 0.23 Delivers Ability To Install/Run AI Agents From The Terminal:
https://www.phoronix.com/news/AMD-GAIA-0.23

FastFlowLM 1.0 Released Now As Part Of The AMD ROCm Umbrella:
https://www.phoronix.com/news/FastFlowLM-1.0

Overshared to an AI? Proton's Tool Will Give You a Reality Check:
https://feed.itsfoss.com/link/24361/17418931/proton-ai-paper-trail

openJiuwen’s Agent Swarm: When AI Agents Finally Work as a Team:
https://feed.itsfoss.com/link/24361/17411448/openjiuwens-agent-swarm
(Oh yeah, just teach AI agents to team up. Surely they can't cause any trouble, right? RIGHT?)

The Coldcard Exploit Explained: Who Lost Bitcoin and Who's at Risk:
https://news.bitcoin.com/featured/the-coldcard-exploit-explained-who-lost-bitcoin-and-whos-at-risk/

Wireshark 4.6.8 Improves Protocol and Capture File Support, Fixes More Bugs:
https://9to5linux.com/wireshark-4-6-8-improves-protocol-and-capture-file-support-fixes-more-bugs

Roundcube Webmail Patches 11 Critical Vulnerabilities in Emergency 1.6.18 and 1.7.3 Updates:
https://www.linuxcompatible.org/story/roundcube-patches-11-critical-vulnerabilities-in-emergency-1618-and-173-update/

Zed v1.15.0 Lands with Self-Hosted AI Predictions, Git Baseline Toggle, and Wayland Drag-and-Drop:
https://www.linuxcompatible.org/story/zed-editor-v1150-adds-selfhosted-ai-predictions-git-diff-baseline-and-wayland-support/

Flutter 3.47 introduces standalone UI packages and Impeller for desktop:
https://alternativeto.net/news/2026/8/flutter-3-47-introduces-standalone-ui-packages-and-impeller-for-desktop/

Podman 6.1 introduces volume renaming, restart command, and robust bugfixes:
https://alternativeto.net/news/2026/8/podman-6-1-introduces-volume-renaming-restart-command-and-robust-bugfixes/

PHP 8.4.25 and 8.5.10 release candidates drop with serious stack-overflow hardening:
https://www.linuxcompatible.org/story/php-8425-and-8510-release-candidates-ship-with-stack-hardening/

PHP 8.6.0 Beta 1 Lands With Partial Function Application, Native Polling API, and a Hard Shift Toward Strict Errors:
https://www.linuxcompatible.org/story/php-860-beta-1-partial-function-application-native-polling-and-stricter-error-handling/

PostgreSQL 19 Beta 3 and Critical Security Patches Released; PG 14 EOL Notice Issued:
https://www.linuxcompatible.org/story/postgresql-19-beta-3-and-critical-security-patches-released-pg-14-eol-notice-issued/

GCC 17 Compiler Adding "-m128bit-atomic" For 128-bit Atomic Memory Operations:
https://www.phoronix.com/news/GCC-Git-m128bit-atomic

(more FOSS news in comments)

#WeeklyNews #FOSSNews #AI #Zed #Flutter #PHP #Podman #PostgreSQL #Dev #AMDGAIA #FastFlowLM #Coldcard #Bitcoin #Wireshark #GCC #ArtificialIntelligence #LLM #AgenticAI #Programming #Development #Coding #IDE #DBMS #SQL #FosseryTech

(more Linux and FOSS news in previous posts of thread)

Lemonade 11.6 Integrates Muse-Glimmer 30B, Experimental TheNoise ROCm Image Generation:
https://www.phoronix.com/news/Lemonade-SDK-11.6

AMD GAIA 0.23 Delivers Ability To Install/Run AI Agents From The Terminal:
https://www.phoronix.com/news/AMD-GAIA-0.23

FastFlowLM 1.0 Released Now As Part Of The AMD ROCm Umbrella:
https://www.phoronix.com/news/FastFlowLM-1.0

Overshared to an AI? Proton's Tool Will Give You a Reality Check:
https://feed.itsfoss.com/link/24361/17418931/proton-ai-paper-trail

openJiuwen’s Agent Swarm: When AI Agents Finally Work as a Team:
https://feed.itsfoss.com/link/24361/17411448/openjiuwens-agent-swarm
(Oh yeah, just teach AI agents to team up. Surely they can't cause any trouble, right? RIGHT?)

