加速實現AI無所不在 英特爾以新Xeon 6、Gaudi加速器和Lunar Lake架構重新定義運算、效能和經濟效益，強化AI PC領導地位

June 4, 2024 Published
A man in a suit holds up a small electronic component on stage, with a futuristic blue and white background. The scene suggests a technology presentation or conference.

At Computex Taiwan, Intel CEO Pat Gelsinger discusses the vast opportunities ahead and the Intel innovation that’s bringing AI everywhere. During the event in June 2024, Intel unveiled cutting-edge technologies and architectures poised to accelerate the artificial intelligence ecosystem. (Credit: Intel Corporation)

英特爾全面布署資料中心、雲端、網路、邊緣與PC的全面運算領域實現最佳化AI運算效能

英特爾在2024台北國際電腦展（COMPUTEX）期間，展示橫跨資料中心、雲端、網路、邊緣運算和PC等領域的多項領先技術和架構。英特爾藉由強大運算效能、業界領先的能源效率，以及降低客戶總持有成本（TCO）等優勢，全面助攻客戶掌握AI商機。

新聞重點：

  • 推出Intel® Xeon® 6處理器：內建效率核心（E-core），針對資料中心的高密度、規模化工作負載進行最佳化，大幅提升效能與能源效率。Intel® Xeon® 6處理器讓機櫃密度提升三倍，機櫃級效能最高提升4.2倍，每瓦效能最高提升2.6倍1，滿足資料中心對效能和能耗的嚴苛要求。
  • 公布Intel® Gaudi® 2和Intel® Gaudi® 3 AI加速器套件定價：Intel® Gaudi®系列加速器不僅具備高效能，且成本低於其他平台的三分之一2。此外，將Xeon處理器與Gaudi AI加速器整合於單一系統，可提供更快速、更實惠且更普及的AI解決方案。
  • 為持續拓展AI PC產品領域，新一代AI PC採用x86架構突破性的能耗表現，為客戶提供廣大的應用相容性，同時大幅提升功耗效率。相較前一代產品，新世代Lunar Lake客戶端處理器架構的系統單晶片（SoC）可降低功耗最高達40%。

英特爾執行長Pat Gelsinger（基辛格）表示：「AI正引領業界邁入最具影響力的創新時代。晶片將再次帶動運算技術快速發展，不僅突破人類潛能，更將在未來數年驅動全球經濟成長。」

Pat Gelsinger進一步表示：「英特爾是全球少數能夠全面掌握AI市場契機，並持續推動創新的企業之一。從半導體製造到PC、網路、邊緣運算和資料中心等領域，英特爾都有全面佈局。最新推出的Xeon、Gaudi和Core Ultra平台，結合英特爾強大的硬體和軟體生態系，可提供靈活、安全、永續且具成本效益的解決方案，協助合作夥伴搶攻未來龐大的商機。」

英特爾實現AI無所不在

在COMPUTEX主題演講中，Pat Gelsinger強調開放標準和英特爾強大的生態系是加速實現AI應用的關鍵驅動力，此一觀點獲得宏碁董事長暨執行長陳俊聖、華碩董事長施崇棠、微軟董事長暨執行長Satya Nadella以及英業達總經理蔡枝安等人的共同支持。

過去六個月內，英特爾取得一系列重大突破。從公布第5代Intel® Xeon®處理器上市到推出Xeon 6系列首款產品、預告Gaudi AI加速器到提供企業客戶高效且具成本效益的生成式AI訓練及推理系統。英特爾也宣布目前已有超過800萬台裝置搭載Intel® Core™ Ultra處理器，顯示AI PC時代已然來臨。

因應此趨勢，英特爾揭露將於今年推出的新一代客戶端架構，並憑藉這一系列突破性進展，加速創新步伐、提升生產效率，進而縮短AI普及化進程，引領產業變革。

A silver Intel Xeon 6 processor is shown against a blue background. The front side displays the Intel logo and model name, while the other side reveals the intricate circuitry pattern.

