Executive Summary
The Open Compute Project (OCP) centers on a cross-industry initiative to help standardize the server hardware “blocks” used for workloads like artificial intelligence and edge deployments. Toward that goal, Intel and other major technology companies, including OEMs, cloud vendors, and hardware manufacturers, created new specifications and standards for the Data Center Modular Hardware System (DC-MHS) and the Data Center Secure Control Module (DC-SCM). DC-SCM is part of the DC-MHS modularity family, but operates out of the Open Compute Hardware Management Project. The modular computing approach can simplify systems management, improve data center energy efficiency, reduce hardware costs, and minimize waste.
Challenge
Demands on the data center have never been greater. Enterprises adopting more complex and resource-intensive workloads, like machine learning (ML) and edge computing, must maximize their system hardware and software for hyperscale efficiency. However, a one-size-fits-all server solution cannot address every customers’ requirements. Modern businesses need better consistency across disparate components, including unified management. Therefore, it’s vital that technology vendors align to standardize their products, simplify customer technology solutions, and get those systems to market as quickly as possible. To save costs, enterprises also need ways to increase the lifespan of their technology investments. The ability to upgrade components within existing server infrastructure can provide customers with improved compute density while embracing environmentally friendly business practices.
The Open Compute Project fosters an ecosystem where industry players collaborate in a safe framework, shaping a versatile and diverse supply chain.
Optimize your cloud infrastructure with the latest technologies, strategies, and tools from Intel.
Solution
Modular computing systems, including servers featuring Intel® Xeon® processors, offer the potential to address all these issues simultaneously. With support from several technology companies, including AMD, Dell, Google, HPE, Jabil, Meta, Microsoft, NVIDIA, and Ampere Technology, among others, Intel was the driving force for the OCP DC-MHS initiative to help drive scalability, sustainability, and security across the data center. These technology vendors can help customers future-proof their platform investments by committing to a joint specification that embraces a modular design philosophy. The approach also reduces development, integration, and validation costs by sharing that investment among multiple technology suppliers. DC-MHS efforts will enable greater interoperability among data center and edge-based deployments by providing a more consistent management interface and form factors among building blocks. Plus, standards-based solutions coming to market can help customers deploy greener and more energy-efficient infrastructure faster. Since server elements can be reused or recycled, the DC-MHS specification can help reduce waste and offer enterprises a path to increase their data centers’ scale and speed while reducing electricity consumption.
“By working together with major technology vendors through the Open Compute Project DC-MHS initiative, we have an incredible opportunity to help enterprises embrace AI’s transformative impact, improve data center efficiency, and enhance their sustainability efforts.” — Zane Ball, Corporate VP & GM Datacenter Engineering & Architecture at Intel
Results
From 2010 to 2018, Intel helped reduce the amount of energy required for data centers worldwide by 20 percent yearly while increasing compute instances by 550 percent and growing electricity consumption by only six percent.1 The cross-industry DC-MHS standard will extend that trend into the future through solutions that deliver greater compute density in an environmentally responsible way. By rightsizing each component and connecting them modularly, Intel reduced a reference system’s carbon footprint by 27 percent compared to a non-modular approach.2
| Platform/Processor | Eagle Stream | Eagle Stream | Birch Stream | |||
|---|---|---|---|---|---|---|
| Specification | 4th Gen Intel® Xeon® Scalable Processors | 5th Gen Intel® Xeon® Scalable Processors | Granite Rapids SP | |||
| PC | Fox Creek Pass | Fox Creek Pass Refresh | Deer Creek City (Pass) | |||
| System | 2U 1N Spread core general purpose server (incl. fan, sb, BB etc.) |
300 | 2U 1N Spread core general purpose server (incl. fan, sb, BB etc.) |
300 | M-FLW, 2U 1N Spread core general purpose server |
360 |
| Processor | 32C/2.1G/270W (6430) | 540 | 32C/2.1G/270W (6530) | 540 | 300W |
600 |
| Memory | 64 GB DDR5-4800 RDIMM (16 Gb, DRx4,14.4W) x16 |
230 30 |
64 GB DDR5-5600 RDIMM (16 Gb, DRx4,16W)x16 |
256 | 64 GB DDR5-6400 - 304 RDIMM (19W)x16 | |
| Storage boot (M.2) | PCIe Gen 4 x4, M.2 2280 480 GB (Micron 7400 PRO, 7.25W)x2 |
15 | PCIe Gen 4 x4 M.2 2280 480 GB (Micron 7400 PRO, 7.25W)x2 |
15 | PCIe Gen 4 x4 M.2 2280 480 GB (Micron 7400 PRO, 7.25W)x2 |
15 |
| PCIe AIC | PCIe gen5x16 AIC 25W +75W |
100 | PCIe gen5x16 AIC 25W +75W |
100 | PCIe gen5x16 AIC 75Wx2 |
150 |
| OCP | OCP NIC 3.0 (25W)x1 PCIE gen5x16, 200G/400G |
25 | OCP NIC 3.0 (25W)x1 PCIE gen5x16, 200G/400G |
25 | OCP NIC 3.0 (35W)x1 PCIE gen5x16, 400G/800G |
70 |
| Baseboard overall | count in System | 0 | count in System | 0 | count in System | 0 |
| Total system power cal DC | 1330 | 1356 | 1659 | |||
| PSU efficiency | 80 Plus Titanium 0.91 | 80 Plus Titanium 0.91 | 80 Plus Titanium 1410 | |||
| AC inlet power (W) | 1461 | 1490 | .90 | |||
| Tool input (W) | 1400 | 1450 | 1566 |
Key Takeaways
- The joint DC-MHS initiative paves the way for new, more turnkey data center solutions with unified systems management.
- Validated, modular data center solutions can help enterprises enhance compute power and scale while lowering their infrastructure spend.
- Because modular components can be reused or recycled, enterprises can lower their carbon footprint and reduce waste.