The Coldcard Exploit Explained: Who Lost Bitcoin and Who's at Risk:
https://news.bitcoin.com/featured/the-coldcard-exploit-explained-who-lost-bitcoin-and-whos-at-risk/

Wireshark 4.6.8 Improves Protocol and Capture File Support, Fixes More Bugs:
https://9to5linux.com/wireshark-4-6-8-improves-protocol-and-capture-file-support-fixes-more-bugs

Roundcube Webmail Patches 11 Critical Vulnerabilities in Emergency 1.6.18 and 1.7.3 Updates:
https://www.linuxcompatible.org/story/roundcube-patches-11-critical-vulnerabilities-in-emergency-1618-and-173-update/

Zed v1.15.0 Lands with Self-Hosted AI Predictions, Git Baseline Toggle, and Wayland Drag-and-Drop:
https://www.linuxcompatible.org/story/zed-editor-v1150-adds-selfhosted-ai-predictions-git-diff-baseline-and-wayland-support/

Flutter 3.47 introduces standalone UI packages and Impeller for desktop:
https://alternativeto.net/news/2026/8/flutter-3-47-introduces-standalone-ui-packages-and-impeller-for-desktop/

Podman 6.1 introduces volume renaming, restart command, and robust bugfixes:
https://alternativeto.net/news/2026/8/podman-6-1-introduces-volume-renaming-restart-command-and-robust-bugfixes/

PHP 8.4.25 and 8.5.10 release candidates drop with serious stack-overflow hardening:
https://www.linuxcompatible.org/story/php-8425-and-8510-release-candidates-ship-with-stack-hardening/

PHP 8.6.0 Beta 1 Lands With Partial Function Application, Native Polling API, and a Hard Shift Toward Strict Errors:
https://www.linuxcompatible.org/story/php-860-beta-1-partial-function-application-native-polling-and-stricter-error-handling/

PostgreSQL 19 Beta 3 and Critical Security Patches Released; PG 14 EOL Notice Issued:
https://www.linuxcompatible.org/story/postgresql-19-beta-3-and-critical-security-patches-released-pg-14-eol-notice-issued/

GCC 17 Compiler Adding "-m128bit-atomic" For 128-bit Atomic Memory Operations:
https://www.phoronix.com/news/GCC-Git-m128bit-atomic

(more FOSS news in comments)

#WeeklyNews #FOSSNews #AI #Zed #Flutter #PHP #Podman #PostgreSQL #Dev #AMDGAIA #FastFlowLM #Coldcard #Bitcoin #Wireshark #GCC #ArtificialIntelligence #LLM #AgenticAI #Programming #Development #Coding #IDE #DBMS #SQL #FosseryTech

(more Linux and FOSS news in previous posts of thread)

Lemonade 11.6 Integrates Muse-Glimmer 30B, Experimental TheNoise ROCm Image Generation:
https://www.phoronix.com/news/Lemonade-SDK-11.6

AMD GAIA 0.23 Delivers Ability To Install/Run AI Agents From The Terminal:
https://www.phoronix.com/news/AMD-GAIA-0.23

FastFlowLM 1.0 Released Now As Part Of The AMD ROCm Umbrella:
https://www.phoronix.com/news/FastFlowLM-1.0

Overshared to an AI? Proton's Tool Will Give You a Reality Check:
https://feed.itsfoss.com/link/24361/17418931/proton-ai-paper-trail

openJiuwen’s Agent Swarm: When AI Agents Finally Work as a Team:
https://feed.itsfoss.com/link/24361/17411448/openjiuwens-agent-swarm
(Oh yeah, just teach AI agents to team up. Surely they can't cause any trouble, right? RIGHT?)

The Coldcard Exploit Explained: Who Lost Bitcoin and Who's at Risk:
https://news.bitcoin.com/featured/the-coldcard-exploit-explained-who-lost-bitcoin-and-whos-at-risk/

Wireshark 4.6.8 Improves Protocol and Capture File Support, Fixes More Bugs:
https://9to5linux.com/wireshark-4-6-8-improves-protocol-and-capture-file-support-fixes-more-bugs

Roundcube Webmail Patches 11 Critical Vulnerabilities in Emergency 1.6.18 and 1.7.3 Updates:
https://www.linuxcompatible.org/story/roundcube-patches-11-critical-vulnerabilities-in-emergency-1618-and-173-update/

Zed v1.15.0 Lands with Self-Hosted AI Predictions, Git Baseline Toggle, and Wayland Drag-and-Drop:
https://www.linuxcompatible.org/story/zed-editor-v1150-adds-selfhosted-ai-predictions-git-diff-baseline-and-wayland-support/

