486SX 25 MHz (without FPU, co-processor... nothing), 16 MB, Win 95.
I'd need a program with an integer decoder. Is there any chance?

Very hardly. I always heard that the minimum requirements for real-time MP3 decoding were 486DX4-100 or Pentium 90

I think even a 486DX2-66 wouldn't process enough MIPS to decode MP3 in real time no matter how much you hack the decoder to speed it up.

Anyway, MAD is an integer MP3 decoder. Give it a try. :B

MAD! What was I thinking?
OK, thanks Roberto.

All FhG decoders (including the one in Winamp) are integer-based as far as I know. Very likely faster than MAD.

I remember I could play mp3s on my old 486 (DX2 50MHz; 16MB Ram; Win95) in real-time. No problem. But I don´t know which player I used.

Maybe you can play with players like Xing and Digideck, that cut lots of "edges" to make the decoding faster. Of course, quality suffers (they reduce accuracy, etc.)

IIRC, there was a 486 version of DigiDeck. Google around.

@S_O: Maybe you used Winamp with the Nitrane decoder. That decoder is hell fast, but we all know by now how buggy and low quality it was.

I recall a small and free mp3 player avaliable for download at the Fhg Institute(Frahunhofer) website's until 2 years ago.
If I'm not mistaken, it was able to play mp3s in i486s machines without a problem.
They even had a 16bit version as well.
(WinPlay???)

//edit by LIF//:
Get it here.

The other thing, if you can find a compatible player, is to ditch the overhead of W95 and
run dos... or linux without Xwindows.

I don't think DOS would be of much help. Even mpxplay requires a 486DX4-100 to decode MP3 (or at least a Pentium 120 for Vorbis).

Pentium 90 needed to decode MP3? Not true.

I used to listen to MP3s on a Pentium 60. (The one which is compatible with the 486 "overdrive" thingybobby. The socket was incompatible with Pentium 66+.) 44kHz 16bit stereo used to work fine in both Winamp and winplay3, with Windows 95.

But there was a catch: I couldn't run anything else while listening to MP3s or the sound would skip! I had to either downmix or downsample if I wanted to do something else.

Nullsoft DOSAMP worked in DOS, but IIRC my DOS sound drivers were flaky and I could only get 22kHz playback. Later, DAMP worked fine IIRC.

I've still got DOSAMP on my HDD.

i used to play mp3z on an old 486dx 66 machine, overclocked to 80mhz. i had to check the "half" quality in the decoder box options in winamp, however. but it sounds to me like what you have is a 386...i don't think they made 486 chips at that speed...but i could be wrong. i highly doubt you will be able to get an mp3 to play with 25mhz without decreasing the quality somehow.

There were definitely 486SX-25's.

http://www.hardwareanalysis.com/content/image/10111/

Take a look at the date on the chip (1989)... it must have been one of the first (if not the first) 486, since 386's were still rather high end and pricey just a year before that in 1988 (I remember clearly, as I was taking some computer science courses that year at a technical college).

Ahh, the good ol' days when 640k really was just about enough for anyone...

i stand corrected.

btw - i read somewhere that the quote you are alluding to was never said by bill gates...its just urban legend.

Could be... if it's an urban legend, it's been around a long time. I think he did say it (or something like it), but of course it's possible he didn't.

As I said, earlier versions of Winamp used nitrane, that used to cut around the edges in order to play faster. So, yes, you could decode them in real time, but with reduced quality (reduced decoder precision).

Probably the decoder was running at "half" or "quarter quality" mode. You didn't even have to enable that, the decoder switched to a lower quality mode automatically if there wasn't enough processing power. Same thing with the current decoder, although it doesn't cut around the edges anymore.

And probably the same thing happened with WinPlay3.



Anyway: Yes, for full quality MP3 playback, you need more than a Pentium 60. Or a 486 other than DX4-100/120

http://www.urbanlegends.com/celebrities/bi...tes_memory.html

Could just be me, but Gates' denial smacks of avoidance and "thou dost protest too much." It comes across as a marketing statement, not a factual statement.

Not that it matters -- who really cares whether he said it or not? Talk about beating the skeleton of a dead horse...

(OK, I was the one who sort of brought it up...).

SX=sucks. 486SX processors actually ran SLOWER, for the most part, than 386DX.

agreed. i remember i had an old 386 25mhz laptop that ran sooo much faster than a 386sx desktop that i had...man that thing was SLOW, even in windows 3.1

My old 486 DX2/66 could not play MP3 in real-time at full quality (only at half quality). I bought a Pentium Overdrive on eBay for like $20 a few years ago. Running at 83MHz, the Pentium overdrive played MP3s fine at full quality.

Haha, I remember all the hype about Pentium Overdrive sockets when people were buying 486s. It was sooo important to have one. When the Pentium overdrive finally came out, hardly anyone got one.

