HPE Cray XD675: технические характеристики и документация

HPE Cray XD675

HPE is bringing the power of supercomputing to datacenters of any size with the HPE Cray XD675.  

HPE Cray XD675 is a top-performing GPU-accelerated server, delivering AI and HPC workload solutions to rack-scale, in a rack and roll fashion. 

HPE Cray XD675 is an 8U chassis system that contains a single 2x CPU node with 8x AMD MI300X OAM GPUs. It offers a complete, scalable solution for AI & HPC customers everywhere, with flexibility of fabric, memory, storage and operating system. HPE Cray XD675 provides maximum performance for advanced HPC Simulations, AI Training and Deep Learning. 

Built with Exascale-ready networking technologies, integrated storage, extensive software portfolio and management tools, HPE Cray XD675s can enable customers to innovate and prepare for tomorrow's challenges.

Cray XD675 Server System Key Features

• 8U Single-Node Chassis (Air-Cooled)
• GPUs: 8x AMD MI300X OAM GPUs providing leadership performance for AI Training, Deep Learning and advanced HPC simulations.
• CPUs: Support for 4^th Generation AMD® EPYC® Scalable Processors: "Genoa" up to 400W TDP
• DRAM: Support for up to 24x DDR5 5600MT/s DIMMs (supported at 4800 MT/s with 4^th Generation AMD® EPYC® Scalable Processors)
• High-Speed Fabric: 8x PCIe Gen 5.0 Half-Height, Half-Length slots supporting InfiniBand NDR and Ethernet, providing direct switchable connections between High-Speed Fabric, GPUs, NVMe drives and CPUs.
• Storage: Up to 16 SFF NVME SSDs, 8 SAS/SATA SSDs, and 2 M.2 2280 Kits.
• Power Supplies: 3000W Platinum Kit (6 x 3000W + 2 x 2400W) and 3000W Titanium Kit (6 x 3000W + 2 x 2700W)
• PCIe Expansion: 5x Full-Height, Half-Length PCIe Gen 5.0  embedded 2-port 10G Base-T (RJ45), 1 1GbE NIC, 1x BMC Port, 1x USB3.0 type C, 1x debug port (type C) PWR Button/Health/Power/UID LEDs 

NVIDIA and NVLink are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries. All third-party marks are the property of their respective owners. 

AMD and EPYC are trademarks and/or registered trademarks of Advanced Micro Devices, Inc. in the U.S. and other countries. All third-party marks are property of their respective owners

HPE Cray XD675 Server Top View

Item

Description

1.

8x AMD MI300X OAM Accelerator

HPE Cray XD675 Server Rear View

Item

Description

Item

Description

1.

LP PCIe x8

6.

OCP 3.0 Slot

2.

FHHL PCIe x3

7.

DC-SCM

3.

2400W PSU

8.

2400W PSU

4.

3000W * 6 PSU

9.

FHHL PCIe x2

5.

OCP 3.0 Slot

HPE Cray XD675 Server Front View

Item

Description

3.

SAS / SATA SSD x8

1.

80105 System Fan (x12)

4.

NVMe SSD x8

2.

NVMe SSD x8

Rack Airflow Requirements

HPE Cray XD675

The increasing power of new high-performance processor technology requires increased cooling efficiency for rack-mounted servers. For maximum cooling, HPE racks are recommended to allow these racks to be fully loaded with servers using the latest processors. For detail information please see Cray XD675 System User Guide:

https://www.hpe.com/info/CrayXD675-docs

Thermal Boundaries

The ability to operate up to 35C will come with limitations on choice of CPU and Memory SKUs. The tables below outline the thermal operating ranges for the CPU and Memory SKUs

Notes:

- If a third-party rack is used, observe the following additional requirements to ensure adequate airflow and to prevent damage to the equipment.

- Always use blanking panels to fill all remaining empty front panel U-spaces in the rack. This arrangement ensures proper airflow. Using a rack without blanking panels will result in improper cooling that can lead to thermal damage.