Flutter 3.47 introduces standalone UI packages and Impeller for desktop:
https://alternativeto.net/news/2026/8/flutter-3-47-introduces-standalone-ui-packages-and-impeller-for-desktop/

Podman 6.1 introduces volume renaming, restart command, and robust bugfixes:
https://alternativeto.net/news/2026/8/podman-6-1-introduces-volume-renaming-restart-command-and-robust-bugfixes/

PHP 8.4.25 and 8.5.10 release candidates drop with serious stack-overflow hardening:
https://www.linuxcompatible.org/story/php-8425-and-8510-release-candidates-ship-with-stack-hardening/

PHP 8.6.0 Beta 1 Lands With Partial Function Application, Native Polling API, and a Hard Shift Toward Strict Errors:
https://www.linuxcompatible.org/story/php-860-beta-1-partial-function-application-native-polling-and-stricter-error-handling/

PostgreSQL 19 Beta 3 and Critical Security Patches Released; PG 14 EOL Notice Issued:
https://www.linuxcompatible.org/story/postgresql-19-beta-3-and-critical-security-patches-released-pg-14-eol-notice-issued/

GCC 17 Compiler Adding "-m128bit-atomic" For 128-bit Atomic Memory Operations:
https://www.phoronix.com/news/GCC-Git-m128bit-atomic

(more FOSS news in comments)

#WeeklyNews #FOSSNews #AI #Zed #Flutter #PHP #Podman #PostgreSQL #Dev #AMDGAIA #FastFlowLM #Coldcard #Bitcoin #Wireshark #GCC #ArtificialIntelligence #LLM #AgenticAI #Programming #Development #Coding #IDE #DBMS #SQL #FosseryTech

(more Linux and FOSS news in previous posts of thread)

Lemonade 11.6 Integrates Muse-Glimmer 30B, Experimental TheNoise ROCm Image Generation:
https://www.phoronix.com/news/Lemonade-SDK-11.6

AMD GAIA 0.23 Delivers Ability To Install/Run AI Agents From The Terminal:
https://www.phoronix.com/news/AMD-GAIA-0.23

FastFlowLM 1.0 Released Now As Part Of The AMD ROCm Umbrella:
https://www.phoronix.com/news/FastFlowLM-1.0

Overshared to an AI? Proton's Tool Will Give You a Reality Check:
https://feed.itsfoss.com/link/24361/17418931/proton-ai-paper-trail

openJiuwen’s Agent Swarm: When AI Agents Finally Work as a Team:
https://feed.itsfoss.com/link/24361/17411448/openjiuwens-agent-swarm
(Oh yeah, just teach AI agents to team up. Surely they can't cause any trouble, right? RIGHT?)

The Coldcard Exploit Explained: Who Lost Bitcoin and Who's at Risk:
https://news.bitcoin.com/featured/the-coldcard-exploit-explained-who-lost-bitcoin-and-whos-at-risk/

Wireshark 4.6.8 Improves Protocol and Capture File Support, Fixes More Bugs:
https://9to5linux.com/wireshark-4-6-8-improves-protocol-and-capture-file-support-fixes-more-bugs

Roundcube Webmail Patches 11 Critical Vulnerabilities in Emergency 1.6.18 and 1.7.3 Updates:
https://www.linuxcompatible.org/story/roundcube-patches-11-critical-vulnerabilities-in-emergency-1618-and-173-update/

Zed v1.15.0 Lands with Self-Hosted AI Predictions, Git Baseline Toggle, and Wayland Drag-and-Drop:
https://www.linuxcompatible.org/story/zed-editor-v1150-adds-selfhosted-ai-predictions-git-diff-baseline-and-wayland-support/

Flutter 3.47 introduces standalone UI packages and Impeller for desktop:
https://alternativeto.net/news/2026/8/flutter-3-47-introduces-standalone-ui-packages-and-impeller-for-desktop/

Podman 6.1 introduces volume renaming, restart command, and robust bugfixes:
https://alternativeto.net/news/2026/8/podman-6-1-introduces-volume-renaming-restart-command-and-robust-bugfixes/

PHP 8.4.25 and 8.5.10 release candidates drop with serious stack-overflow hardening:
https://www.linuxcompatible.org/story/php-8425-and-8510-release-candidates-ship-with-stack-hardening/

