Science & Tech · AI · 3 hrs ago
Intel unveils AI accelerator with up to 480GB of memory
Intel has introduced the D270, a data-center accelerator designed to run AI models rather than display conventional 3D graphics.
It is part of the company’s new GPU D-series and is based on Intel’s Xe3P architecture.
Intel’s reference card has 160GB of memory, while versions made by manufacturing partners can offer up to 480GB.
The card uses LPDDR5X memory, which Intel says helps provide capacity for large language models and long conversations.
It can deliver peak FP8 performance of 1.3 petaflops and is designed to use air cooling in standard server systems.
Intel says one card can handle models with roughly 300 billion parameters, while an eight-card server can support models with more than one trillion.
Customer sampling is underway, but Intel has not announced pricing or when the accelerator will be generally available.
Intel has unveiled the GPU D270, an accelerator designed for AI inference rather than 3D graphics.
The card can be configured with up to 480GB of LPDDR5X memory and delivers peak FP8 performance of 1.3 petaflops.
Intel’s reference card has 160GB of memory, while manufacturing partners can develop versions with more.
The D270 uses a 350W air-cooled design intended to fit conventional server infrastructure.
Intel says customer sampling is underway, but it has not announced pricing or a general availability date.
- Who
- Intel unveiled the GPU D270, its first GPU D-series accelerator.
- What
- It is a data-center accelerator for AI inference, with up to 480GB of LPDDR5X memory.
- When
- Not stated
- Where
- It is designed for data centers and conventional server infrastructure.
- Why
- It targets AI inference workloads, including large language models and long-context tasks.
This story does not have two clearly opposing sides.
This story does not have a timeline yet.
- Maximum memory
- Up to 480GB of LPDDR5X
- Reference-card memory
- 160GB of LPDDR5X
- Peak FP8 performance
- 1.3 petaflops
- Power target
- 350W
- Memory bandwidth
- 1.5TB/s
- Architecture
- Xe3P
Sources
Related news
Microsoft president calls for global AI safety rules
Meta releases Muse Wonder, its first frontier AI model
Meta unveils two AI models it says rival leading competitors
Meta launches Muse models, with Muse Wonder ranked second in performance evaluation
Test taker caught trying to use AI smart glasses in Yonsei exam
Anthropic AI submits false murder tip to Philadelphia police
Bodyfriend introduces heat and sleep care products for seasonal change
South Korea’s OGP team applies startup methods inside government
Microsoft plans to deploy Maia 300 AI chip from 2027
China faces difficulty moving tech firms away from Nvidia chips
Nadella calls for human-controlled emergency brakes on advanced AI
