loading
loading
ASHRAE TC 9.9, the technical committee that owns most of the data center industry’s thermal guidance, has posted a freely downloadable Technical Bulletin titled TCS Coolant Integrity and System Readiness Best Practices, and it opens more with a warning than standards-body documents usually do: “It is too easy to violate direct-to-chip cooling system operational and warranty requirements.” Issued in May and now up on the committee’s public documents page, the bulletin catalogs four areas where a technology cooling system (TCS) can drift from its design basis in ways that stay invisible at commissioning and surface as damage months later. The timing tracks the wave of direct-to-chip liquid-cooling deployments that AI-class processor densities are driving into mainstream data centers. If the material sounds familiar, it should. Dell’s Tim Shedd walked through exactly this territory last week on podcast #152, from coolant chemistry and filtration to why commissioning shortcuts show up as failures long after the install crew has left, and mentioned that this bulletin was on the way. He is credited among its contributors, alongside Dell’s Meraj Mohebi, IBM’s Dustin Demetriou, AMD’s Ashwin Siddarth, Vertiv’s Chris Campbell, Oracle’s Vali Sorell, and engineers from Fluid2Chip and DLB Associates, a cross-vendor roster that reflects how shared this problem set has become. The Four Failure Paths Coolant composition comes first because everything downstream is sized against it. Most single-phase deployments are designed around a nominal 25% propylene glycol mix (PG25), and the bulletin details how drifting from the design-basis formulation changes viscosity, density, specific heat, and thermal conductivity, which cascades into altered pressure drop, a shifted pump operating point, degraded heat exchanger effectiveness, and even CDU control stability. Glycol should meet TC 9.9’s minimum quality guidance or the IT equipment manufacturer’s spec, and mixing coolants without confirming the impacts is one of the explicit triggers for what the bulletin calls a structured technical review. Air management is the sleeper. Entrained air cuts heat-transfer effectiveness, cavitates pumps, accelerates corrosion, and stretches commissioning schedules, and the bulletin is specific about the mechanism that catches teams off guard: degassing is most effective at elevated temperatures that a loop may only reach under heavy load, so air problems can appear weeks after startup once the system settles at steady-state temperatures. The guidance covers separator placement on the warm, low-pressure return near the CDU inlet, venting at every high point, and avoiding trap geometries, vertical dead legs, inverted U-bends, and unvented flexible hose runs that can hold air indefinitely. Stainless-steel passivation gets the bulletin’s firmest language. Welding, grinding, and field fabrication damage the chromium-oxide film that makes stainless steel corrosion-resistant, and without proper cleaning and re-passivation per ASTM A380/A967 practice, free iron ends up in the coolant and fouls the microchannel surfaces that cold plates depend on, the same turbulator structures we examined up close in our Chilldyne cold plate deep dive, where fluid quality decides whether those fine features keep moving heat or plug up. The rule the bulletin lands on is absolute: “An individual installed component that is NOT passivated can jeopardize the operation of the system.” System pressure ratings round out the four. The committee recommends a full hydraulic model of the pressure cascade, covering not only steady-state operation but also flush and fill pressures, static height, relief valve settings, expansion tank pre-charge, and the transients associated with connecting and disconnecting equipment. It also points to IEC 62368-1’s requirement for a hydrostatic safety test at 1.5x the rated maximum working pressure, with relief valves set below the rated maximum that the IT equipment manufacturer publishes in its thermal template. The failure mode being designed against is the expensive one: overpressuring the IT equipment itself. From Guidance to Field Checklist The back half of the bulletin becomes operational, and it clearly has been written by people who have been burned. Coolant concentration and quality should be lab-verified at initial fill, after any significant makeup fluid addition, and on a periodic schedule, with the design-basis fluid spec documented in turnover deliverables. Vacuum filling, an increasingly popular technique for removing trapped air, has a caveat we have not seen elsewhere: glycol mixtures can partially separate under vacuum, so the committee recommends sampling and lab testing after a vacuum fill to confirm the blend remains uniform. There is practical adjudication of the automatic-versus-manual air vent debate (cap the automatic vents or isolate them with a valve once the loop is confirmed air-free, rather than swapping hardware), and the whole document closes with a field verification checklist that maps each risk area to what to verify, the acceptable evidence, and why it matters, a page that commissioning agents can lift directly into their QA plans. The bulletin is a companion to TC 9.9’s 2024 Liquid Cooling: Resiliency Guidance for Cold Plate Deployments and carries the standard caveat: it is informational guidance and does not supersede IT equipment manufacturer requirements or local code. For a conversational take on why this plumbing discipline matters, Shedd’s hour-long podcast episode is a good pairing.
document fetched 2026-08-31 02:47 · canonical https://storagereview.com/news/ashrae-tc-9-9-flags-four-ways-direct-to-chip-cooling-loops-go-wrong