PHP 8.6.0 Beta 1 Lands With Partial Function Application, Native Polling API, and a Hard Shift Toward Strict Errors:
https://www.linuxcompatible.org/story/php-860-beta-1-partial-function-application-native-polling-and-stricter-error-handling/

PostgreSQL 19 Beta 3 and Critical Security Patches Released; PG 14 EOL Notice Issued:
https://www.linuxcompatible.org/story/postgresql-19-beta-3-and-critical-security-patches-released-pg-14-eol-notice-issued/

GCC 17 Compiler Adding "-m128bit-atomic" For 128-bit Atomic Memory Operations:
https://www.phoronix.com/news/GCC-Git-m128bit-atomic

(more FOSS news in comments)

#WeeklyNews #FOSSNews #AI #Zed #Flutter #PHP #Podman #PostgreSQL #Dev #AMDGAIA #FastFlowLM #Coldcard #Bitcoin #Wireshark #GCC #ArtificialIntelligence #LLM #AgenticAI #Programming #Development #Coding #IDE #DBMS #SQL #FosseryTech

RT @TeksEdge: 🔥 Besitzer von Strix Halo 👉 Ihr habt einen weiteren lokalen KI-Beschleuniger in eurem Rechner, der ungenutzt bleibt. AMDs FastFlowLM kann unterstützte Modelle vollständig auf dem Ryzen AI XDNA2 NPU ausführen, ohne die Radeon-GPU zu nutzen. FastFlowLM ist wie Ollama, wurde aber speziell für AMDs NPU entwickelt. 👀 ⚡ LLMs 👁️ Vision 🎙️ Audio 🧩 Embeddings 🧠 MoE 📚 Bis zu 256K Kontext 🔋 Bis zu 10×+ Energieeffizienz 🪶 ~16MB Laufzeitzeit Unterstützte Ryzen AI-Chips umfassen: ✅ Strix Halo ✅ Strix Point ✅ Kraken Point ✅ Gorgon Point Euer Strix Halo verfügt bereits über: 🧠 CPU 🎮 Radeon 8060S GPU ⚡ XDNA2 NPU Die meisten lokalen KI-Arbeitslasten belasten die GPU und ignorieren die NPU weitgehend. Jetzt könntet ihr folgendes tun: ⚡ NPU → kleine LLM-/Vision-/Sprach-Arbeitslast 🎮 GPU → große lokale LLMs Derselbe PC. Zwei unabhängige KI-Beschleuniger arbeiten gleichzeitig. 🎯 Strix Halo ist nicht nur eine 128GB-Box mit einheitlichem GPU-Speicher. Dieser NPU wird zunehmend zu nützlicher KI-Hardware für den lokalen Einsatz.

mehr auf Arint.info

#AMD #FastFlowLM #LocalAI #NPU #RyzenAI #StrixHalo #arint_info

https://x.com/TeksEdge/status/2087335758231122143#m

RT @TeksEdge: 🔥 Besitzer von Strix Halo 👉 Ihr habt einen weiteren lokalen KI-Beschleuniger in eurem Rechner, der ungenutzt bleibt. AMDs FastFlowLM kann unterstützte Modelle vollständig auf dem Ryzen AI XDNA2 NPU ausführen, ohne die Radeon-GPU zu nutzen. FastFlowLM ist wie Ollama, wurde aber speziell für AMDs NPU entwickelt. 👀 ⚡ LLMs 👁️ Vision 🎙️ Audio 🧩 Embeddings 🧠 MoE 📚 Bis zu 256K Kontext 🔋 Bis zu 10×+ Energieeffizienz 🪶 ~16MB Laufzeitzeit Unterstützte Ryzen AI-Chips umfassen: ✅ Strix Halo ✅ Strix Point ✅ Kraken Point ✅ Gorgon Point Euer Strix Halo verfügt bereits über: 🧠 CPU 🎮 Radeon 8060S GPU ⚡ XDNA2 NPU Die meisten lokalen KI-Arbeitslasten belasten die GPU und ignorieren die NPU weitgehend. Jetzt könntet ihr folgendes tun: ⚡ NPU → kleine LLM-/Vision-/Sprach-Arbeitslast 🎮 GPU → große lokale LLMs Derselbe PC. Zwei unabhängige KI-Beschleuniger arbeiten gleichzeitig. 🎯 Strix Halo ist nicht nur eine 128GB-Box mit einheitlichem GPU-Speicher. Dieser NPU wird zunehmend zu nützlicher KI-Hardware für den lokalen Einsatz.

