Actually kinda bullshit that most 7zip benchmarks run across all cores when realistically most people will be working with LZMA1 files that can't make use of threads to speed up decompression.
Tell your local redump set packager to use LZMA2!
EDIT: LZMA2 alone don't do shit, more testing underway.
Size: 8525905920
(8131 MiB)
7z x Method = LZMA:24 real 6m24.943s
7z a -t7z -m0=lzma2 -mx=9 real 3m50.738s (and 17GiB of ram use!)
7z x Method = LZMA2:26 real 2m5.788s
Better, but it's still not really using threads.
pbzip2 real 0m53.717s
pbzip2 -d real 0m19.985s
EDIT: That ain't shit either.
lbzip2 real 0m33.681s
lbzip2 -d real 0m19.202s
Better implementation of the same thing, just way faster compression. Ram use was negligible, none of that 17GiB nonsense. And I saw even closer to 2x compression speed on a 6c12t system.
End Edit
And it goes wide, wonderful! But what about the filesizes?
7Z LZMA:24 7854112521 7490MiB
7Z LZMA2:26 7762319712 7402MiB
BZIP2 8037017095 7664MiB
So I'm saying it right now. It's 2023, romsets are not scene bullshit. There's no reason to cling to inferior methods because that's the way it's always been. If you cared THAT much about saving every single byte, you'd be using LZMA2 already. But you're not. So,
Romsets should use bzip2
And really they should be in tar as well.
Oh, and I guess benchmarks should be in bzip2 instead of 7z.