AMD Instinct MI350P for Enterprise AI Adoption

Excited to see Enterprise AI solutions powered by MI350P successfully delivered by the team. This milestone reflects the incredible engineering execution, cross-functional collaboration, and relentless focus required to bring next-generation AI infrastructure to customers at scale. As AI adoption accelerates across industries, delivering high-performance, efficient, and scalable platforms becomes increasingly critical. Proud of the team’s contribution in helping advance the AI ecosystem and enabling the future of enterprise AI computing. Congratulations to everyone involved in making this happen! Learn more: https://bit.ly/4deZM2L

AMD Instinct MI350P PCIe GPUs are designed to help #enterprises run AI inference on existing air-cooled infrastructure, without rebuilding around dedicated accelerator platforms.

Meet the #16T1 The WOLF-16T1 (VPX3U-THOR-SBC) is a rugged #3UVPX SBC built for next-generation #C5ISR and AI-enabled mission systems. Powered by NVIDIA® IGX Thor™, it delivers up to 2070 FP4 TFLOPS of AI compute with an embedded Blackwell GPU, 128 GB ECC LPDDR5X memory, TSN-enabled networking, and a functional safety-capable PCIe Gen4 switch for secure, high-bandwidth edge processing. In the Wild Deploy the #16T1 in robotics, sensor fusion, autonomy, and safety-critical edge AI platforms where ruggedization, security, and real-time performance are essential. Operating from -40°C to +85°C and built to withstand shock and vibration, the 16T1 brings advanced AI compute to rugged SOSA®-aligned VPX systems. #wolftracks #wolfadvancedtechnology #3U #igxthor

A new PCIe‑based AI accelerator card has been revealed, featuring 384 GB of onboard memory and a 240 W power envelope, capable of running massive 700‑billion‑parameter models. This new design shows growing interest in high‑memory accelerators outside traditional GPU form factors. #AIAccelerators #PCIe #AIHardware #Datacenters #GPU #GPUs #Innovation #Technology #TechnologyNews #AI https://lnkd.in/d-CP3q6G

Skymizer Unveils Low-Power PCIe AI Accelerator for Local LLM Inference 📌 Skymizer is challenging industry giants with the HTX301, a low-power PCIe accelerator capable of running 700B parameter models locally. By leveraging efficient LPDDR4/5 memory and a 28nm process, this device offers an energy-efficient alternative to power-hungry GPU clusters, consuming just 240W. It provides enterprises a cost-effective way to maintain data sovereignty without expensive infrastructure overhauls. 🔗 Read more: https://lnkd.in/ePGp4RwD #Skymizer #Htx301 #Pcieaccelerator #Llminference #Largelanguagemodels

In a smart factory, AI vision systems are inspecting products in real time. In retail, analytics platforms are processing video streams to understand customer behavior. In development labs, engineers are running increasingly complex AI models locally. Our industrial motherboards, supporting AMD EPYC™ 4005 and Ryzen™ 9000 Series processors, are built to meet these demands across real-world deployments. From full-scale ATX platforms powering AI-driven analytics, to compact Mini-ITX systems enabling edge inference in space-constrained environments, the lineup delivers the performance and expandability needed for modern workloads, backed by DDR5, PCIe Gen5, and rich industrial I/O Read more: https://lnkd.in/gCU2NwsU #ASRockIndustrial #AMD

Tech tip: If you don't want to risk melting your power cables or flipping a breaker, you can set a max wattage for your GPUs using nvidia-smi. If you are doing AI work with a consumer GPU it will use less than half the max TDP. This is because half the chip is sitting dark: texture units, RT cores, render ops, video output. Don't believe the hype about "5090's need 650 watts" example: # Set to 400W sudo nvidia-smi -pl 400

▶️ Watch Now: Blackwell GPUs + Premio's Rugged Edge Systems Performance, validation, and design—explained What does a major leap in GPU performance actually mean in the real world? In our latest video, Premio breaks down what’s new with NVIDIA RTX PRO Blackwell GPUs—and how that performance translates into real-world AI outcomes. From physical AI to agentic AI, see how you can now deploy faster, more intelligent systems at the edge with confidence. What You’ll Learn: ✅ What’s New in Blackwell: Key performance gains and architecture advancements ✅ Real-World Impact: How higher compute unlocks smarter AI decisions at the edge ✅ Validated Systems: Premio GPU computers designed and tested for Blackwell GPUs ✅ Premio’s Rugged Design: Optimized for high-power GPU performance ▶️Watch the full video: https://lnkd.in/gA6WakTB

Skymizer, a Taiwanese company, has introduced the innovative HTX301 PCIe AI accelerator card, which impressively runs language models with up to 700 billion parameters while consuming only 240 watts of power. This technological feat is accomplished using decade-old 28-nanometer chips and LPDDR4 and LPDDR5 memory instead of more costly memory options like HBM or GDDR. The HTX301 relies on the HyperThought platform and employs advanced compression techniques, significantly outperforming open-source alternatives. Positioned to disrupt the industry, Skymizer aims to offer a cost-effective, on-premises solution that could eliminate the necessity for large-scale GPU clusters. #Ai #Hardware #Innovation https://lnkd.in/g5PTDVTE

AI workloads are getting smaller, faster, and increasingly local. 🚀 The new MINISFORUM M2 brings high-performance computing and local AI capability into a compact 520g mini PC. Powered by Intel® Core™ Ultra 7 356H (Core Ultra Series 3 / "Panther Lake" Series), M2 delivers up to 90 TOPS of AI performance and verified local deployment for mainstream open-source models including Qwen3.5-35B and Gemma-4-26B. 👏 Tested local inference performance includes: • Qwen3.5-35B-A3B — 23.34 tok/s • GPT-OSS 20B — 21.08 tok/s • Gemma-4-26B-A4B — 22.25 tok/s Key highlights: • Support up to 128GB DDR5 memory • Dual PCIe 4.0 SSD slots • Dual 2.5GbE • Wi-Fi 7 & Bluetooth 5.4 • Triple 4K display output • USB4 connectivity From creators to homelabs and edge deployment, M2 shows how capable modern mini PCs have become in the era of local AI. 🔗 Learn more: https://lnkd.in/eK3g6R5Y #MINISFORUM #MiniPC #LocalAI #IntelPantherLake #EdgeComputing #Homelab