Raised floor load rating and load test field guide

Access-floor load ratings, and why one number is never enough

An access-floor load rating is the load a panel and its understructure carry within a specific, defined test, and the trap is that there is no single rating. The same floor is rated several ways: concentrated load through a small indentor, uniform load spread over the panel, rolling load from a wheel, ultimate load as an overload factor, pedestal axial load straight down the column, and drop impact load. A panel can pass one and fail another. The spec calls out each by name because each protects against a different event.

The Ceilings and Interior Systems Construction Association, CISCA, publishes Recommended Test Procedures for Access Floors that define how each load is measured. Read that carefully: the procedures define the method, not a pass or fail. They tell a lab how to apply the load and what to measure. The number a panel has to hit comes from the project specification and the manufacturer's published rating tested to that method.

The wrong rating fails the floor in a quiet way. Someone quotes the concentrated number, the biggest one on the data sheet, and sizes the room around it. Then a loaded rack rolls in on casters during fit-out and the governing case was never the static point load, it was the rolling load over many passes, and that number was lower and nobody checked it. The tile takes a set, the panel edge crushes, and the failure traces back to a rating that answered the wrong question.

This guide goes deep on the loads. The full turnover record, level, grounding, air-seal, ESD, and seismic, lives in the raised-floor acceptance packet, and the loads are one chapter of it. Here that chapter gets its own walk-through, because matching the floor to the equipment is where the expensive mistakes hide.

What is a concentrated (point) load rating?

A concentrated load rating, also called the point load, is the load a single panel carries through a small indentor without deflecting or taking a permanent set past the allowed limits. It is the number people quote because it is usually the largest, and it stands in for a rack foot, a leveling stud, or a caster sitting at rest on the panel.

The CISCA concentrated load method applies the load through a steel indentor representing roughly one square inch of contact, commonly a 1 in square pad or a 1.128 in diameter round pad, which works out to about the same one-square-inch footprint. The load is placed at the panel's worst points, the center, the midpoints of the edges, and the corners over the pedestal, because where you push changes how the panel responds. The center is usually the soft spot for deflection; the corner over a pedestal head is stiff but concentrates load into the understructure.

Two limits decide the rating, and they are not the same thing. Deflection is how far the panel moves while the load sits on it. Permanent set is how much deformation stays after the load comes off. A common published panel might carry a 1,250 lbf concentrated design load with deflection under load held to around 0.080 in and average permanent set held to around 0.010 in, but those figures are class and manufacturer specific. Read them off the panel's published table and the project spec, not off habit. Field ratings commonly run from about 1,000 lbf up past 2,000 lbf depending on the panel.

Concentrated (point) load

Load applied through a roughly 1 sq in indentor (1 in square or 1.128 in diameter), representing a rack foot or caster at rest

Deflection

How far the panel surface moves down while the load is applied, measured under load

Permanent set

The deformation that remains in the panel after the load is removed; the floor never recovers it

How is a uniform (distributed) load rated?

Uniform load is weight spread evenly across the panel, rated in pounds per square foot, lb/sq ft, and it covers the static case where the load is not a point but a field. Think rows of stationary equipment without feet, batteries on a tray, boxes, or a slab of UPS gear sitting flat. The load is shared across the whole panel area instead of driving into one spot, so the panel holds far more of it than it holds as a concentrated point.

The arithmetic is simple and worth doing on the floor. Uniform load is total weight divided by the area it sits on. A 2,000 lb cabinet on a footprint of 2 ft by 4 ft is 8 sq ft, so the average uniform pressure is 250 lb/sq ft. That number looks comfortable against a typical uniform rating, and that is exactly the trap: the cabinet does not sit on 8 sq ft of floor. It sits on four feet, and each foot is a concentrated load.

So uniform load is the rating that almost never governs in a data hall and almost always gets quoted because it sounds reassuring. Use it for genuinely distributed loads. For anything on feet, casters, or a narrow base, the concentrated and rolling cases are the ones that decide whether the floor holds.