At Computex in Taipei, Taiwan, on June 4, 2024, Intel launched the Intel Xeon 6 processors with Efficient-cores (E-cores). For companies looking to refresh aging infrastructure to help reduce costs and free up space, Intel Xeon 6 with E-cores offers significant rack density advantages, enabling a 3-to-1 rack-level consolidation. (Credit: Intel Corporation)

Image of a computer processor with a rectangular shape, featuring a central gold-colored section surrounded by a blue and green circuit board design, with various small components visible around the edges.

At Computex in Taipei, Taiwan, on June 4, 2024, Intel launched the Intel Xeon 6 processors with Efficient-cores (E-cores), which use less power and occupy less rack space in aging infrastructure compared with earlier generations. These Intel Xeon 6 processors with E-cores free up space for innovative new AI projects. (Credit: Intel Corporation)

An Intel Xeon 6 processor chip with a silver metallic surface and the Intel Xeon logo in black. The chip is rectangular with connectors on the top and bottom edges, set against a white background.

At Computex in Taipei, Taiwan, on June 4, 2024, Intel shared details on the new Intel Xeon 6 processors with Performance-cores (P-cores) and Efficient-cores (E-cores). The Intel Xeon 6 platform will give customers choice, addressing a broad array of use cases and workloads, from compute-intensive high performance computing and AI to scalable cloud-native applications. (Credit: Intel Corporation)

Close-up image of a computer processor with a blue and silver design. Numerous small components and circuits are visible on its surface, set against a dark background.

At Computex in Taipei, Taiwan, on June 4, 2024, Intel launched the Intel Xeon 6 processors with Efficient-cores (E-cores), which use less power and occupy less rack space in aging infrastructure compared with earlier generations. These Intel Xeon 6 processors with E-cores free up space for innovative new AI projects. (Credit: Intel Corporation)

A person wearing a brown hijab and white protective suit is seated at a workstation, inspecting electronic circuit boards. They are wearing blue gloves and surrounded by industrial equipment and metal racks.

A worker in the Intel Kulim Assembly Test facility in Kulim, Malaysia, inspects Intel Xeon 6 processors with Efficient-cores (code-named Sierra Forest) in May 2024. The first member of the Intel Xeon 6 processor family was introduced June 4, 2024, at Computex in Taipei, Taiwan. (Credit: Intel Corporation). (Credit: Intel Corporation)

Gloved hands holding a tray with four large computer processors, each with visible contact pins. The processors are neatly arranged in a black carrier on a light gray surface, showcasing the intricate design and technology.

A worker in the Intel Kulim Assembly Test facility in Kulim, Malaysia, inspects Intel Xeon 6 processors with Efficient-cores (code-named Sierra Forest) in May 2024. The first member of the Intel Xeon 6 processor family was introduced June 4, 2024, at Computex in Taipei, Taiwan. (Credit: Intel Corporation). (Credit: Intel Corporation)

A person wearing safety goggles, a mask, and blue gloves holds up a microchip in a cleanroom environment. The focus is on the microchip, with blurred background equipment and lighting visible.

A worker in the Costa Rica Assembly Test (CRAT) facility in San Jose, Costa Rica, displays an Intel Xeon 6 processor with Efficient-cores (code-named Sierra Forest) in May 2024. The first member of the Intel Xeon 6 processor family was introduced June 4, 2024, at Computex in Taipei, Taiwan. (Credit: Intel Corporation).

An overhead view of a silicon wafer with numerous glowing, grid-like patterns in orange and red hues, set against a white background. The wafer is circular and reflective, showcasing the intricate design of microchips.

A photo shows a wafer of Intel Xeon 6 processors with Efficient-cores (code-name Sierra Forest). Intel launched the first member of the Intel Xeon 6 processor family on June 4, 2024, at Computex in Taipei, Taiwan. Intel Xeon 6 with E-cores offers scalability and flexibility with a modular system on a chip architecture. (Credit: Intel Corporation)

An angled view of a silicon wafer, showing an array of rectangular microchips with a colorful reflective surface. The background is plain white.