Software Development Tools (Programming languages, debuggers, libraries)

Additional 3^rd party software developmental tools:

• AMD ROCm 6.2+

Security

• Secure Start - only boot Signed FW
• UEFI Secure Boot
• NIST SP800-193 Protect, Detect, Recover for BIOS and BMC Tamper-free updates - components digitally signed and verified using CNSA strength RSA4096 signing key.
• Secure Recovery - recover critical firmware to known good state on detection of compromised firmware. Event recorded in the event log.
• UEFI TPM 2.0 (Trusted Platform Module 2.0) HPE Cray XD675 supports OpenSSL 3.0 provide for secure remote connectivity session to system
• SPI BUS monitoring to ensure BIOS and BMC FW are always signed and secure.
• Secure out-of-the-box high-entropy RMC administrator password is unique for each system.
• Support for multiple Redfish role-based access control (RBAC) accounts ensuring that passwords need not be shared and can provide separation of duties.

HPE Services

No matter where you are in your digital transformation journey, you can count on HPE Services to deliver the expertise you need when, where and how you need it. From planning to deployment, ongoing operations and beyond, our experts can help you realize your digital ambitions.

https://www.hpe.com/services

Consulting Services

No matter where you are in your journey to hybrid cloud, experts can help you map out your next steps.  From determining what workloads should live where, to handling governance and compliance, to managing costs, our experts can help you optimize your operations. 

https://www.hpe.com/services/consulting

HPE Managed Services

HPE runs your IT operations, providing services that monitor, operate, and optimize your infrastructure and applications, delivered consistently and globally to give you unified control and let you focus on innovation. 

HPE Managed Services | HPE

Operational services

Optimize your entire IT environment and drive innovation.  Manage day-to-day IT operational tasks while freeing up valuable time and resources. Meet service-level targets and business objectives with features designed to drive better business outcomes.

https://www.hpe.com/services/operational

HPE Complete Care Service

HPE Complete Care Service is a modular, edge-to-cloud IT environment service designed to help optimize your entire IT environment and achieve agreed upon IT outcomes and business goals through a personalized experience. All delivered by an assigned team of HPE Services experts.  HPE Complete Care Service provides:

• A complete coverage approach -- edge to cloud
• An assigned HPE team
• Modular and fully personalized engagement
• Enhanced Incident Management experience with priority access
• Digitally enabled and AI driven customer experience

https://www.hpe.com/services/completecare 

HPE Tech Care Service

HPE Tech Care Service is the operational support service experience for HPE products. The service goes beyond traditional support by providing access to product specific experts, an AI driven digital experience, and general technical guidance to not only reduce risk but constantly search for ways to do things better. HPE Tech Care Service delivers a customer-centric, AI driven, and digitally enabled customer experience to move your business forward. HPE Tech Care Service is available in three response levels. Basic, which provides 9x5 business hour availability and a 2-hour response time. Essential which provides a 15-minute response time 24x7 for most enterprise level customers, and Critical which includes a 6-hour repair commitment where available and outage management response for severity 1 incidents.

https://www.hpe.com/services/techcare

HPE Lifecycle Services

HPE Lifecycle Services provide a variety of options to help maintain your HPE systems and solutions at all stages of the product lifecycle.  A few popular examples include:

• Lifecycle Install and Startup Services: Various levels for physical installation and power on, remote access setup, installation and startup, and enhanced installation services with the operating system.
• HPE Firmware Update Analysis Service: Recommendations for firmware revision levels for selected HPE products, taking into account the relevant revision dependencies within your IT environment.
• HPE Firmware Update Implementation Service: Implementation of firmware updates for selected HPE server, storage, and solution products, taking into account the relevant revision dependencies within your IT environment.
• Implementation assistance services: Highly trained technical service specialists to assist you with a variety of activities, ranging from design, implementation, and platform deployment to consolidation, migration, project management, and onsite technical forums.
• HPE Service Credits: Access to prepaid services for flexibility to choose from a variety of specialized service activities, including assessments, performance maintenance reviews, firmware management, professional services, and operational best practices.

