New MES Guidance Highlights Cost Implications of Cold Plate Flatness in Data Centers

Introduction


As data centers advance in technology, particularly with the integration of artificial intelligence (AI) processors, the construction and sourcing processes become increasingly complex. MES, Inc., a leading global manufacturing sourcing and supply chain company, has recently published compelling guidance aimed at providing clarity on the specifications related to cold plate flatness. This guidance is crucial for engineering teams focused on optimizing cooling efficiency without unnecessarily inflating manufacturing costs.

The Challenge of Cold Plate Flatness


In the quest for efficiency, data center teams often face the dilemma of tightening cold plate flatness tolerances. Tighter tolerances may appear beneficial at first; however, according to MES, they can lead to increased manufacturing expenses without producing a significant improvement in cooling effectiveness. Hiten Shah, the President and CEO of MES, emphasizes that beginning with a thermal model is key. If this model doesn’t demonstrate measurable benefits, imposing stricter tolerances can inadvertently inflate costs.

Tolerances become particularly challenging as AI processor packages increase in size. Consider, for instance, aiming for a 0.025 mm flatness tolerance across a 30 mm die, which translates to a precision rate of 1 in 1,200. As die size expands to 70 mm, the tolerance ratio becomes even more demanding—1 in 2,800. Additionally, various factors, such as residual stress, heat from machining, and distortion caused by fixture setups, compound the challenges of maintaining flatness.

Cost Implications


One key takeaway from the guidance is that cost rises incrementally. A notable shift occurs when a tolerance specification changes from 0.025 mm to 0.015 mm. At this stage, the transition from precision CNC machining to methods such as grinding or dedicated fly cutting necessitates additional resources for setup, fixturing, and inspection. In practical terms, MES suggests that a 0.025 mm tolerance is generally adequate for most direct-to-chip applications, suggesting that organizations should tailor their final specifications based on comprehensive thermal modeling analyses.

Moreover, the interface between the cold plate and the thermal interface material (TIM) significantly influences cooling efficiency. Thermal conductivity varies substantially between TIM and metals like copper. The TIM layer usually spans 25 to 75 microns in thickness, a stark contrast to copper's 388 W/m•K conductivity. Even a mild dishing of 0.040 mm can inadvertently add up to 40 microns of TIM locally, or rough surfaces might trap microscopic voids that hinder optimal thermal transfer.

Recommendations for Data Center Teams


The MES guidance offers specific recommendations aimed at optimizing both flatness and surface finish specifications. It advises engineering teams to specify both the flatness and finish concurrently, emphasizing the necessity to measure flatness across the die footprint due to its direct impact on heat dissipation.

Additionally, Brad Layne, MES’s Engineering Manager, underscores that a cold plate can meet flatness tolerances yet still present thermal resistance if the surface finish hasn’t been explicitly defined. This underlines the importance of clarity in specifications, particularly regarding critical die interfaces, which should include parameters such as Ra and Rz values.

Accounting for how machining sequences affect results is also vital; for instance, as-brazed flatness can range from 0.100 to 0.250 mm, as the vacuum braze cycle can relieve residual stresses and alter the base shape. The guidance strongly recommends performing machining on the mounting face after brazing and documenting this procedure clearly on the production drawings.

Conclusion


As MES continues to serve diverse industries across multiple regions, including the U.S., Mexico, Poland, and India, this guidance reflects their commitment to enhancing manufacturing practices. For those engaged in the high-stakes world of AI data centers, understanding the relationship between cold plate specifications and the associated costs is essential. The complete guidance, including illustrations and checklists for specifications, is publicly accessible through MES’s website. By aligning precision with cost-efficiency, data center teams can ensure optimal performance while managing their manufacturing budgets effectively.

Topics Business Technology)

【About Using Articles】

You can freely use the title and article content by linking to the page where the article is posted.
※ Images cannot be used.

【About Links】

Links are free to use.