A photo shows a wafer of Intel Xeon 6 processors with Efficient-cores (code-name Sierra Forest). Intel launched the first member of the Intel Xeon 6 processor family on June 4, 2024, at Computex in Taipei, Taiwan. Intel Xeon 6 with E-cores is optimized for higher performance per watt for throughput-oriented workloads like microservices as well as networking and media. (Credit: Intel Corporation)

A silver Intel Xeon 6 processor is shown against a blue background. The front side displays the Intel logo and model name, while the other side reveals the intricate circuitry pattern.
Image of a computer processor with a rectangular shape, featuring a central gold-colored section surrounded by a blue and green circuit board design, with various small components visible around the edges.
An Intel Xeon 6 processor chip with a silver metallic surface and the Intel Xeon logo in black. The chip is rectangular with connectors on the top and bottom edges, set against a white background.
Close-up image of a computer processor with a blue and silver design. Numerous small components and circuits are visible on its surface, set against a dark background.
A person wearing a brown hijab and white protective suit is seated at a workstation, inspecting electronic circuit boards. They are wearing blue gloves and surrounded by industrial equipment and metal racks.
Gloved hands holding a tray with four large computer processors, each with visible contact pins. The processors are neatly arranged in a black carrier on a light gray surface, showcasing the intricate design and technology.
A person wearing safety goggles, a mask, and blue gloves holds up a microchip in a cleanroom environment. The focus is on the microchip, with blurred background equipment and lighting visible.
An overhead view of a silicon wafer with numerous glowing, grid-like patterns in orange and red hues, set against a white background. The wafer is circular and reflective, showcasing the intricate design of microchips.
An angled view of a silicon wafer, showing an array of rectangular microchips with a colorful reflective surface. The background is plain white.

下載所有圖片 （ZIP，26 MB）

現代化資料中心的AI應用：Intel Xeon 6處理器為高密度、規模化工作負載帶來卓越效能與能源效率

隨著數位轉型加速發展，迫使企業升級傳統資料中心系統。企業不僅要降低成本、實現永續目標，還需要將實體空間的使用率最大化，同時創造全新數位能力。

英特爾專為因應現代化資料中心的挑戰而設計的Xeon 6平台和處理器家族，提供E-core和P-core兩種規格，可處理包含AI、高效能運算需求以及可擴充的雲端原生應用等廣泛應用情境和工作負載。E-core和P-core建立於相容架構上，不僅可共享軟體堆疊，還受益於開放的軟硬體供應商生態系統支援。

首款發布的Xeon 6處理器為Intel Xeon 6 E-core處理器（代號Sierra Forest），Xeon 6 P-core處理器（代號Granite Rapids）則預計於下一季度推出。

Intel Xeon 6 E-core處理器擁有高核心密度和卓越的每瓦效能，不僅可提供高效能運算，還可大幅降低能源成本。此款處理器在效能和能源效率方面的最佳化設計，可適用於雲端原生應用、內容傳遞網路、微服務架構和消費者數位服務等高密度、規模化工作負載環境中。

Intel Xeon 6 E-core處理器還具備高密度優勢，採用E-core的Xeon 6處理器讓機櫃密度提升三倍。與第二代Intel® Xeon®處理器在媒體轉碼工作負載的表現相比3，Intel Xeon 6 E-core處理器機櫃級效能提升4.2倍，每瓦效能提升2.6倍。藉由降低功耗、擴大機櫃使用空間等特點，Xeon 6處理器可為創新AI應用提供更強大的運算力和更高效的基礎架構。

An electronic circuit board featuring an Intel Gaudi processor at the center, surrounded by various components and connectors, with a blue and yellow color scheme.

Intel introduced the Intel Gaudi 3 AI accelerator on April 9, 2024, at the Intel Vision event in Phoenix, Arizona. The accelerator delivers 4x AI compute for BF16 and 1.5x increase in memory bandwidth compared with its predecessor. (Credit: Intel Corporation)

Image of an Intel Gaudi chip on a circuit board with visible microcomponents and metallic connectors. The chip is centered, displaying its branding prominently, surrounded by detailed electronic pathways.