mehr auf Arint.info

#AMD #FastFlowLM #LocalAI #NPU #RyzenAI #StrixHalo #arint_info

https://x.com/TeksEdge/status/2087335758231122143#m

RT @TeksEdge: 🔥 Besitzer von Strix Halo 👉 ihr habt einen weiteren Local-AI-Beschleuniger in eurem Rechner, der ungenutzt bleibt. AMDs FastFlowLM kann unterstützte Modelle vollständig auf dem Ryzen AI XDNA2-NPU ausführen, ohne die Radeon-GPU zu nutzen. FastFlowLM funktioniert ähnlich wie Ollama, ist jedoch speziell für AMDs NPU entwickelt. 👀 ⚡ LLMs 👁️ Vision 🎙️ Audio 🧩 Embeddings 🧠 MoE 📚 Bis zu 256K Kontext 🔋 Bis zu 10×+ Energieeffizienz 🪶 ~16MB Laufzeitzeit Unterstützte Ryzen AI-Chips umfassen: ✅ Strix Halo ✅ Strix Point ✅ Kraken Point ✅ Gorgon Point Euer Strix Halo verfügt bereits über: 🧠 CPU 🎮 Radeon 8060S-GPU ⚡ XDNA2-NPU Die meisten Local-AI-Workloads belasten die GPU und ignorieren die NPU weitgehend. Jetzt könnt ihr potenziell Folgendes tun: ⚡ NPU → kleine LLM-/Vision-/Sprach-Workloads 🎮 GPU → große lokale LLMs Derselbe PC. Zwei unabhängige KI-Beschleuniger arbeiten gleichzeitig. 🎯 Strix Halo ist nicht nur eine 128GB-Box mit einheitlichem GPU-Speicher. Dieser NPU wird nun auch als nützliche KI-Hardware für den lokalen Einsatz relevant.

mehr auf Arint.info

#AMD #FastFlowLM #LocalAI #NPU #RyzenAI #StrixHalo #arint_info

https://x.com/TeksEdge/status/2087335758231122143#m

RT @TeksEdge: 🔥 Besitzer von Strix Halo 👉 ihr habt einen weiteren Local-AI-Beschleuniger in eurem Rechner, der ungenutzt bleibt. AMDs FastFlowLM kann unterstützte Modelle vollständig auf dem Ryzen AI XDNA2-NPU ausführen, ohne die Radeon-GPU zu nutzen. FastFlowLM funktioniert ähnlich wie Ollama, ist jedoch speziell für AMDs NPU entwickelt. 👀 ⚡ LLMs 👁️ Vision 🎙️ Audio 🧩 Embeddings 🧠 MoE 📚 Bis zu 256K Kontext 🔋 Bis zu 10×+ Energieeffizienz 🪶 ~16MB Laufzeitzeit Unterstützte Ryzen AI-Chips umfassen: ✅ Strix Halo ✅ Strix Point ✅ Kraken Point ✅ Gorgon Point Euer Strix Halo verfügt bereits über: 🧠 CPU 🎮 Radeon 8060S-GPU ⚡ XDNA2-NPU Die meisten Local-AI-Workloads belasten die GPU und ignorieren die NPU weitgehend. Jetzt könnt ihr potenziell Folgendes tun: ⚡ NPU → kleine LLM-/Vision-/Sprach-Workloads 🎮 GPU → große lokale LLMs Derselbe PC. Zwei unabhängige KI-Beschleuniger arbeiten gleichzeitig. 🎯 Strix Halo ist nicht nur eine 128GB-Box mit einheitlichem GPU-Speicher. Dieser NPU wird nun auch als nützliche KI-Hardware für den lokalen Einsatz relevant.

mehr auf Arint.info

#AMD #FastFlowLM #LocalAI #NPU #RyzenAI #StrixHalo #arint_info

https://x.com/TeksEdge/status/2087335758231122143#m

(more Linux and FOSS news in previous posts of thread)

GrapheneOS update: new exec spawning, app toggle, July 2026 security patch:
https://alternativeto.net/news/2026/7/grapheneos-update-new-exec-spawning-app-toggle-july-2026-security-patch/

OPNsense 26.7 debuts on FreeBSD 15.1 with firewall, networking & future-proof enhancements:
https://alternativeto.net/news/2026/7/opnsense-26-7-debuts-on-freebsd-15-1-with-firewall-networking-and-future-proof-enhancements/

