When I compared a 160 MHz AM5x86 and a Pentium Overdrive 83, the POD was about twice as fast in FPU tests I think. Pentium's FPU is far superior to any 486 clone.
However, MP3 decoding should not need a FPU. MP3s are played on a lot of devices without an FPU.
wrote:Those 5x86 chips are "almost" Pentium level of FPU performance aren't they?
I suppose you're referring to the Cyrix 5x86? The AMD 5x86 is basically just a higher clocked 486. The Cyrix 5x86 is basically a castrated Cyrix 6x86, and those aren't exactly known for their fantastic FPU performance.
wrote:However, MP3 decoding should not need a FPU. MP3s are played on a lot of devices without an FPU.
Yep, using integer optimised decoders (usually on ARM hardware) - or with hardware mp3 decoding 😀
I would imagine that most general purpose mp3 playback libraries for dos/windows are not integer based, so will struggle on early processors with weak (ie non Pentium) floating point hardware. There are integer based decoders out there though, but haven't specifically looked at any for Dos/Windows. mpg123 supports integer-only decoding, but it's a unix program.
My collection database and technical wiki:
https://www.target-earth.net
I had hoped that the ultimate 486 comparison would have cleared a lot of 486 vs. pentium fpu myths, but I guess not. Anyway, here's another benchmark score.
When playing 128kbps, 44khz, stereo, Winamp 2.95, no quality downscaling, waveoutput (not directsound output), AWE64Gold ISA sound, Windows NT4 sp6a, Cyrix 5x86-133:
Cyrix 5x86-133 utilization at 55% as seen in the Task Manager
wrote:mpg123 supports integer-only decoding, but it's a unix program.
libmad is integer-based too.
http://www.underbit.com/products/mad/
Winamp libmad plugin
http://www.winamp.com/plugin/mad-libmad-0-15-1b/221520
I am not sure if Winamp's own mp3 decoders have been or are integer based. It's fairly likely.
I used to run XMMS w/ libmad on my Sharp Zaurus SL-5500 PDA back in 2002. That has a Intel StrongARM CPU w/o FPU. Played great and was my first digital audio player. That CPU is surely similar in speed to a 486 on a per clock basis (it ran 206 MHz though).
wrote:I had hoped that the ultimate 486 comparison would have cleared a lot of 486 vs. pentium fpu myths, but I guess not.
feipoa, how about you put the link to your ultimate 486 thread in your signature? 😉
You deserved the honor hehe! ;D
Brilliant idea!
Plan your life wisely, you'll be dead before you know it.
wrote:Brilliant idea!
Cheers mate, welcome to the club ;D
wrote:When playing 128kbps, 44khz, stereo, Winamp 2.95, no quality downscaling, waveoutput (not directsound output), AWE64Gold ISA sound, Windows NT4 sp6a, Cyrix 5x86-133:
Cyrix 5x86-133 utilization at 55% as seen in the Task Manager
Could you try the libmad plugin for Winamp? I'm curious if it's more or less demanding than the old Winamp 2.95 decoder.
I am also curious which of those old decoders are accurately decoding the MP3s. I would not be surprised if they took shortcuts back then. libmad claims to be spot on accurate. I don't know how to test for this though.
wrote:The slowest MHz CPU I've successfully played MP3s on was a Pentium 60. Never successfully played one on a 486, even a DX4 100.
MP3 is fairly floating-point intensive, and Pentiums are a lot better at that than 486.
I was thinking the same thing. Maybe upgrading to a POD would help if all he has is a 486 motherboard. Even if it's integer based, the integer performance of the POD can't possibly be worse than a fast 486 and is probably somewhat better, so nothing would be lost by doing the upgrade.
wrote:I should have mentioned the Pentium 60 managed it with no quality reduction, and it was under DOS using DOSamp.
Those 5x86 chips are "almost" Pentium level of FPU performance aren't they? That'd explain why they seem to have more luck than Intel 486 chips.
Not really. That was the one big advantage Intel held over the others, that their floating point units were stronger. A POD would probably decimate most of the really fast 486's in floating point intensive operations. If you could find a golden Intel 486 that was capable of overclocking to those speeds, it would probably even be stronger where floating point was concerned.
Could you try the libmad plugin for Winamp? I'm curious if it's more or less demanding than the old Winamp 2.95 decoder.
Using the libmad plug-in with Winamp 2.95 results in many crackles. Libmad totally bypassed the recycle bin when I booted his butt.
2.95 is old? I just started using it from 1.X a year ago.
Plan your life wisely, you'll be dead before you know it.
wrote:Using the libmad plug-in with Winamp 2.95 results in many crackles. Libmad totally bypassed the recycle bin when I booted his butt.
2.95 is old? I just started using it from 1.X a year ago.
well I guess that means libmad is too much for the 5x86. Or the Winamp plugin implementation is inefficient. It worked well on my old PDA but that was probably Linux, XMMS and likely a faster CPU.
2.95 is about 10 years old. I use it or Sonique on 9x systems.