Intel introduced the Intel Gaudi 3 AI accelerator on April 9, 2024, at the Intel Vision event in Phoenix, Arizona. The accelerator meets the unserved demands for choice while offering versatility through open and community-based software and open industry-standard Ethernet, helping businesses flexibly scale their systems. (Credit: Intel Corporation)

A computer chip labeled Intel is positioned on a reflective surface with colorful lighting, showcasing intricate circuits and connectors.

Intel introduced the Intel Gaudi 3 AI accelerator on April 9, 2024, at the Intel Vision event in Phoenix, Arizona. It is designed to bring global enterprises choice for generative AI, building on the performance and scalability of its Gaudi 2 predecessor. (Credit: Intel Corporation)

Close-up of a computer chip labeled Intel Gaudi on a circuit board, illuminated with a blue light.

Intel tackles the generative AI gap by introducing the Intel Gaudi 3 AI accelerator at the Intel Vision event on April 9, 2024, in Phoenix, Arizona. Gaudi 3 gives customers choice with open community-based software and industry-standard Ethernet networking to scale their systems more flexibly. (Credit: Intel Corporation)

Infographic showcasing the Intel Gaudi 3 AI chip features: 17x faster inference, 7.4x faster training than Nvidias H100, 128GB HBM2e, 3.7TB/s memory bandwidth, 180TB FLOPS, 24x 600GB/s fabric, extensive AI software stack support.

Intel introduced the Intel Gaudi 3 AI accelerator on April 9, 2024, at the Intel Vision event in Phoenix, Arizona. The AI accelerator is designed to break down proprietary walls to bring choice to the enterprise generative AI market. (Credit: Intel Corporation)

An electronic circuit board featuring an Intel Gaudi processor at the center, surrounded by various components and connectors, with a blue and yellow color scheme.
Image of an Intel Gaudi chip on a circuit board with visible microcomponents and metallic connectors. The chip is centered, displaying its branding prominently, surrounded by detailed electronic pathways.
A computer chip labeled Intel is positioned on a reflective surface with colorful lighting, showcasing intricate circuits and connectors.
Close-up of a computer chip labeled Intel Gaudi on a circuit board, illuminated with a blue light.
Infographic showcasing the Intel Gaudi 3 AI chip features: 17x faster inference, 7.4x faster training than Nvidias H100, 128GB HBM2e, 3.7TB/s memory bandwidth, 180TB FLOPS, 24x 600GB/s fabric, extensive AI software stack support.

下載所有圖片 （ZIP，27 MB）

Intel Gaudi AI加速器提供高效能生成式AI，大幅降低總體擁有成本

現今，生成式AI的應用布署和運行不僅更快速，成本也更低。x86架構作為主流基礎架構，已大規模運行於各類資料中心，並在其中扮演整合AI功能的關鍵角色，可確保在合理成本下與其他系統和套件互通，並為使用者提供開放生態系的優勢。

Intel Xeon處理器是AI工作負載中CPU節點的絕佳選擇。Intel Xeon處理器與專為AI工作負載設計的Intel Gaudi AI加速器可協同運作，提供與現有IT基礎架構無縫整合的解決方案。

Intel Gaudi架構作為唯一通過MLPerf基準測試的Nvidia H100替代方案，可提供客戶生成式AI所需的卓越效能。相較於其他選擇，Intel Gaudi不僅性價比更具優勢，還能以更低的總體擁有成本（TCO），實現更快速布署。

在定價方面，英特爾提供極具競爭力的方案。一套包含八個Intel Gaudi 2加速器和一個通用基板（Universal Baseboard，UBB）的標準AI套件，售價為65,000美元，約為同級平台的三分之一。更高階的AI套件包含八個Intel Gaudi 3加速器和一個UBB的套件，售價為125,000美元，約同級平台的三分之二4。

