lame benchmarks (pun intended)

[johnklos]

I want to see Vampire benchmarks, too!

It's interesting to see the progression:

Code: [Select]

33 MHz m68040 (Mac Quadra 605):
3582.799u 6700.611s 3:08:39.78 90.8%    0+0k 0+798io 6pf+0w

50 MHz m68060 (Amiga 1200, Blizzard 1260):
1324.752u 513.886s 31:25.96 97.4%       0+0k 0+556io 0pf+0w

1.5 GHz PowerPC G4 (Mac mini), no Altivec:
31.992u 0.470s 0:37.09 87.5% 0+0k 0+17io 5pf+0w

1.2 GHz Raspberry Pi 3:
30.575u 0.320s 0:34.71 88.9%    0+0k 0+16io 1pf+0w

2.05 GHz AMD Athlon 5350:
10.957u 0.119s 0:11.11 99.5% 0+0k 0+6io 0pf+0w

2.3 GHz NVIDIA Jetson TK1:
10.686u 0.129s 0:10.97 98.4%    0+0k 0+7io 2pf+0w

3.6 GHz AMD FX-8150:
5.641u 0.170s 0:05.81 100.0% 0+0k 0+3io 0pf+0w

4 GHz Core i7 6700k:
3.220u 0.050s 0:03.27 100.0%    0+0k 0+6io 0pf+0w

3.4 GHz AMD Threadripper 1950x:
2.956u 0.309s 0:03.62 89.7% 0+0k 0+3io 0pf+0wThat breaks down to this for performance (audio frame per second per MHz):

Code: [Select]

m68040: .031
m68060: .1145
PowerPC G4 (7447a): .216
ARM Cortex A53: .284
AMD Athlon 5350: .464
ARM Cortex A15: .423
AMD FX-8150: .5
Core i7 6700k: .805
AMD Threadripper: .948That means that, per MHz on LAME encoding, an AMD Threadripper is about thirty times faster than an m68040

[johnklos]

I should note that this doesn't take turbo clocks in to account.