Notes: To review the list of Lifecycle Services available for your product go to:
https://www.hpe.com/services/lifecycle 

For a list of the most frequently purchased services using service credits, see the HPE Service Credits Menu

Other Related Services from HPE Services:

HPE Education Services

Training and certification designed for IT and business professionals across all industries. Broad catalogue of course offerings to expand skills and proficiencies in topics ranging from cloud and cybersecurity to AI and DevOps. Create learning paths to expand proficiency in a specific subject.  Schedule training in a way that works best for your business with flexible continuous learning options.

https://www.hpe.com/services/training

Defective Media Retention

An option available with HPE Complete Care Service and HPE Tech Care Service and applies only to Disk or eligible SSD/Flash Drives replaced by HPE due to malfunction. 

Consult your HPE Sales Representative or Authorized Channel Partner of choice for any additional questions and services options.

Parts and Materials

HPE will provide HPE-supported replacement parts and materials necessary to maintain the covered hardware product in operating condition, including parts and materials for available and recommended engineering improvements. 

Parts and components that have reached their maximum supported lifetime and/or the maximum usage limitations as set forth in the manufacturer's operating manual, product quick-specs, or the technical product data sheet will not be provided, repaired, or replaced as part of these services.

How to Purchase Services

Services are sold by Hewlett Packard Enterprise and Hewlett Packard Enterprise Authorized Service Partners:

• Services for customers purchasing from HPE or an enterprise reseller are quoted using HPE order configuration tools.
• Customers purchasing from a commercial reseller can find services at https://ssc.hpe.com/portal/site/ssc/

AI Powered and Digitally Enabled Support Experience

Achieve faster time to resolution with access to product-specific resources and expertise through a digital and data driven customer experience. 

Sign into the HPE Support Center experience, featuring streamlined self-serve case creation and management capabilities with inline knowledge recommendations. You will also find personalized task alerts and powerful troubleshooting support through an intelligent virtual agent with seamless transition when needed to a live support agent. 

https://support.hpe.com/hpesc/public/home/signin

Consume IT On Your Terms

HPE GreenLake edge-to-cloud platform brings the cloud experience directly to your apps and data wherever they are-the edge, colocations, or your data center. It delivers cloud services for on-premises IT infrastructure specifically tailored to your most demanding workloads. With a pay-per-use, scalable, point-and-click self-service experience that is managed for you, HPE GreenLake edge-to-cloud platform accelerates digital transformation in a distributed, edge-to-cloud world.

• Get faster time to market
• Save on TCO, align costs to business
• Scale quickly, meet unpredictable demand
• Simplify IT operations across your data centers and clouds

To learn more about HPE Services, please contact your Hewlett Packard Enterprise sales representative or Hewlett Packard Enterprise Authorized Channel Partner.  Contact information for a representative in your area can be found at "Contact HPE" https://www.hpe.com/us/en/contact-hpe.html

For more information

http://www.hpe.com/services

This section lists some of the steps required to configure a Factory Integrated Model. To ensure only valid configurations are ordered, Hewlett Packard Enterprise recommends the use of a Hewlett Packard Enterprise approved configurator. Contact your local sales representative for information on configurable product offerings and requirements.

Notes: FIO indicates that this option is only available as a factory installable option.