Intel Gaudi 3加速器可大幅提升生成式AI模型的訓練和推理效能，協助企業充分發揮專屬資料的潛在價值。根據預估，在8,192組加速器的叢集環境中，Intel Gaudi 3可比同等規模的Nvidia H100 GPU叢集縮短40％模型訓練時間5；在64組加速器叢集中，Intel Gaudi 3在Llama2-70B模型上的訓練速度，預計可比Nvidia H100快15％6。此外，Intel Gaudi 3在運行Llama-70B和Mistral-7B等大型語言模型（LLM）時，推理性能平均較Nvidia H100提高2倍7。

Intel Gaudi 3已推出至市場，除了已與戴爾、慧與科技、聯想和美超微合作外，今日更宣布與華碩、鴻海、技嘉科技、英業達、廣達電腦和緯創展開合作，擴大產品線布局。

Close-up image of a computer chip with two prominent square components at the top, featuring alphanumeric codes. The chip has a blue and gold outline with other electronic markings on a plain black background.

On June 4, 2024, at Computex, Intel revealed new details regarding the next generation of the new mobile processor, code-named Lunar Lake. Lunar Lake brings a new ground-up redesign that delivers greater power and efficiency, and leading AI compute. It will power the next generation of AI PCs with unprecedented x86 power efficiency and no-compromise application compatibility. (Credit: Intel Corporation)

Illustration of a computer chip with a sleek, rectangular design. The chip features a prominent central section with intricate circuit patterns highlighted in blue, surrounded by a metallic frame.

Lunar Lake: 2024年8月の Hot Chips イベントで、インテルは、インテル® Gaudi 3、Lunar Lake プロセッサー搭載クライアント、インテル Xeon 6 システムオンチップ搭載のエッジを使用したクラウド全体の AI アーキテクチャの専門知識をカバーする 4 つのセッションを発表しました。(クレジット: Intel Corporation)

Collage of features for Intels Lunar Lake chip: integrated Wi-Fi, NPU 4.0, Enhanced Intel Thread Director, Xe2 GPU, up to 32GB memory, 120 TOPS AI, Ecores, P-cores, new packaging, and power delivery. Intel logos present.

An illustration shows the angled view of the Lunar Lake package. On June 4, 2024, at Computex, Intel revealed new details regarding the next generation of the new mobile processor, code-named Lunar Lake. Lunar Lake brings a new ground-up redesign that delivers greater power and efficiency, and leading AI compute. (Credit: Intel Corporation)

Close-up image of a computer chip with two prominent square components at the top, featuring alphanumeric codes. The chip has a blue and gold outline with other electronic markings on a plain black background.
Illustration of a computer chip with a sleek, rectangular design. The chip features a prominent central section with intricate circuit patterns highlighted in blue, surrounded by a metallic frame.
Collage of features for Intels Lunar Lake chip: integrated Wi-Fi, NPU 4.0, Enhanced Intel Thread Director, Xe2 GPU, up to 32GB memory, 120 TOPS AI, Ecores, P-cores, new packaging, and power delivery. Intel logos present.

下載所有圖片 （ZIP，7 MB）

加速筆記型電腦的終端AI：新架構帶來3倍AI運算效能提升，實現前所未有的能源效率

除了維持資料中心領域領先地位，英特爾也積極擴大邊緣裝置和PC領域的AI影響力。英特爾的邊緣布署案例已超過9萬個，交付生態系夥伴的CPU高達2億個，此成果展現英特爾多年來持續提供企業客戶多元化選擇的承諾。

AI個人電腦正為全方位運算體驗帶來革命性轉變，位於AI PC創新時代前端的英特爾，其優勢不僅在於提供更快處理速度與更流暢設計的軟硬體解決方案，更著眼於打造可即時學習、進化的終端裝置。這些智慧型裝置可預測使用者需求，並根據其偏好不斷調適，開創一個兼具生產力、效率和創造力的嶄新時代。

根據預測，2028年搭載AI功能的PC將佔所有PC數量約80%。英特爾深刻洞察此趨勢，已著手打造最佳化AI PC軟硬體平台。英特爾以Core Ultra平台為基礎，與100家以上獨立軟體供應商（ISV）緊密合作，共同打造300項AI加速功能和500種AI模型。

