Intel揭开了针对工作站市场的两个新的ARC Pro GPU，与NVIDIA在AI驱动的计算中竞争

新的Arc Pro B50和B60系列工作站GPU的起价分别为299美元和500美元，并适合寻求本地AI处理的客户

Intel Corporation has unveiled two new graphics processing units (GPUs) targeting the workstation market and setting the stage for renewed competition with Nvidia in AI-powered computing.

Intel Corporation推出了针对工作站市场的两个新的图形处理单元（GPU），并为与NVIDIA的重新竞争奠定了基础。

The tech giant estimates that entry-level workstation GPUs account for roughly 30% of the overall workstation market, and it believes the size warrants the need to increase its market presence.

这家科技巨头估计，入门级工作站GPU占整个工作站市场的30％，并且认为规模值得增加其市场业务。

The new Arc Pro B50 and B60 series workstation GPUs start at $299 and $500, respectively, and cater to customers looking for on-premise AI processing that does not rely on cloud providers like Amazon, Google, or Microsoft.

新的Arc Pro B50和B60系列工作站GPU的起价分别为299美元和500美元，并迎合了寻求本地AI处理的客户，而Amazon，Google或Microsoft不依赖云提供商。

Intel has been developing its Arc GPU architecture for several years, initially releasing B-series chips named Battlemage for gamers and an A-series tailored toward enterprise-level graphics tasks.

英特尔已经开发其ARC GPU体系结构已有几年了，最初为游戏玩家发布了名为Battlemage的B系列芯片，以及针对企业级图形任务量身定制的A系列。

According to Intel, the newly announced B50 and B60 cards are designed for AI workloads due to their enhanced memory and improved AI-specific performance metrics.

根据英特尔的说法，新宣布的B50和B60卡是为AI工作负载而设计的，因为它们的内存增强并改善了特定于AI的性能指标。

The Arc Pro B50, the mainstream AI workstation chip, comes with 16GB of memory and delivers up to 170 trillion operations per second (TOPS). Intel reiterated that this GPU is for AI-assisted software engineering and other design-centric workflows.

ARC Pro B50是主流AI工作站芯片，具有16GB的内存，每秒可提供高达170万亿的操作（顶部）。英特尔重申，此GPU适用于AI辅助软件工程和其他以设计为中心的工作流程。

The more powerful Arc Pro B60 has 24GB of memory and can handle up to 197 TOPS for customers running highly demanding AI applications. Intel released benchmark comparisons that showed performance gains over its previous-generation A50 6GB GPU and Nvidia’s competing RTX A1000 8GB.

功率更强大的Arc Pro B60具有24GB的内存，对于运行高度要求的AI应用程序的客户，最多可以处理197个上层。英特尔发布了基准比较，该比较显示了其前一代A50 6GB GPU和NVIDIA竞争的RTX A1000 8GB的性能提高。

Per the results, the new Arc Pro B50 delivers up to 3.4 times the performance in graphics workloads compared to the A50, along with improvements over the RTX A1000 across various tasks. Intel also reported similar performance advantages in multiple AI inference benchmarks.

根据结果​​，与A50相比，新的Arc Pro B50的性能最高是图形工作负载的3.4倍，并且在各种任务中对RTX A1000的改进。英特尔还报告了多个AI推理基准测试的性能优势。

“We expect this to be very competitive and a very attractive solution, particularly for design and engineering,” said Vivian Lien, Intel’s vice president and general manager of client graphics, in a press conference.

“我们希望这将非常有竞争力，并且是一个非常有吸引力的解决方案，尤其是对于设计和工程，”英特尔留置权公司的客户图形副总裁兼总经理Vivian Lien在新闻发布会上说。

Nvidia continues to lead the market with its RTX Pro AI chips and workstation Nvidia’s DGX Spark and DGX Station desktop machines. The chip manufacturer describes DGX Spark as a “compact supercomputer” powered by the Grace Blackwell GB10 chip.

NVIDIA继续使用其RTX Pro AI芯片和Workstation Nvidia的DGX Spark和DGX站台式机机器领导市场。芯片制造商将DGX Spark描述为由Grace Blackwell GB10芯片提供动力的“紧凑型超级计算机”。

The DGX Station, outfitted with the more advanced GB300 Grace Blackwell Ultra, provides up to 784 GB of memory, which dwarfs Intel’s current offerings.

DGX站配备了更高级的GB300 Grace Blackwell Ultra，可提供多达784 GB的内存，使Intel的当前产品相形见war。

At Computex Taipei on Monday, Nvidia made several announcements, including advancements in humanoid robotics and expanded access to its NVLink technology.

周一，NVIDIA在Computex台北上发表了几次公告，包括人类机器人技术的进步以及扩大了对其NVLink技术的访问。

NVLink helps businesses connect with semi-custom AI servers leveraging Nvidia’s GPU infrastructure. It was developed to enable high-speed data exchange between chips and is used in its GB200 system, which integrates two Blackwell GPUs with a Grace CPU.

NVLink帮助企业与利用NVIDIA的GPU基础架构的半定期AI服务器联系。它的开发是为了启用芯片之间的高速数据交换，并在其GB200系统中使用，该系统将两个Blackwell GPU与Grace CPU集成在一起。

“NV link fusion is so that you can build semi-custom AI infrastructure, not just semi-custom chips,” the CEO reckoned.

首席执行官认为，“ NV Link Fusion是为了使您可以构建半定期AI基础架构，而不仅仅是半定期芯片。”

CEO Jensen Huang called the physical AI sector the “next trillion-dollar industry.” He asserted that the company is currently focused on software platforms designed to train and operate humanoid robots in industrial settings.

首席执行官詹森黄（Jensen Huang）称实体AI部门为“下一个万亿美元的行业”。他断言，该公司目前专注于旨在在工业环境中培训和操作人形机器人的软件平台。

During President Donald Trump’s visit to Saudi Arabia last week, the company also made headlines, announcing plans to supply several hundred thousand AI processors over five years to Humain, an AI startup owned by the Saudi sovereign wealth fund.

在上周唐纳德·特朗普（Donald Trump）总统访问沙特阿拉伯期间，该公司还成为头条新闻，宣布计划在五年内为数十万AI处理器提供劳恩（Humain），这是由沙特主权财富基金拥有的AI AI初创公司。

Nvidia also unveiled DGX Cloud Lepton, an AI platform featuring a compute marketplace that connects developers to tens of thousands of GPUs via a global network of cloud providers.

NVIDIA还揭开了DGX Cloud Lepton，这是一个AI平台，该平台采用计算市场，通过全球云提供商网络将开发人员连接到数以万计的GPU。

During his keynote at Computex, Huang announced the opening of a new office in Taiwan and a partnership with Foxconn to build an AI supercomputer in the country.

在Computex的主题演讲中，黄宣布在台湾开设了一个新办公室，并与富士康建立了合作伙伴关系，以在该国建立AI超级计算机。

“We are delighted to partner with Foxconn and Taiwan to help build Taiwan’s AI infrastructure and to support TSMC and other leading companies to advance innovation in the age of AI and robotics,” Huang concluded.

Huang总结说：“我们很高兴与Foxconn和Taiwan合作，以帮助建立台湾的AI基础设施，并支持TSMC和其他领先公司，以推动AI和机器人时代的创新。”