Uniform (distributed) load

Weight spread evenly over the panel, rated in lb/sq ft (psf) or kPa

w

Uniform pressure: total weight W divided by the footprint area A it bears on

Rolling load and the 10-pass and 10,000-pass tests

Rolling load is dynamic, the load a wheel applies as it rolls across the panel, and it is reported as two numbers because a panel takes one load for a few passes and a lower load for many passes. Repeated wheeling accumulates deformation the way a single pass never does. A panel that shrugs off one trip of a cart can fail under a thousand trips on the same line, so a single rolling number tells you almost nothing.

The CISCA rolling load method runs two separate wheel tests on separate panels. One uses a smaller, harder wheel, on the order of 3 in diameter by about 1-13/16 in wide, run for 10 passes. The other uses a larger wheel, on the order of 6 in diameter by about 2 in wide, run for 10,000 passes. Acceptance is tied to a deformation limit, commonly a combination of local and overall set on the order of 0.040 in after the test. The applied load is set by the panel class; the wheel sizes, pass counts, and the deformation limit are the parts that stay consistent across the method. Confirm the exact wheel, load, and pass scheme against the edition the spec cites.

Manufacturer data sheets publish both numbers. A given panel may list a 10-pass rolling load well above its concentrated rating and a 10,000-pass rolling load lower than the 10-pass figure, because endurance is the harder test. When you accept a floor, read both. The 10-pass number covers the occasional move; the 10,000-pass number covers a route that sees traffic day after day.

Why is rolling load usually the worst case?

On a data hall floor the worst event is almost never the rack sitting in place. It is the rack moving to that place. A loaded cabinet rolled across the floor on casters during fit-out concentrates its whole weight on a few small contact patches that are also moving, and moving load finds every soft panel, every unsupported cut edge, and every pedestal that is a hair low. The static in-service load is a known quantity sitting still. The install rolling load is the wild card.

It gets worse because the install case stacks the odds against you. The floor is new, the panels may not be fully locked, the understructure adhesive may not be at full cure, and the heaviest single object the floor ever sees is often the fully built rack being walked down the aisle on a pallet jack or its own casters. A rack that weighs 2,500 lb on four casters is putting north of 600 lb on each small wheel while it rolls, and that is a rolling load, not the gentle uniform number from the data sheet.

Plan the move path like it is a load case, because it is. Identify the route a heavy rack takes from the dock to its position, confirm the 10,000-pass rolling rating along that path if it sees repeat traffic, and where a route crosses cut panels or perimeter tiles, treat those as the weak links. On big fit-outs crews lay temporary load-spreading plates or plywood runners over the travel path so the wheel load spreads across more than one panel. That is cheap. A crushed panel under a half-million-dollar rack is not.

Ultimate load and the safety factor

Ultimate load is the overload the system survives without collapsing. Failure here means the system will no longer accept the load: the panel folds, the understructure buckles, the pedestal punches. It is not a working number you ever load the floor to. It is the margin between the rating you use and the point where the floor lets go. On manufacturer data sheets the published ultimate load is commonly about twice the rated concentrated load, for example a 1,250 lbf rated panel listed around 2,500 lbf ultimate, and that rated value is itself based on a deflection limit rather than on collapse.

Building-code evaluation handles the margin a different way, as a safety factor on strength. The ICC-ES acceptance criteria for access floors, AC48, derive the allowable concentrated load by dividing the tested ultimate strength by a safety factor of 3, and the evaluation reports also check that a panel survives a point load of at least twice its design rating. So two distinct numbers are in play: the data sheet ultimate at roughly twice the rated load, and the code-evaluation allowable at the tested ultimate strength divided by three. Do not assume the data sheet ultimate and the AC48 allowable derivation are the same figure.

The mistake is reading the ultimate number as usable capacity. It is the opposite. It is the number that says how much abuse the floor tolerates before catastrophe, and it exists so the everyday rating has real headroom. When a spec quotes an ultimate value, confirm whether it is the data sheet ultimate, roughly twice the rated load, or a code-evaluation figure, and confirm the safety factor the project's evaluation report actually used. CISCA itself defines only the test methods; it assigns no multiplier or safety factor of its own.