基於上述優勢，英特爾在2024年COMPUTEX大會中公布了Lunar Lake處理器架構細節。Lunar Lake作為新世代AI PC的旗艦處理器，不只在繪圖與AI處理方面都有顯著提升，在輕薄型產品的高效能運算方面也有卓越表現，此款處理器可降低系統單晶片功耗高達40%3，並提供3倍以上AI運算能力8。Lunar Lake處理器預計將於2024年第三季度開始出貨。

Lunar Lake全新架構將提供以下功能：

  • 新的P-core和E-core：可大幅提升效能和能源效率。
  • 第4代英特爾NPU：提供高達48 TOPS的AI運算效能，Lunar Lake NPU較上一代的運算能力提升4倍，可最佳化生成式AI使用體驗。
  • 全新GPU架構（代號Battlemage）：融合了Xe2 GPU核心和Xe Matrix Extension（XMX）陣列兩項創新技術。Xe2 GPU核心可提升遊戲和繪圖性能1.5倍，新的XMX陣列則作為第二個AI加速器，提供高達67 TOPS運算效能，以因應AI內容創作所需的龐大吞吐量（Throughput）。
  • 先進的低功耗島（Low Power Island）：專為處理背景作業和生產力任務設計的全新運算叢集，具備極高能源效率，可有效延長筆記型電腦的電池續航力。

 

在其他公司正準備進軍AI個人電腦市場之際，英特爾已經搶得先機大量出貨。英特爾2024年第一季度交付的AI個人電腦處理器數量，已超過所有競爭對手的總出貨量。即將推出的Lunar Lake處理器，將為來自20家OEM廠商、超過80款不同型號的AI個人電腦提供強大運算效能。英特爾也預計在今年出貨超過4,000萬個Core Ultra處理器，進一步鞏固其在AI PC領域的領先地位。

正如英特爾共同創辦人之一Gordon Moore（高登·摩爾）所言：「所有已完成的事都能夠被超越」。英特爾始終堅持不懈追求更卓越的境界。憑藉涵蓋客戶端、邊緣、資料中心和雲端的全球布署，以及建立在開放標準基礎上的強大生態系統、具經濟效益的解決方案，致力落實AI無所不在的願景、積極擘畫AI未來發展藍圖。今日宣布的重大進展，不僅代表英特爾的技術躍進，更是對所有客戶和合作夥伴的誠摯邀請，共同掌握這波前所未有的商機，攜手開創下一個創新時代。

Forward-Looking Statements

This release contains forward-looking statements that involve a number of risks and uncertainties, including with respect to Intel’s product roadmap and anticipated product sales and competitiveness and projected growth and trends in markets relevant to Intel’s business. Such statements involve many risks and uncertainties that could cause our actual results to differ materially from those expressed or implied, including those associated with:

  • the high level of competition and rapid technological change in our industry;
  • the significant long-term and inherently risky investments we are making in R&D and manufacturing facilities that may not realize a favorable return;
  • the complexities and uncertainties in developing and implementing new semiconductor products and manufacturing process technologies;
  • our ability to time and scale our capital investments appropriately and successfully secure favorable alternative financing arrangements and government grants;
  • implementing new business strategies and investing in new businesses and technologies;
  • changes in demand for our products;
  • macroeconomic conditions and geopolitical tensions and conflicts, including geopolitical and trade tensions between the US and China, the impacts of Russia's war on Ukraine, tensions and conflict affecting Israel and the Middle East, and rising tensions between mainland China and Taiwan;
  • the evolving market for products with AI capabilities;
  • our complex global supply chain, including from disruptions, delays, trade tensions and conflicts, or shortages;
  • product defects, errata and other product issues, particularly as we develop next generation products and implement next-generation manufacturing process technologies;
  • potential security vulnerabilities in our products;
  • increasing and evolving cybersecurity threats and privacy risks;
  • IP risks including related litigation and regulatory proceedings;
  • the need to attract, retain, and motivate key talent;
  • strategic transactions and investments;
  • sales-related risks, including customer concentration and the use of distributors and other third parties;
  • our significantly reduced return of capital in recent years;
  • our debt obligations and our ability to access sources of capital;
  • complex and evolving laws and regulations across many jurisdictions;
  • fluctuations in currency exchange rates;
  • changes in our effective tax rate;
  • catastrophic events;
  • environmental, health, safety, and product regulations;
  • our initiatives and new legal requirements with respect to corporate responsibility matters; and
  • other risks and uncertainties described in this release, our 2023 Form 10-K, and our other filings with the SEC.