486SXes came out after DXes. There was a series of 486DXes that were defective (the FPU and I believe something else were bugged). They disabled the FPU and sold them as a "discount processor". The Pentium FDIV bug wasn't the first time Intel had trouble with their FPUs. I had a DX2/66 but it turns out the DX50s (different from the DX2/50s) were faster. They were uncommon though because the 50MHz bus wreaked havoc with lots of VLB cards of the era.

Anyhow, there were still 486es in use when mp3s started circulating when I was in college (DX4/100's at least) and they played them OK IIRC. A 486SX25 is probably a long shot, but are the low-power embedded processors in portable players that much faster? In this thread it's mentioned that they use about 30MHz to decode mp3s...


HAHA, those things were surreal.

30mHz DSPs. Keep in mind these are processors specialized in signal processing, not general purpose processors like CPUs.

Besides, these DSPs aren't running a fully featured operational system in the background.

I remember somewhere reading about an MP3 player for Playstation. It might have just been in development, but the PS certainly doesn't have a 60MHz processor.

I never heard of anything like that. As far as I know, the PlayStation can decode ADPCM at most. (although MP2 would be a possibility)

And it's CPU runs at 33mHz. Certainly not enough for MP3, unless using some low quality mode.

Try SkyOS B)

???

http://skyos.org/about.php

The PSX mp3 player was an external one.

FYI: My sister can run Windows 95B w/ CoolPlayer on her Pentium 60 w/ 64 megs of RAM, without a problem. It plays both MP3 and Ogg Vorbis smoothly.

Later, when Msoft was pushing OS/2, gates was quoted as saying that if you want multitasking, you need 4 meg of RAM, like it was a basic law of nature. At the time I was running a Sun with less than that which certainly had multitasking.

On my first job we had a computer with 1/2 meg serving 10 users at a time (but 10 meg hard drive for swap space)

A -real- Pentium 60 is enough to playback MP3 (perhaps the overdrive things were much slower) - I had one of the first real Pentium 60s and it played MP3 (well... 128kbit/s ) without any problems at full quality. I used the Winamp versions that were available at that time (perhaps they used the Nitrane decoder, I'm not sure about that). I remember that playing MP3s caused about 60% load (in average)... I was very pleased to see my next PC only had 15% load while playing MP3s.

Regards, fileman.

Oh, I forgot something: there are player modules for the parallel port available. So the PC (can be a 286, I guess) is just to be used to get the data from HD or disc and send it to the parallel port.

Ah... I found a link:
http://www.heise.de/ct/ftp/projekte/mp3pla...1/default.shtml
(german)

Regards, fileman.

I started out with a 486/66. Using a variety of players I could never get it to play mp3's without skipping.

I had to decode to wav if I wanted full quality.

First, thanks everybody.
This is a little "experiment" with an old PC for a friend of mine. Hehe, when I opened the box... there were ten years of history and dust!
About the decoders: The quality is not my first goal, obviously. I've tried a few of them, but the processor hasn't power enough for real time decoding. I keep searching a Windows player, but it appears that a few MHz more are needed... even WinPlay 16 bits doesn't work without integrated co-processor.
My next movement will be a mini-dedicated OS in a floppy disk. More fun.
And if nothing work, I always have the solution of decode to WAV.

In that case, I suggest something like DOSAMP or DAMP.

I've got DOSAMP 0.8 and DAMP 0.95 WIP 7.

- MAD uses 32 bit x 32 bit => 64 bit multiplications.
- These need around 40 clocks on a Intel 486 (actually 3 clocks more than on a Intel 386).
- On a 25 MHz CPU these are 625000 multiplications per second.
- per Sample and Channel these are 625000 / 44100 / 2 = 7 multiplications

Not the smallest possibility of a chance ...
It looks like also a quarter sampling rate synthesis is not possible.

Not knowing anything about this whatsoever then, what sort of speed are the processors in items like iRivers, Car Stereo's that accept MP3 data discs etc,...?

I know they dont have to maintain an OS at the same time, but I am curious

70MHz ARM for entry-level models, 90MHz ARM for higher models

Well, it's a fact: A 486SX @ 25MHz can't decode MP3 in real time.
This test was made by professionals. Don't try to do this in your house!

Here's what I just went through/found out through personal experience:

I had a 486 DX2/66 I wanted to use for playing mp3s. I installed DOS and tried every DOS based mp3 player I could find. I finally found that mpxplay came *the closest* to being able to play most mp3s. However, depending on the bitrate there was always some skipping with mpxplay (I lightened the load on the CPU by disabling all display-type features of mpxplay as much as I could)

I had to overclock the cpu to 80 MHz and then I could safely play any mp3 I tried. And again, this is all in DOS only mode - no windows overhead.

So in the end I'm pretty sure you need at least a DX/2 80 MHz 486 for mp3s (unless you find some quality decreasing mode that I couldn't)...by the way mpxplay can also play mpcs (for sure) and oggs (I think).

I have an old PS/Valuepoint here (used to be a 486SX/25) with a 60 Mhz Pentium Overdrive processor. A year or so back I tried to get it to play mp3s. I finally managed, using mpg123 under FreeBSD - it made some strange warbling noises while filling the buffer (the first few seconds of the first song in the playlist), but otherwise it played flawlessly.