Pedestal axial load and overturning

The pedestal is the column, and its axial load rating is the straight-down compression the pedestal assembly carries to the slab without deforming. The panels collect the load and hand it to the pedestal heads, so the pedestal rating has to keep pace with whatever the panels carry, especially under a rack that lands four feet right over four pedestal heads. Common published pedestal axial ratings run on the order of 5,000 lb to 7,500 lb per pedestal, but that is spec and manufacturer specific, so read the rated value, not a remembered one.

Axial load is only half of what a pedestal has to resist. The CISCA procedures also cover a pedestal overturning moment test, because a tall pedestal under lateral load wants to tip, and the base-to-slab connection is what stops it. A pedestal that is plumb and bonded resists both the push down and the push sideways. One set in adhesive that never properly cured to a clean slab can carry the axial load on a calm day and still let go under the lateral nudge of a rolling rack.

Check the pedestal as part of the load path, not as a separate item. Push-test the heads for rocking, confirm the base is bonded to a properly prepped slab per the manufacturer's installation instructions, and confirm the pedestal height and type match what the panel rating assumed. A panel rated for 1,250 lbf on a pedestal that is loose or undersized is not a 1,250 lbf floor.

Drop impact load

Drop impact load is the panel's resistance to a weight dropped onto it, the dropped tool, the dropped equipment foot, the corner of a crate set down hard. The CISCA drop impact method drops a defined weight from a defined height onto a small contact area on the panel and checks that the system survives without failure. A commonly cited version drops about 150 lb from 36 in onto a one-square-inch area, with some panel specs calling for 175 lb or 200 lb from the same height. Confirm the value against the specified class.

Impact rarely governs the floor selection, but it is the test that catches a brittle panel. A panel can carry a high static concentrated load and still crack under a sharp impact if the core is brittle, which is why heavier cementitious-filled steel panels generally take impact better than lighter cores. If the room will see heavy equipment handled by hand, the impact rating is worth a look, not just the headline concentrated number.

What CISCA performance class should the spec call out?

There is a point of confusion worth clearing up: CISCA publishes the test methods, not a set of performance classes with assigned pass/fail values. The procedures define how to run the concentrated, uniform, rolling, ultimate, pedestal axial, overturning, and impact tests. What a panel has to achieve is set by the project specification and the manufacturer's rated values tested to those methods.

In practice a spec calls out the floor by its rated concentrated load and the matching rolling load, and manufacturers sell panels in tiers, often labeled by the concentrated rating in pounds. You will see panels described around 1,000, 1,250, 1,500, 2,000 lbf and higher, each with its own published rolling, ultimate, and impact numbers. So when a spec says a class, what it usually means is a design concentrated load plus the companion rolling-load requirement, the deflection and permanent-set limits, and the ultimate safety factor.

The relationship to watch is that a higher concentrated class does not guarantee a proportionally higher rolling number. Two panels with the same concentrated rating can have different 10,000-pass rolling ratings depending on the core and the understructure. So do not let the spec call out only the concentrated class and assume the rolling case is covered. Specify the rolling load the room needs by name, and confirm the chosen panel's published rolling table meets it. The exact class names and numbers are manufacturer and spec specific; verify them against the cut sheet and the contract documents.

How much weight can a raised floor hold, and matching it to the rack

How much a raised floor holds is not a single answer, it is the answer to which load case, and the work is translating the rack into the loads the floor actually sees. Start with the rack's wet weight, fully populated, with PDUs, cabling, and any in-rack cooling, because the empty cabinet weight on the brochure is not what lands on the floor. Then find the footprint and how the weight reaches the floor: feet, leveling studs, casters, or a base frame.

Now split the weight into the cases that matter. The static in-service case is the wet weight divided across the leveling feet, and each foot is a concentrated load the panel sees at rest. The install case is the wet weight divided across the casters while the rack rolls, and that is a rolling load over whatever path it travels. A 2,500 lb rack on four feet is roughly 625 lb per foot static, comfortably inside a 1,250 lbf concentrated panel. The same rack on four casters being wheeled in is roughly 625 lb per wheel rolling, and now the question is the panel's rolling rating, not its concentrated one.