Given these risks and uncertainties, readers are cautioned not to place undue reliance on  such forward-looking statements. Readers are urged to carefully review and consider the  various disclosures made in this release and in other documents we file from time to time  with the SEC that disclose risks and uncertainties that may affect our business.

Unless specifically indicated otherwise, the forward-looking statements in this release are  based on management's expectations as of the date of this release, unless an earlier date is  specified, including expectations based on third-party information and projections that  management believes to be reputable. We do not undertake, and expressly disclaim any  duty, to update such statements, whether as a result of new information, new  developments, or otherwise, except to the extent that disclosure may be required by law.

Disclaimers:

AI runs best on Intel across the compute continuum from the data center, cloud and network to the edge and PC as of May 2024, based on broad compatibility, extensive software options, unique architecture, and impressive performance of Intel offerings, which combine to deliver the best overall AI experience, including in comparison to competitive offerings.  See intel.com/performanceindex for details.  Results may vary.

1 See [7T1] at intel.com/processorclaims: Intel® Xeon® 6. Results may vary.

2 Pricing estimates based on publicly available information and Intel internal analysis.

3 Disclaimer for footnote: Power measurements are based on Lunar Lake reference platform using YouTube 4K 30fps AV1. See backup for details. Results may vary.

4 Pricing guidance for cards and systems is for modeling purposes only. Please consult your original equipment manufacturer (OEM) of choice for final pricing. Results may vary based upon volumes and lead times.

5 Source for Nvidia H100 GPT 3 performance: https://mlcommons.org/benchmarks/training/, v3.1, closed division round. Accessed April 30, 2024​

Intel Gaudi 3 measurements and projections by Habana Labs, April 2024; Results may vary​

Intel Gaudi 3 performance projections are not verified by MLCommons Association. The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See http://www.mlcommons.org/ for more information.​

6 Source for Nvidia H100 LLAMA2-70B performance https://developer.nvidia.com/deep-learning-performance-training-inference/training , April 29,2024, “Large Language Model” tab.​

Intel Gaudi 3 measurements and projections by Habana Labs, Apr 2024; Results may vary​

7 Source for Nvidia performance: Overview — tensorrt_llm documentation (nvidia.github.io), May 2024. Reported numbers are per GPU.​

Intel Gaudi 3 projections by Habana Labs, April 2024; Results may vary​

8 Based on total number of platform Tops on Lunar Lake vs. prior generation.

Related Posts

News 人工智慧

Intel 在 CES 2025 上擴展了 AI 計算機和邊緣運算的領導地位

Intel 為企業與消費者突破 AI 效能與電源效率的界限，開創 AI 運算的新時代。

January 6, 2025
News 人工智慧

英特爾展示採用Xeon 6處理器與Gaudi 3加速器的最新系統與解決方案

December 12, 2024
News 人工智慧Close-up of an Intel Gaudi processor on a circuit board. The chip is centrally positioned, surrounded by intricate circuitry and small components in blue, yellow, and metallic colors.

英特爾發表新一代AI解決方案Xeon 6和Gaudi 3

開啟高效能企業AI系統與解決方案新紀元

September 24, 2024
News 人工智慧

英特爾攜手生態系合作夥伴推出最新商用AI PC裝置與應用

新一代Intel vPro平台與新版OpenVINO工具套件協助企業輕鬆布署終端AI

August 27, 2024