Step 1: Choose a Chassis

HPE Cray Supercomputing XD675 Configure-to-order Server

P68867-B21

Step 2: Choose GPU Option

AMD Instinct MI300X 8-way Accelerator for HPE Cray Supercomputing XD675

P69044-B21

Step 3: Select M.2 Storage Options (max 1)

Micron 7450 960GB NVMe Gen4 Read Intensive M.2 2280 SSD for HPE Cray XD

P69525-H21

HPE Cray XD MR9660-16i 16-port 8GB Cache Tri-Mode 24G PCIe4 x8 Controller

S4J77A

Notes:

-  M.2 Kit(Min 0 // Max 2)

- Controller only required if required by SSD option

Step 4: Select a Processor (Dual Processor Configuration Only)

AMD EPYC 9654 2.15GHz 96-core 360W Processor Kit for HPE Cray Supercomputing

P56262-B21

AMD EPYC 9534 2.4GHz 64-core 280W Processor Kit for HPE Cray Supercomputing

P56264-B21

AMD EPYC 9634 2.00GHz 84-core 290W Processor Kit for HPE Cray Supercomputing

P56267-B21

AMD EPYC 9454 2.35GHz 48-core 290W Processor Kit for HPE Cray Supercomputing

P56270-B21

AMD EPYC 9334 2.5GHz 32-core 210W Processor Kit for HPE Cray Supercomputing

P56273-B21

Notes: Certain limitations may apply to select processors. Please refer to the thermal table above or contact your HPE sales representative

Step 5: Select from the following Memory options (minimum 2304 GB Recommended)

SK Hynix 64GB (1x64GB) Dual Rank x4 DDR5-5600 CAS-52-52-52 R Standard Memory Kit for HPE Cray XD

S4J61A

HPE Cray XD 96GB (1x96GB) Dual Rank x4 DDR5-5600 CAS-52-52-52 R Standard Memory Kit   

S4L14A

SK Hynix 128GB (1x128GB) Quad Rank x4 DDR5-5600 CAS-52-52-52 R 3DS Std Memory Kit for HPE Cray XD

S4J68A

Notes:

- Minimum System Memory of 2304GB Recommended

- Certain limitations may apply to select Memory options. Please refer to the thermal table above or contact your HPE sales representative

Step 6: Choose from the following SFF Storage and RAID options

Hot Plug SFF  

NVME SSDs (max 16)

HPE Cray XD675 1.6TB NVMe Mixed Use SFF U.3 SSD

S4K93A

HPE Cray XD675 3.2TB NVMe Mixed Use SFF U.3 SSD

S4K94A

HPE Cray XD675 1.92TB NVMe Read Intensive SFF Self-encrypting SSD

S4K95A

HPE Cray XD675 3.84TB NVMe Read Intensive SFF Self-encrypting SSD

S4K96A

HPE Cray XD675 7.68TB NVMe Read Intensive SFF Self-encrypting SSD

S4K97A

HPE Cray XD675 15.36TB NVMe Read Intensive SFF Self-encrypting SSD

S4K98A

Step 7: Choose Factory Configuration Setting

Cray Compute Node Identifier (no max)

HPE Cray Compute Node FIO Configuration

R9H92A

Step 8: Choose Networking Card Configuration

InfiniBand & Ethernet PCIe (Max 13)

Broadcom BCM57504 Ethernet 10/25Gb 4-port SFP28 Adapter for HPE

P26264-B21

Intel E810-CQDA2 Ethernet 100Gb 2-port QSFP28 Adapter for HPE

P21112-B21

HPE Data Processing Unit InfiniBand NDR/Ethernet 400Gb 1-port QSFP112 HHHL B3140H Adapter

P66387-H21

HPE Data Processing Unit InfiniBand NDR200/Ethernet 200Gb 2-port QSFP112 FHHL B3220 Adapter

P66386-H21

HPE InfiniBand NDR/Ethernet 400Gb 1-port OSFP PCIe5 x16 MCX75310AAS-NEAT Adapter

P45641-H23

Broadcom 57608 Ethernet 400Gb 1-port QSFP-DD Adapter for HPE

P74287-H21

Step 9: Select Power Cables (Max 8)

A power supply kit consisting of 8x Titanium or Platinum 3,000-Watt Supplies is already embedded within the chassis.