Caster geometry changes the answer. Small, hard casters concentrate the load into a tiny patch and read as a harsher rolling load than large, soft wheels that spread it. The footprint matters too: a foot or caster landing in the center of a panel is the panel's softest point, while one landing over a pedestal head is supported. Where you cannot control where the feet land, size to the worst case, a foot at panel center. Lay the rack footprint over the floor grid and you can see which feet sit center-span and which sit over pedestals, and that map is what tells you whether the rated floor really carries this rack.

Deflection and permanent set: what passes and what a failed panel looks like

Deflection and permanent set are the two acceptance numbers behind every concentrated and rolling rating, and they answer different questions. Deflection is movement while the load is on, and a panel is allowed some. A stiff panel might move only a few hundredths of an inch under its rated load and spring fully back. That is not a failure. That is the panel doing its job within its elastic range.

Permanent set is the deformation that stays after the load lifts, and that is the one that ends careers. The panel never recovers it. A common published limit holds average permanent set on the order of 0.010 in at the rated concentrated load, but the controlling number is in the spec and the panel table. Set accumulates: each overload event leaves a little more, which is exactly why the rolling test runs to 10,000 passes and limits the combined set to roughly 0.040 in. The endurance number is a set budget, not a single-event number.

A failed panel shows itself before a meter touches it. The surface dishes in the center, the laminate or covering wrinkles or lifts at a corner, the panel rocks because its edges no longer sit flat on the stringers, and a straightedge across the joints catches a lip that was not there at install. Run a straightedge across the traffic paths and feel for the dish. Where a panel reads soft underfoot or a cart catches a step, pull it, check the underside for cracking and check the pedestal heads under it, and log it by coordinate. A panel with visible permanent set does not get accepted because it tested fine once. It is already past its budget.

Stringers, stringerless, and how the whole system carries the load

The rating on the panel is a system rating, not a panel-alone rating, because the panel hands its load to the understructure and the understructure hands it to the slab. Change the understructure and you change what the floor carries. This is why a panel's published numbers are tied to a specific pedestal and stringer configuration, and swapping in a different understructure can void the rating the spec relied on.

Bolted-stringer systems lock the panels to a grid of stringers bolted to the pedestal heads. They carry the most lateral and rolling load, which is why deep data halls usually spec them, and the CISCA stringer load test reflects that the stringer is a load-carrying member, commonly proven to carry a concentrated load at center span, on the order of 550 lbf, within a permanent-set limit. A missing or loose stringer bolt drops the rolling and lateral rating right there, and the panels over it start to rock. Snap-on stringer systems clip rather than bolt, lay faster, and carry less lateral load, and the failure is a stringer that has popped its clip. Stringerless cornerlock systems carry no stringers at all and lean on the pedestal heads and panel corners, so they are the most exposed to a single bad pedestal and to rolling load that wants to walk a panel sideways.

The field lesson is that you cannot accept the load rating by looking only at the panel. Confirm the understructure matches the rated configuration, the stringer bolts are present and tight or the clips fully seated, the stringers are not sprung or bowed, and the pedestals are plumb and bonded. A 2,000 lbf panel on a stringerless system with one rocking pedestal is not carrying 2,000 lbf where that pedestal is.

When do you run a field proof load test?

Most of the time you accept the floor against its tested rating and its installation, not by loading it in the field, because the panel was proven to the CISCA methods in a lab and the rating carries a 3x ultimate margin. You verify the installation, the right panel, the right understructure, plumb bonded pedestals, tight stringers, and you trust the rating the system was built to. A field proof load test is the exception, run when the rating is in doubt, the installation is questioned, the panels are reused or of unknown origin, or the spec or the owner calls for an on-floor demonstration.

When you do run one, set it up to prove the case the room cares about, which is usually the concentrated point load at panel center, and sometimes the rolling case along a defined path. For a static proof, apply a known load through a defined contact pad at the panel's weak point, hold it for the specified dwell time, and measure deflection under load and permanent set after removal with a dial indicator or precise level reference against the limits in the spec. Use a calibrated load: weights, a load cell with a jack, or a measured ballast, not a guess. For a rolling proof, run a wheel of the specified size and load along the path for the specified passes and measure the set.

