LAME VBR Performance Study

Three albums · V0 through V4 · x64 builds

VBR encoding performance across official and current builds

The same three WAV albums were encoded at V0 through V4 with official LAME releases from 3.95.1 to 4.0. The current LAME 4.0 cVBRb MinGW and Visual Studio builds were then tested separately with the same inputs and settings.

Current builds Official releases Download data

Measured summary

Official 4.0 and 3.101 are effectively identical in this test. Their average times differ by no more than a few hundredths of a second at each setting.

The Visual Studio cVBRb build is the fastest current build. It is around 14–15% faster than the cVBRb MinGW build across V0–V4.

3.100 retains the 3.99.5 output curve but is substantially slower. Official 4.0 restores and improves the later-release throughput.

Current LAME 4.0 builds

Average elapsed seconds per album. Lower values indicate faster encoding.

Line chart comparing official LAME 4.0 with cVBRb MinGW and Visual Studio encoding times from V0 to V4 — The two cVBRb builds produce effectively identical output sizes, so their principal measured difference here is throughput.

Build	V0	V1	V2	V3	V4	Mean
Official 4.0	31.48 s	30.20 s	29.34 s	28.44 s	27.63 s	29.42 s
4.0 cVBRb MinGW	32.22 s	30.86 s	30.13 s	29.16 s	28.37 s	30.15 s
4.0 cVBRb Visual Studio	27.46 s	26.48 s	25.68 s	25.14 s	24.35 s	25.83 s

Current-build differences

Setting	Official 4.0	cVBRb MinGW	cVBRb Visual Studio	VS gain over MinGW	VS gain over official 4.0
`-V0`	31.48 s	32.22 s	27.46 s	17.3%	14.6%
`-V1`	30.20 s	30.86 s	26.48 s	16.5%	14.0%
`-V2`	29.34 s	30.13 s	25.68 s	17.3%	14.2%
`-V3`	28.44 s	29.16 s	25.14 s	16.0%	13.1%
`-V4`	27.63 s	28.37 s	24.35 s	16.5%	13.4%

Official-release performance

The complete three-album timing history from 3.95.1 through 4.0.

Line chart showing average VBR encoding times for official LAME releases 3.95.1 through 4.0 — 3.98.4 and 3.99.5 mark the major throughput improvement over the earlier releases. 3.100 regresses sharply, while 3.101 and 4.0 restore faster performance.

Bar chart showing mean encoding time across V0 through V4 for official LAME releases — Mean elapsed time across all five VBR settings provides a compact view of each release's overall throughput on the test system.

Official release	V0	V1	V2	V3	V4	Mean
3.95.1	68.37 s	70.46 s	71.64 s	84.17 s	83.02 s	75.53 s
3.96.1	69.84 s	72.03 s	73.42 s	86.56 s	85.35 s	77.44 s
3.97	62.35 s	63.33 s	64.25 s	73.59 s	73.41 s	67.39 s
3.98.4	36.56 s	35.49 s	34.98 s	33.66 s	32.98 s	34.73 s
3.99.5	35.52 s	34.12 s	33.06 s	32.16 s	31.33 s	33.24 s
3.100	67.36 s	66.56 s	65.02 s	64.10 s	63.30 s	65.27 s
3.101	31.55 s	30.23 s	29.31 s	28.47 s	27.60 s	29.43 s
4.0	31.48 s	30.20 s	29.34 s	28.44 s	27.63 s	29.42 s

Fastest official release across 15 runs

There are three albums and five VBR settings. A win is assigned to the shortest elapsed time for each album/setting combination.

Release	Fastest-run wins
3.95.1	0
3.96.1	0
3.97	0
3.98.4	0
3.99.5	0
3.100	0
3.101	8
4.0	8

3.100 compared with 3.99.5

These two releases produce the same average output-size curve in the companion study, but their timing differs substantially.

Setting	3.99.5	3.100	3.100 slower by
`-V0`	35.52 s	67.36 s	89.7%
`-V1`	34.12 s	66.56 s	95.1%
`-V2`	33.06 s	65.02 s	96.7%
`-V3`	32.16 s	64.10 s	99.3%
`-V4`	31.33 s	63.30 s	102.1%

Download the measured data

Official summary CSV Official raw CSV cVBRb summary CSV cVBRb raw CSV

Results describe this test system, these x64 binaries and the three submitted WAV albums. Timing differences should be retested when the compiler, libraries, processor or build configuration changes.