Select Power Cords (required - max 8)

HPE C19 - C20 WW 250V 16Amp 2.5m Jumper Cord

295633-B22

HPE C19 - C20 WW 250V 16Amp Flint Gray 2.0m Jumper Cord

AF574A

HPE C19 - C20 WW 250V 16Amp Flint Gray 1.20m Jumper Cord

AF575A

Step 10: Transceivers (Max 20)

HPE InfiniBand NDR/Ethernet 1x400Gb OSFP Multi-mode 50m HCA-side Transceiver

P49764-B21

HPE InfiniBand NDR/Ethernet 400G QSFP112 MPO12 850nm Multi-mode 50m APC Transceiver

P65334-B21

Step 11: Racks

HPE 48U 800mmx1200mm G2 Enterprise Shock Rack

P9K58A

HPE 42U 800mmx1200mm G2 Enterprise Shock Rack

P9K46A

HPE 42U 800mmx1075mm G2 Enterprise Shock Rack

P9K42A

HPE Power Distribution Units

Power Distribution Units (PDUs) are an integral piece to this data center solution and HPE offers several types. Basic PDUs provide reliable power with 0U or 1U installation options. Metered PDUs have added intelligence to precisely track power usage and switched PDUs provide both local and remote power management. There are additional metered PDUs that are recommended for this solution that are not part of the mainstream PDU product offering. They are as follows:

HPE Switched 3-phase 66.5kVA/60309 5-wire 100A/277V 21-breaker Vertical NA PDU

R8P19A

HPE Metered 3Ph 66.5kVA/60309 100A 5-wire 480/277V Outlets (21) SDG23/Vertical NA PDU

879034-B21

HPE Metered 3Ph 39.9kVA/60309 60A 5-wire 480/277V Outlets (21) SDG23/Vertical NA PDU 

880459-B21

HPE Metered 3Ph 57.6kVA/60309 100A 5-wire 80A/230V Outlets (3) C13 (18) C19/Vertical NA PDU

880460-B21

HPE Metered 3Ph 34.5kVA/60309 60A 5-wire 48A/230V Outlets (3) C13 (18) C19/Vertical NA FIO PDU

880461-B21

HPE Cray Supercomputer 60A 415V 3 Phase 24 CX PDU

R4N30A

HPE Mtrd 3P 69.1kVA 125A 96A230V FIO PDU

880462-B21

HPE Metered 3Ph 45.1kVA/60309 63A 5-wire 63A/230V Outlets (3) C13 (18) C19/Vertical INTL FIO PDU

880463-B21

HPE G2 Metered/Switched 3Ph 17.3kVA/60309 4-wire 48A/208V Out (12) C13 (12) C19/Vertical NA/JP PDU

P9S22A

HPE G2 Metered 3Ph 17.3kVA/60309 60A 4-wire 48A/208V Outlets (12) C13 (12) C19/Vertical NA/JP PDU

P9R86A

HPE G2 Metered Modular 3Ph 17.3kVA/60309 60A 4-wire 48A/208V Outlets (6) C19/1U Horizontal NA/JP PDU

P9R80A

HPE G2 Metered Modular 3Ph 22kVA/60309 5-wire 32A/230V Outlets (6) C19/1U Horizontal INTL PDU

P9R81A

Thermal limitations

For a full list of thermal limitations please see the HPE Cray XD675 thermal guidelines.

https://www.hpe.com/support/xd675-thermal

Environmental-friendly Products and Approach - End-of-life Management and Recycling

Hewlett Packard Enterprise offers end-of-life product return, trade-in, and recycling programs, in many geographic areas, for our products. Products returned to Hewlett Packard Enterprise will be recycled, recovered or disposed of in a responsible manner. 

The EU WEEE Directive (2012/19/EU) requires manufacturers to provide treatment information for each product type for use by treatment facilities. This information (product disassembly instructions) is posted on the Hewlett Packard Enterprise web site 

These instructions may be used by recyclers and other WEEE treatment facilities as well as Hewlett Packard Enterprise OEM customers who integrate and re-sell Hewlett Packard Enterprise equipment.