FreeBSD Laptop Support Continues Improving With WiFi, GPU & Audio Driver Work:
https://www.phoronix.com/news/FreeBSD-Laptops-June-2026

FreeBSD Intern Working On Porting AMD ROCm To The BSD World:
https://www.phoronix.com/news/FreeBSD-Intern-AMD-ROCm

FreeBSD Desktop Installer Option Working Through NVIDIA Driver Handling, Licensing:
https://www.phoronix.com/news/FreeBSD-Desktop-Install-NVIDIA

GNU Hurd Makes Progress On AArch64, Writing Translators In Rust:
https://www.phoronix.com/news/GNU-Hurd-Q2-2026

scrcpy 4.1 launches with VP8/VP9 video encoding, FFmpeg 8.1, and several enhancements:
https://alternativeto.net/news/2026/7/scrcpy-4-1-launches-with-vp8-vp9-video-encoding-ffmpeg-8-1-and-several-enhancements/

Mesa 26.2.0-rc1 and 26.1.5 Released: OpenCL 3.1 Support, Vulkan Updates, and Stability Fixes:
https://www.linuxcompatible.org/story/mesa-2620rc1-and-2615-released-opencl-31-support-vulkan-updates-and-stability-fixes/

FastFlowLM Joins AMD to Advance AI Inference:
https://www.amd.com/en/blogs/2026/fastflowlm-joins-amd-to-advance-ai-inference.html

ROCm 7.14: TheRock Goes Production and Expands AMD’s AI Software Platform:
https://rocm.blogs.amd.com/ecosystems-and-partners/rocm-7.14-blog/README.html

Open Book Touch is Crowdfunding: A Buttonless, Open Hardware Answer to Kindle:
https://feed.itsfoss.com/link/24361/17379351/open-book-touch-crowdfunding

#WeeklyNews #OpenSource #FOSSNews #FOSS #OpenSourceNews #News #GrapheneOS #OPNSense #FreeBSD #GNUHurd #BSD #Scrcpy #Mesa #FastFlowLM #ROCm #OpenBookTouch #FosseryTech

(more Linux and FOSS news in previous posts of thread)

GrapheneOS update: new exec spawning, app toggle, July 2026 security patch:
https://alternativeto.net/news/2026/7/grapheneos-update-new-exec-spawning-app-toggle-july-2026-security-patch/

OPNsense 26.7 debuts on FreeBSD 15.1 with firewall, networking & future-proof enhancements:
https://alternativeto.net/news/2026/7/opnsense-26-7-debuts-on-freebsd-15-1-with-firewall-networking-and-future-proof-enhancements/

FreeBSD Laptop Support Continues Improving With WiFi, GPU & Audio Driver Work:
https://www.phoronix.com/news/FreeBSD-Laptops-June-2026

FreeBSD Intern Working On Porting AMD ROCm To The BSD World:
https://www.phoronix.com/news/FreeBSD-Intern-AMD-ROCm

FreeBSD Desktop Installer Option Working Through NVIDIA Driver Handling, Licensing:
https://www.phoronix.com/news/FreeBSD-Desktop-Install-NVIDIA

GNU Hurd Makes Progress On AArch64, Writing Translators In Rust:
https://www.phoronix.com/news/GNU-Hurd-Q2-2026

scrcpy 4.1 launches with VP8/VP9 video encoding, FFmpeg 8.1, and several enhancements:
https://alternativeto.net/news/2026/7/scrcpy-4-1-launches-with-vp8-vp9-video-encoding-ffmpeg-8-1-and-several-enhancements/

Mesa 26.2.0-rc1 and 26.1.5 Released: OpenCL 3.1 Support, Vulkan Updates, and Stability Fixes:
https://www.linuxcompatible.org/story/mesa-2620rc1-and-2615-released-opencl-31-support-vulkan-updates-and-stability-fixes/

FastFlowLM Joins AMD to Advance AI Inference:
https://www.amd.com/en/blogs/2026/fastflowlm-joins-amd-to-advance-ai-inference.html

ROCm 7.14: TheRock Goes Production and Expands AMD’s AI Software Platform:
https://rocm.blogs.amd.com/ecosystems-and-partners/rocm-7.14-blog/README.html

Open Book Touch is Crowdfunding: A Buttonless, Open Hardware Answer to Kindle:
https://feed.itsfoss.com/link/24361/17379351/open-book-touch-crowdfunding

