ChipFlip
Same GPU, same memory, same specification sheet. $1,110 apart. The Gigabyte RTX 5090 asks $4,600.00 used and the ASUS RTX 5090 asks $3,490.00, and the gap between them is almost entirely about which cooler is bolted on.
Board partners differ less than the price gap suggests
Nvidia sets the silicon, the memory configuration, and the power limits. What a partner controls is the cooler, the factory clock bump, the power delivery headroom, and the warranty. On a 5090 those things matter more than usual because the card pulls a lot of power and generates real heat, but they don't produce a $1,110 difference in performance.
Expect a few percent between the best and worst 5090 coolers under sustained load. Not thirty percent.
What the gap probably reflects
Specific SKU scarcity, seller pricing, and condition rather than capability. On a card this expensive, one seller's asking price can drift far from the market without anything correcting it quickly, because there simply aren't many transactions.
Thin markets produce wide spreads. That's true of every halo product.
Where partner choice does matter
Physical size, since 5090 coolers are enormous and case clearance is a real constraint. Power connector orientation, which affects whether a side panel closes. And warranty transferability, which varies by partner and is worth checking before buying used at this price.
Cooling performance matters too, but less than the size question. Any 5090 cooler from a major partner is adequate; not all of them fit.
Bottom line
Take the ASUS at $3,490. It's the same GPU for $1,110 less, and no cooler difference justifies that. Measure your case clearance before you commit either way, because that's the constraint that will actually bite you on a card this size.