Both machines live in the same neighbourhood. The Cat 3516 is published from roughly 1450 to 2500 kW (60 Hz); the Cummins QSK60 is rated at about 2000 kW standby from a 60.2-litre V-16. Pick a 3516 trimmed to the ~2000 kW standby point and you have a genuine like-for-like pair: two large-frame, four-stroke, turbocharged diesels aimed at the same mission-critical duty. That is the only fair way to compare them. What follows are the four claims that get repeated in procurement meetings, and the mechanism that decides whether each one survives contact with a real machine room.

Myth 1

The bigger displacement engine runs cooler, so it needs less radiator.

Displacement tells you swept volume, not how much heat ends up in the coolant. A diesel rejects heat down three roughly parallel paths: into the jacket water around the liners and head, into the charge-air after the turbo (the aftercooler), and out through friction and alternator losses. The radiator and its airflow have to carry the jacket-water and charge-air shares away continuously. A larger, slower-turning engine at the same electrical output may actually push a comparable or larger heat load into coolant, because more cylinders and more metal mean more surface shedding heat into the jacket — not less.

When this reverses: in a generously ventilated outdoor canopy with cold ambient air, the larger engine's extra thermal margin becomes free headroom. If your site sits at 1500 m and −10 °C winters with open airflow, displacement does help — there the air is doing the work the radiator size implies.

Myth 2

Tier 2 with no aftertreatment means it's the dirty, outdated choice.

The Cummins QSK60 is EPA Tier 2 certified for stationary emergency standby with no DPF or SCR required. People read "no aftertreatment" as a compromise. For a standby smelter set that runs a few dozen hours a year, the mechanism runs the other way. A DPF needs exhaust hot enough to regenerate; a standby engine that starts cold, ramps to load, and shuts down rarely gets there. SCR needs a urea dosing loop that must stay thawed and topped up to function. Both are systems that fail quietly between the once-a-year times you actually need the genset.

When this reverses: the moment the set becomes a prime mover — peak-shaving the smelter daytime, running hundreds or thousands of hours a year, or sitting in a non-attainment air district. Then aftertreatment is mandatory and runs hot enough to behave, and the calculus flips entirely toward the emissions-optimized build.

Myth 3

Whichever set has the higher peak kVA will catch a big block load better.

Block-load acceptance is not a kVA contest; it is a question of how far the engine speed and bus voltage dip when a large step lands, and how fast the governor and AVR recover — the territory of ISO 8528-5. A potline contactor or a large pump dropping on in one step can be a brutal transient. What absorbs it is the engine's torque reserve at the instant of the step plus the responsiveness of the fuel system, not the nameplate peak.

When this reverses: if your loads are soft-started, VFD-fed, or staged by a PLC so no single step exceeds a modest fraction of rating, transient performance stops being the deciding axis. Then both sets coast through it and you choose on fuel, service network, or controls instead.

Myth 4

Two 2 MW sets that parallel are interchangeable once you bus them together.

Paralleling is where control architecture stops being a footnote. Cummins generator publishes PowerCommand 3.3 with isochronous load sharing and paralleling arrays from 2 MW to 20+ MW in N+1 and 2N, with Modbus/SNMP and black-start. Caterpillar's EMCP 4.2 board consolidates management, diagnostics, and metering on the Cat side. Two sets sharing a bus must agree continuously on frequency and reactive load; the control layer, not the iron, is what keeps them from fighting.

When this reverses: for a single non-paralleled standby set, the paralleling pedigree is irrelevant. A lone 2000 kW machine feeding one bus through one breaker can be either brand on its own merits; the control argument only bites when you put two or more on the same copper.

Topology/standards per the cited standards; all product ratings are manufacturer-stated values from the cited datasheets, current to 2026-06; derived/illustrative figures are labelled as such. This is not an independent head-to-head test. Caterpillar is a brand affiliated with this site; competitor names are used for identification only.