#WeeklyNews #OpenSource #FOSSNews #FOSS #OpenSourceNews #News #GrapheneOS #OPNSense #FreeBSD #GNUHurd #BSD #Scrcpy #Mesa #FastFlowLM #ROCm #OpenBookTouch #FosseryTech

(more Linux and FOSS news in previous posts of thread)

GrapheneOS update: new exec spawning, app toggle, July 2026 security patch:
https://alternativeto.net/news/2026/7/grapheneos-update-new-exec-spawning-app-toggle-july-2026-security-patch/

OPNsense 26.7 debuts on FreeBSD 15.1 with firewall, networking & future-proof enhancements:
https://alternativeto.net/news/2026/7/opnsense-26-7-debuts-on-freebsd-15-1-with-firewall-networking-and-future-proof-enhancements/

FreeBSD Laptop Support Continues Improving With WiFi, GPU & Audio Driver Work:
https://www.phoronix.com/news/FreeBSD-Laptops-June-2026

FreeBSD Intern Working On Porting AMD ROCm To The BSD World:
https://www.phoronix.com/news/FreeBSD-Intern-AMD-ROCm

FreeBSD Desktop Installer Option Working Through NVIDIA Driver Handling, Licensing:
https://www.phoronix.com/news/FreeBSD-Desktop-Install-NVIDIA

GNU Hurd Makes Progress On AArch64, Writing Translators In Rust:
https://www.phoronix.com/news/GNU-Hurd-Q2-2026

scrcpy 4.1 launches with VP8/VP9 video encoding, FFmpeg 8.1, and several enhancements:
https://alternativeto.net/news/2026/7/scrcpy-4-1-launches-with-vp8-vp9-video-encoding-ffmpeg-8-1-and-several-enhancements/

Mesa 26.2.0-rc1 and 26.1.5 Released: OpenCL 3.1 Support, Vulkan Updates, and Stability Fixes:
https://www.linuxcompatible.org/story/mesa-2620rc1-and-2615-released-opencl-31-support-vulkan-updates-and-stability-fixes/

FastFlowLM Joins AMD to Advance AI Inference:
https://www.amd.com/en/blogs/2026/fastflowlm-joins-amd-to-advance-ai-inference.html

ROCm 7.14: TheRock Goes Production and Expands AMD’s AI Software Platform:
https://rocm.blogs.amd.com/ecosystems-and-partners/rocm-7.14-blog/README.html

Open Book Touch is Crowdfunding: A Buttonless, Open Hardware Answer to Kindle:
https://feed.itsfoss.com/link/24361/17379351/open-book-touch-crowdfunding

#WeeklyNews #OpenSource #FOSSNews #FOSS #OpenSourceNews #News #GrapheneOS #OPNSense #FreeBSD #GNUHurd #BSD #Scrcpy #Mesa #FastFlowLM #ROCm #OpenBookTouch #FosseryTech

(more Linux and FOSS news in previous posts of thread)

GrapheneOS update: new exec spawning, app toggle, July 2026 security patch:
https://alternativeto.net/news/2026/7/grapheneos-update-new-exec-spawning-app-toggle-july-2026-security-patch/

OPNsense 26.7 debuts on FreeBSD 15.1 with firewall, networking & future-proof enhancements:
https://alternativeto.net/news/2026/7/opnsense-26-7-debuts-on-freebsd-15-1-with-firewall-networking-and-future-proof-enhancements/

FreeBSD Laptop Support Continues Improving With WiFi, GPU & Audio Driver Work:
https://www.phoronix.com/news/FreeBSD-Laptops-June-2026

FreeBSD Intern Working On Porting AMD ROCm To The BSD World:
https://www.phoronix.com/news/FreeBSD-Intern-AMD-ROCm

FreeBSD Desktop Installer Option Working Through NVIDIA Driver Handling, Licensing:
https://www.phoronix.com/news/FreeBSD-Desktop-Install-NVIDIA

GNU Hurd Makes Progress On AArch64, Writing Translators In Rust:
https://www.phoronix.com/news/GNU-Hurd-Q2-2026

scrcpy 4.1 launches with VP8/VP9 video encoding, FFmpeg 8.1, and several enhancements:
https://alternativeto.net/news/2026/7/scrcpy-4-1-launches-with-vp8-vp9-video-encoding-ffmpeg-8-1-and-several-enhancements/

Mesa 26.2.0-rc1 and 26.1.5 Released: OpenCL 3.1 Support, Vulkan Updates, and Stability Fixes:
https://www.linuxcompatible.org/story/mesa-2620rc1-and-2615-released-opencl-31-support-vulkan-updates-and-stability-fixes/