Acceptance is the same two numbers as the lab test: deflection within the allowed value under load, and permanent set within the allowed value after the load comes off, with no cracking, no rocking, and no damage to the understructure. Record the load applied, the contact pad, the location by coordinate, the dwell time, the readings, and who witnessed it. A proof test with no record is a story. A proof test with the coordinate, the load, and the readings is evidence the floor carries what the spec demanded.

Cutouts and how an opening near the load cuts capacity

A cut panel is a weakened panel, and the load rating on the data sheet assumes a full, uncut panel. Cut a hole in it for cables or a PDU and you remove material that was carrying load and helping the panel span between pedestals. The closer the cutout is to where the load lands and the larger it is relative to the panel, the more it drops the panel's real capacity. A big cutout near a panel edge, with a rack foot or a caster crossing right beside it, is a panel set up to fail.

How the cut was made and supported is the difference between a clean opening and a soft spot. Cut edges should be trimmed and, where the panel design requires it, supported with the manufacturer's edge trim or additional pedestals so the cut edge is not carrying load it was never meant to carry. A panel with a large unsupported cutout has lost a chunk of its rolling and concentrated rating exactly where you can no longer see the original number stamped on it.

So treat any heavy load near a cutout as its own check. Keep rack feet and caster paths away from large openings where you can, add pedestals or edge support around cutouts that sit in load paths, and never assume a cut panel still meets its tag. The rating belonged to the panel before the saw touched it.

High-density AI racks and when the slab replaces the access floor

Racks are getting heavier fast, and the floor loading problem is growing with them. A conventional rack tops out around 3,000 lb, but AI racks packed with GPUs, high-speed networking, and in-rack liquid cooling routinely run past 4,000 lb, and heavy-duty AI enclosures rated to around 5,000 lb of static load are now on the market. The power density tells the same story: a current high-density AI rack can pull well over 100 kW, and the cooling and copper that go with that weight ride on the same floor.

Liquid cooling adds its own dead weight. A coolant distribution unit flooded with fluid can weigh on the order of 3 tons, and the manifolds, hoses, and in-row gear add more, so the floor sees both heavier racks and heavier support equipment in the same hall. The static concentrated load per foot climbs, and the install rolling load climbs with it, because someone still has to move that 4,000 lb rack from the dock to its slot.

This is why fewer new AI halls are built on raised floor at all. Strengthening an access floor to carry these loads, heavier panels, more pedestals, tighter understructure, gets expensive fast, and the airflow reason for the underfloor plenum weakens once the racks move to liquid cooling and do not need a pressurized cold-air plenum. The trend is toward pouring the equipment on a slab, or slab-on-grade, with overhead power and cabling, and reserving raised floor for halls that still rely on underfloor air or that carry loads a properly rated floor handles. The call is a structural one: when the rack weight and the rolling case push past what a rated access floor carries economically, the slab wins. Confirm the structural slab itself carries the point loads, because moving to slab moves the load question to the building structure under the IBC and ASCE 7, it does not erase it.

Tying every load result to a grid coordinate

A load result without a location is a result nobody can defend, so every reading ties to a grid coordinate the same way the rest of the acceptance packet does. The floor is already a grid, so the room gets a coordinate system, commonly letters on one axis and numbers on the other, and any tile reads as something like AA-01. That coordinate links the photo of the panel, the load reading or proof-test result, the punch item, and the signoff back to one physical tile.

Key the load record to the same grid the rest of the packet uses, set the origin and axes once, and put the grid on the plan that opens the record. The payoff comes a year later when a rack is going in over row 7 and someone asks whether that floor was ever proven for the load. The coordinate walks them from the plan to the rating, to the proof test if there was one, to the punch, to the signature. No coordinate, no chain.

Field checklist

What to document

A load record that cannot answer a question a year out did not do its job. Capture enough that someone who was never on the floor can reconstruct what it was rated for, what was installed, and what it was proven to carry, all by coordinate against the spec it was held to.

Record the specified load class and every value in it, the installed panel and understructure that meet it, the rack weights and footprints the floor was matched to, the load-case breakdown, the move paths and their rolling rating, any field proof test with the load applied and the deflection and permanent set measured, and the punch list with responsible party and closeout. Where a panel was cut, record the support added. Where the owner accepted an open item, record the exception in writing, by coordinate, with the signature that accepted it.

