Intel D270 GPU AI Accelerator–Up to 480 GB LPDDR5X

來源原標題

Intel D270 GPU AI Accelerator–Up to 480 GB LPDDR5X

作者

hurricanepootis

原始標籤

hackernews

來源訊號

HN points 301　HN 留言 175

原始連結

https://www.intel.com/content/www/us/en/products/details/discrete-gpus/data-center-gpu/d-series.html

摘要原文

Intel® GPU D-series Scale inference, not infrastructure Intel® GPU D270 │ Inference-Optimized Data Center GPU Overview Use cases Features Get started Agentic AI is changing data center workloads Agents read far more than they write: long contexts, multistep workflows, compute dominated by prefill. As demands grow, challenges expand beyond bandwidth. Intel D-series GPUs are built for this new era of agentic AI, efficiently handling larger contexts, more concurrent sessions, and more tokens within practical power and cost limits. Memory up to 480 GB LPDDR5X per card Power 350 W Air-cooled Compute 32 Xᵉ Xᵉ3P cores Compute 256 XMX Xᵉ3P matrix engines Scaling agentic AI Agentic fleets Over thousands of steps, 100K token contexts, and dozens of concurrent sessions, keep agents resident to help sustain throughput across the agentic loop. Reasoning over more context Whole-context reasoning Keep more of the model and long context resident, for richer reasoning with fewer GPUs and no high bandwidth memory (HBM). Economics of inference Efficient token serving Dense, air-cooled, standard PCIe infrastructure helps maximize tokens per rack and watt, scaling serving capacity and keeping costs in check. More inference within the power and racks you already run As agentic AI drives longer prompts and more concurrent sessions, customers need scalable infrastructure without trade-offs in power, cost, or deployment complexity. Intel GPU D270 offers up to 480 GB LPDDR5X memory for power-efficient, memory-bound inference in data center environments. Built for AI Built for agentic workloads Die area reallocated from graphics and ray‑tracing pipelines to Xᵉ3P cores and matrix compute, improving inference efficiency. Built for compute Compute where it matters More prompt processing per watt: 32 Xᵉ3P cores and 256 XMX engines accelerate prefill, curbing long‑context and agentic costs. Built for the workload Precision for every workload From native low‑precision inference to FP64 scientific computing, the workload determines the precision—not the hardware. Built for efficiency Standard PCIe 5.0 Standard form factor, standard interconnect, using existing servers. No proprietary networking or bespoke systems required. Built for reliable power 350W air-cooled Passive cooling in the racks you already run, with no liquid‑cooling retrofit needed between evaluation and deployment. Built for your data center Open by design Zero-day support for leading open models, served on the frameworks you already run: vLLM and the open PyTorch ecosystem. Workload-first compute for modern AI Deliver an intelligent evolution in data center infrastructure. Built for the new era of AI inference.  Sign up to learn more about AI and data center innovations Stay up to date on Intel GPU D-series news and the latest AI resources for enterprise. Thanks for signing up! You'll start receiving updates soon. Frequently Asked Questions What is Intel GPU D270? Intel GPU D270 is an inference-optimized data center GPU designed for the agentic AI era. It is built to support enterprise AI, SaaS, GPUaaS, and government or sovereign-cloud deployments that need memory-rich, power-efficient AI acceleration. What types of AI workloads is Intel GPU D270 designed to support? Intel GPU D270 is designed for production AI inference workloads such as agentic AI, long-context retrieval-augmented generation (RAG), enterprise AI assistants, coding assistants, document understanding, multimodal AI, diffusion models, and high-concurrency serving. Why is Intel GPU D270 relevant for agentic AI? Agentic AI often turns one task into multiple rounds of model calls, tool use, retrieval, reasoning, and response generation. Intel GPU D270 is designed for the inference requirements that become important in those environments, including prefill processing, larger context, memory capacity, concurrency, and power efficiency. What does “inference-optimized” mean for Intel GPU D270? Inference-optimized means the GPU is designed for running AI models in production rather than training them at massive scale. Intel GPU D270 focuses on the needs of deployed AI applications, including prompt processing, high-concurrency serving, memory capacity, efficient power use, and system-level performance. How does Intel GPU D270 help with long-context AI workloads? Long-context AI workloads can require more memory to support larger prompts, retrieval-heavy workflows, and KV cache growth during inference. Intel GPU D270 is designed with high available memory capacity to help support larger models, longer context, and more concurrent sessions. Can Intel GPU D270 be used in existing data center infrastructure? Intel GPU D270 is designed around a standard PCIe 5.0 form factor and air-cooled deployment profile, helping enterprises evaluate AI inference acceleration within existing data center environments without requiring a specialized liquid-cooling retrofit. Who is the primary audience for Intel GPU D270? The primary audience includes AI system architects, AI platform architects, infrastructure and cloud engineering leaders, enterprise AI decision-makers, SaaS providers, GPUaaS providers, government or sovereign-cloud teams, and developers deploying production inference workloads. What should customers do if they want to learn more about Intel GPU D270? Customers should connect with Intel to learn more about Intel GPU D270, evaluate how it may fit their AI inference roadmap, and receive updates as additional product information becomes available. Disclaimers Performance projections vs RTX Pro 6000 are based on publicly available information and Intel internal analysis as of Feb 2026. Performance varies by use, configuration, and other factors. Memory capacity comparison at comparable bandwidth. (301 points, 175 comments on Hacker News)