FastFlowLM Joins AMD to Advance AI Inference:
https://www.amd.com/en/blogs/2026/fastflowlm-joins-amd-to-advance-ai-inference.html

ROCm 7.14: TheRock Goes Production and Expands AMD’s AI Software Platform:
https://rocm.blogs.amd.com/ecosystems-and-partners/rocm-7.14-blog/README.html

Open Book Touch is Crowdfunding: A Buttonless, Open Hardware Answer to Kindle:
https://feed.itsfoss.com/link/24361/17379351/open-book-touch-crowdfunding

#WeeklyNews #OpenSource #FOSSNews #FOSS #OpenSourceNews #News #GrapheneOS #OPNSense #FreeBSD #GNUHurd #BSD #Scrcpy #Mesa #FastFlowLM #ROCm #OpenBookTouch #FosseryTech

Chạy Qwen3-VL-4B-Instruct độc quyền trên AMD Ryzen AI NPU với FastFlowLM, mang lại hiệu suất nhanh hơn 10 lần và tiết kiệm năng lượng. Hỗ trợ độ dài ngữ cảnh lên đến 256k tokens. #FastFlowLM #AMD #RyzenAINPU #Qwen3VL #AI #NPU #TiếtKiệmNăngLượng #HiệuSuấtCao #TríTuệNhânTạo

https://www.reddit.com/r/LocalLLaMA/comments/1ok2f23/running_qwen3vl4binstruct_exclusively_on_amd/

"Chạy Qwen3-VL-4B-Instruct Exclusively trên AMD Ryzen AI NPU! Tốc độ và hiệu suất cao hơn 10 lần, hỗ trợ độ dài ngữ cảnh lên đến 256k token. #FastFlowLM #Qwen3VL #AMD #RyzenAINPU #AI #MachineLearning #TríTuệNhânTạo"

https://www.reddit.com/r/LocalLLaMA/comments/1ok2f23/running_qwen3vl4binstruct_exclusively_on_amd/

Volt FastFlowLM (FLM) chạy các mô hình AI (Whisper, GPT-OSS, Qwen3...) toàn bộ trên AMD Ryzen AI NPU, không cần GPU. Là công nghệ ưu青睐 hiệu suất, 10x tiết kiệm điện, hỗ trợ ngữ cảnh 256k token. CL-shaped & Server Mode tương thích OpenAI. Nhỏ gọn 16MB, cài 20s. Thử ngay tại GitHub & YouTube Demo. #AI #Tech #AMDNPUT #FastFlowLM #Whisper #GPTOSS #MoE #NLPhack #FAI

https://youtu.be/0t8ijUPg4A0?si=539G5mrICJNOwe6Z

FastFlowLM (FLM) - một runtime siêu tốc cho GPT-OSS, Gemma3, Qwen3, DeepSeek-R1, LLaMA3.x trên AMD Ryzen AI NPU! 🚀

⚡️ Nhanh hơn 10 lần, siêu nhẹ (14MB), cài đặt trong 20 giây. Hỗ trợ context dài đến 256k tokens!

#AI #MachineLearning #RyzenAI #FastFlowLM #TríTuệNhânTạo #HọcMáy #AMD

https://youtu.be/ksYyiUQvYfo?si=zfBjb7U86P947OYW

📬 FastFlowLM unter der Lupe: Lokale KIs im Test
#KünstlicheIntelligenz #Softwareentwicklung #FastFlowLM #FFLM #HuggingFace #NeuralProcessorUnit #NPU #WebUI https://sc.tarnkappe.info/0bf527

📬 FastFlowLM unter der Lupe: Lokale KIs im Test
#KünstlicheIntelligenz #Softwareentwicklung #FastFlowLM #FFLM #HuggingFace #NeuralProcessorUnit #NPU #WebUI https://sc.tarnkappe.info/0bf527

📬 FastFlowLM unter der Lupe: Lokale KIs im Test
#KünstlicheIntelligenz #Softwareentwicklung #FastFlowLM #FFLM #HuggingFace #NeuralProcessorUnit #NPU #WebUI https://sc.tarnkappe.info/0bf527

📬 FastFlowLM unter der Lupe: Lokale KIs im Test
#KünstlicheIntelligenz #Softwareentwicklung #FastFlowLM #FFLM #HuggingFace #NeuralProcessorUnit #NPU #WebUI https://sc.tarnkappe.info/0bf527