The same player under Linux didn't cut it though - probably some differences in the sound drivers or something. Neither did any DOS or Win3.1 based player I could find. So it's definitely a close race for those beasts

Now i will share my experience with yours. I was there at the time playing with those things, and i still have old CPUs and motherboards in working conditions (but i stripped those stupid Ni-CD batteries before they melted the circuits with acid leaks).

From my experience, the fastest player i found some years ago was mpg123 which i tried it with some linux distro using 2.2 .x kernels. I could make a Cyrix 486DX4-100 chip decode in realtime 128kbps stereo files at full 44.1khz. But i couldn't do that with stock binary mpg123, i had to compile from sources, and use 486 optimization. Yes, i tried both i386 and i586 optimizations, and it actually was worse. It had to be i486. I recall realtime decoding will use about 80% cpu usage, i guess a 486DX4-75 or even DX2-66 could do, its a matter of trying.

That very same CPU, using win95 and winplay3 (thats the FHG player, first it was a playing time limited demo program with 80$ cost, then they later released post 2.0x version uncripled and free) was not be able to achieve the same, it had to use "half" quality (22khz) or "mono". I also tried winamp of then, and i was much, much worse, and it also wanted lots of memory... Cubic Player using MS-DOS (a tracker player) added mp2/mp3 capabilities, but it was also similar to winamp in perfomance. Only mpg123 made a significant difference. Oh speaking of mp2, i used a "Phillips" windows player and mp2 files were much much easier to reproduce with full quality and any bitrate. (Consider toolame!

About urban myths... The 386sx is a 32bit CPU with a 16bit bus. This cpu has no cache and could use 16bit 286 style memory. It doesn't have a FP unit either, but you could buy a separate math coprocessor (387) for real FP operations in the same way you could with 286, and 8086. Typically those with CAD programs would.

The 386dx is a 32bit CPU with a 32bit bus. I has no cache either and same story about not having an FP, but math coprocessor available. Due to the low memory speeds, external cache (known nowdays as L2) improved perfomance dramatically. It was common to see motherboards with 64kb to 256kb sram cache.

The 486dx is a 32bit CPU with a 32bit bus. So far bus speed was tied to the CPU speed (well actually 386 needed twice). a 486dx2 cpu is a CPU with half bus speed (ie. dx2-66 used 33mhz). The 486 had 8k write thru cache (L1), motherboards would typically come with 256kb sram cache or more, like they did with 386. The 486DX had the "math-coprocessor" builit in, in other words, has its own FP unit so didn't need an external one.

The 486sx is EXACTLY the same as the 486dx, only without the FP unit (or it was disabled/damaged, etc. In fact you can run a 486dx like an sx...) just like earlier x86 CPUs without FP, you could simply add an external math-coprocessor and will behave identically to the 486DX (minus L1 cache perfomance?)

The 486dx4 is like 486dx2, only this time it runs the bus speed at a THIRD (hehehe, dx3 was trademarked or something silly).

Cyric 486slc is simply like a 386sx with 1k L1 writeback cache, and 486dlc is the same but like a 386dx, They could be used as a chip upgrade (socket compatible).

There was an Intel 486dx-50 and an later a 486dx2-50, the former obviously faster, but a lot hotter and mobo manufacturers of then had more trouble working with speeds beyond 33Mhz.

It would be fun to use a "DSP in a PCI/ISA" slot as a "sound coprocesor", then you could use really old machines

Oh yes, most of these chips were fanless, fan coolers started showing with dx4 and pentium line. I wonder if the V-Dragon chip is fanless too? Cheap PCs are coming

i recall that most 486 proc's had fans...i had around 4 or 5 486dx2 66's and they all had fans. i was only able to overclock one of them to 80mhz tho.

Well, but this one doesn't have fan... it doesn't have heatsink... it doesn't have socket... it doesn't have...

Are you sure it is a CPU? Maybe you are loking at a chipset?

Yes my friend, the processor is directly welded with the motherboard, just like the chipset. I said that this is a true historic piece.

Funny my first computer was EXACTLY like that. 486SX, soldered to the mainboard, no heatsink, no fan.

Funny, that's pretty much what it looks like in my four year old iMac Anyway, have you considered AAC? They seem to be easier to decode than MP3, and do it faster on my computer.

On theory at least, AAC demands more processing power than MP3 for decoding.

That's why it needs a 25mHz DSP for real-time decoding, while MP3 demands a 20mHz one.

More info:
http://www.iis.fraunhofer.de/amm/techinf/a..._dsp/index.html

Well, MP3 must be the only compressed audio format that my friend knows... (and maybe WMA)
But since I'll have this powerful machine a few days, I'll try MPC, FLAC and other formats, just for curiosity.
At least it can play WAV and CD Audio.

A 486 probably had just 1 integer multiplier, which is only active when it gets to a multiply instruction. THe last DSP chip I designed (about the time those 25 mhz 486's came out) had over a dozen multipliers which all ran pretty much every clock cycle.