Common mistakes

• Speccing or quoting the uniform load when the real case is a rack on feet or casters, which is concentrated and rolling, not distributed.
• Sizing the floor to the concentrated rating and never checking the rolling load, the case a loaded rack on casters actually applies.
• Ignoring the install rolling load entirely, when moving the rack across a new floor is usually the heaviest single event the floor sees.
• Reading the ultimate load as usable capacity instead of the 3x overload margin it actually is.
• Placing rack feet or caster paths next to large cutouts and trusting the uncut panel's rating to still apply.
• Trusting a panel tag without confirming the installed understructure matches the rated configuration the number assumed.
• Using the empty cabinet weight off the brochure instead of the wet, fully populated weight that lands on the floor.
• Assuming a higher concentrated class automatically means a higher rolling rating, when the two do not track together.
• Accepting a panel that tested fine once but shows visible permanent set, which means it is already past its set budget.
• Running a field proof test with an uncalibrated load and no coordinate, so the result cannot be defended later.

Standards and references

CISCA, the Ceilings and Interior Systems Construction Association, publishes the Recommended Test Procedures for Access Floors, which define the concentrated, uniform, rolling, ultimate, pedestal axial, pedestal overturning moment, stringer, and drop impact load test methods that a spec calls out by name. The procedures define the methods, not pass/fail values, so the rated numbers come from the manufacturer's data tested to those methods and from the project specification. Confirm the specified class and the panel's rated values rather than assuming any single number, and verify the test edition the spec references, since the section organization can differ between editions.

The verified anchors worth carrying: the concentrated test uses a roughly one-square-inch indentor, a 1 in square or 1.128 in diameter pad; the rolling test runs two wheels, on the order of a 3 in wheel for 10 passes and a 6 in wheel for 10,000 passes, against a combined deformation limit commonly near 0.040 in; the ultimate load is at least three times the rated concentrated load without collapse; and code-evaluation practice such as the ICC-ES acceptance criteria derives the allowable working load from the tested ultimate divided by a safety factor of 3. The deflection, permanent-set, rolling, pedestal axial, and impact values themselves are class and manufacturer specific.

The structure under and around the floor follows the building code. Seismic and gravity design for the access floor as a nonstructural component, and for a slab where the equipment moves off raised floor, follow the IBC and the referenced ASCE 7 loading. For the full acceptance picture, level, grounding, air-seal, ESD per ANSI/ESD S20.20, and seismic anchorage, see the raised-floor acceptance packet. Confirm the applicable standards, editions, and the adopted code with the AHJ, because the exact requirement is set by the jurisdiction and the project, not by the rule of thumb.

Units, terms, and conversions

Access-floor load work crosses imperial and metric and a few trade synonyms, so the same rating reads differently across a spec, a data sheet, and a drawing set.

Concentrated and rolling loads are given in pounds-force, lbf, in US data sheets and in newtons or kilonewtons in metric ones. Uniform load is pounds per square foot, lb/sq ft or psf, or kilopascals, kPa. Deflection and permanent set are in inches or millimeters, and they are not interchangeable: deflection is under load, permanent set is what stays after. Rated concentrated load and point load mean the same thing, raised floor and access floor mean the same thing, and the grid coordinate such as AA-01 is not a unit but it is the key the whole load record turns on, so define it once and use it everywhere.

lbf (pounds-force)

The unit for concentrated, rolling, ultimate, and pedestal loads in US data sheets; newtons in metric

lb/sq ft (psf)

Pounds per square foot, the unit for uniform distributed load; kilopascals in metric

Deflection

Movement of the panel surface while the load is applied; some is allowed and recoverable

Permanent set

Deformation remaining after the load is removed; it does not recover and it accumulates

CISCA class

A spec's design concentrated load and the companion rolling, deflection, set, ultimate, and pedestal values; the values are manufacturer and spec specific

Rolling load

Dynamic load from a wheel rolled across the panel, reported at 10 passes and 10,000 passes

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What load rating does a data center raised floor need?
• What is rack readiness in a data center?
• What is a raised access floor?
• Do data centers still use raised floors?
• What floor resistance passes ANSI/ESD S20.20?