What balanced actually gains

Balanced outputs are sold on a number: +6 dB, four times the power. This list checks that number against the makers’ own figures. Every row is one manufacturer’s balanced power divided by its own single-ended power at the same load, so nothing here is compared across brands or across loads.

37 of these 124 reach the +6 dB the marketing implies, and 13 gain nothing measurable at all — the same amplifier with a different socket on it. A balanced output is a design decision, not a law of physics, and the spread below is what that looks like when you read the spec sheets instead of the box.

Sorted on one published figure

SourcePriceBalanced gainBalanced power at that load
FiiO M17$1799+7.8 dB at 32Ω3.0 W
RME ADI-2/4 Pro SE$2499+7.5 dB at 32Ω10.6 W
FiiO KA13$79+7.2 dB at 32Ω260 mW
Sony NW-ZX707$898+6.6 dB at 32Ω115 mW
Sony NW-WM1AM2$1399+6.2 dB at 32Ω125 mW
Sony NW-WM1ZM2$3700+6.2 dB at 32Ω125 mW
ddHiFi TC44Grip$160+6.2 dB at 32Ω330 mW
Tanchjim Space$70+6.0 dB at 32Ω500 mW
Truthear SHIO$70+6.0 dB at 32Ω500 mW
Moondrop Dawn Pro$75+6.0 dB at 32Ω500 mW
Qudelix 5K$109+6.0 dB at 32Ω500 mW
Qudelix T71$229+6.0 dB at 32Ω2.0 W
SMSL VMV P2 MK2$1160+6.0 dB at 32Ω8.0 W
Schiit Ragnarok 2$1499+6.0 dB at 32Ω24.0 W
Headamp GS-X Mini$1795+6.0 dB at 32Ω5.1 W

The table shows the first 15 of 124. The rest are in the full source list, where every figure is labelled with where it came from.

156 of the 280 sources in this database are not on this list because their makers publish no separate rating for a balanced output — either the product does not have one, or it has one and the spec sheet treats both sockets as the same thing. Without two figures from the same maker at the same load there is nothing here to divide.

This is not a ranking of sound quality and cannot be. It sorts one number that manufacturers printed, and manufacturers do not print anything about how their products sound. Every row links to that product’s own page, where each figure is labelled with the document it came from.

Same database, different question