I didn't get a chance to try this. But have you tried winplay3? Its seems pretty darn fast for playing MP3s on slow systems.

i am rather interested in this, especially improving mp3 playback performance on 486 rigs, but i can't bother installing winnt4 on a 486 rig.

i have the following test platform:
mvp3 board with 2mb cache
256mb sdram
geforce2ultra
ymf724 soundcard
win2000sp4
winamp 2.95
192kbps mp3 file for test sample
on this platform, the cyrix 6x86-66 is the only cpu that can i can observe noticeable cpu usage.
using directsound output: ~40% cpu usage from the winamp.exe process
using waveout output: also ~40% cpu usage from the winamp.exe process. however, there is an additional ~10% cpu usage from the system process, due to lack of hardware directsound acceleration.

all other socket7 cpus that i tested, including the second-slowest k5-100, have <5% cpu usage that can't be measured precisely.

i also believe that winplay3 does no good compared to winamp, except for the lack of directsound support(thats the same as winamp with waveout). but i am not sure if it would be perform better on 486 rigs.

noshutdown wrote:

on this platform, the cyrix 6x86-66 is the only cpu that can i can observe noticeable cpu usage.

That sounds right. A Pentium 60 is sufficient for playing high bitrate MP3s.

486s struggle though.

I no longer have any 486s, but I do remember trying to play MP3s on a DX2-66 a few years ago.

Damn near impossible, at least under Win 95.

i486-DX2, 66MHz
16MB RAM
no L2 (not sure how cache dependant MP3 playback is)
SB AWE64
Win95 A running an early release of WinAmp, I think a 2.x version. All visualizations and extras were disabled.

32kb/s, 22KHz, stereo: playable, but Windows GUI redraws were very slow during playback.
48Kb/s, 22KHz, Stereo: would skip every 2 seconds. CPU must've been completely maxed out. Anything higher than 48Kb/s was unplayable.

My observations from this recent topic:

I have this 486 VLB, now equipped with a Cx5x86-100MHz. Also with 16MB (60ns), but with a fresh install of Windows 95 (4.00.950B […]
Show full quote

I have this 486 VLB, now equipped with a Cx5x86-100MHz. Also with 16MB (60ns), but with a fresh install of Windows 95 (4.00.950B), using video mode 1024x768x8 ** edit: with a Cirrus Logic 5428 VLB **.
...
Winplay3 needs Frequency set to half to play a 128kbps MP3.
Winamp 2.05 just manages a 128kbps MP3 at full quality, but only by buffering the entire MP3 in memory.
In comparison: MPXplay in DOS properly plays a 256kbps MP3, when set to buffer the entire file.
...
** Newer observation with Winamp 2.05: **
Indeed a 128kbps MP3 also works with a 1024kB buffer. Maybe it is because I installed the proper Windows IDE driver today.
...
Benchmarks here. This is with a motherboard from 1994, not one of those fancy 1996 PCI boards.

I remember the biggest improvement for MP3 playing was the switch from 70ns to 60ns FPM RAM.

smeezekitty wrote:

I didn't get a chance to try this. But have you tried winplay3? Its seems pretty darn fast for playing MP3s on slow systems.

Winplay3 in Windows NT 4.0 had about a 10% reduction in CPU usage compared to Winamp, however it had an audible skip sound every 3-5 seconds. Perhaps it works better in Win9x?

I've seen a 386DX-40 play an MP3 at 11KHz, with an SB16. Quite an amazing feat, i tell you.

OP updated to show results with Win95c vs. NT4. There was no reduction in CPU usage except when using DirectSound.

feipoa wrote:

OP updated to show results with Win95c vs. NT4. There was no reduction in CPU usage except when using DirectSound.

directsound won't do any good unless you use a pci soundcard with hardware audio acceleration.
i always wanted to try playing mp3 with pci soundcard on a 486 rig, but its still on the long way being put together, hopefully can be done in 2-3 months. would you give it a try?

Could you use a Monster Sound MX400 on a 486? would the drivers load?
That card boosted about MP3 Acceleration, I dunno if was anything but market hype, but its in the Diamond wiki here.
http://en.wikipedia.org/wiki/Diamond_Multimedia

I pulled the PCI sound card out of my 486 after comparing GLQuake results.

IBM 5x86c-100-HF at 100 MHz / Voodoo3 / Win95c

GLQuake results with:

AWE64Gold (ISA) = 19.2 fps
Aureal Vortex2 (PCI) = 18.8 fps
Yamaha 718 w/XR385 (ISA) = 18.0 fps

feipoa wrote:

I pulled the PCI sound card out of my 486 after comparing GLQuake results. […]
Show full quote

I pulled the PCI sound card out of my 486 after comparing GLQuake results.

IBM 5x86c-100-HF at 100 MHz / Voodoo3 / Win95c

GLQuake results with:

AWE64Gold (ISA) = 19.2 fps
Aureal Vortex2 (PCI) = 18.8 fps
Yamaha 718 w/XR385 (ISA) = 18.0 fps

guess quake is not very good at making use of directsound, compared to more modern winamp.
so maybe you can give it another try, and try some other pci soundcards like ymf724 and ess. i also want to try that later, but it can take a few more months for me to finish collecting 486 parts that i need.

AlphaWing wrote:

Could you use a Monster Sound MX400 on a 486? would the drivers load?
That card boosted about MP3 Acceleration, I dunno if was anything but market hype, but its in the Diamond wiki here.
http://en.wikipedia.org/wiki/Diamond_Multimedia

I remember this too with ESS Canyon3D cards. That's the DSP on MX400. The problem is I don't think it works. You see, you would need either an audio program that accesses the DSP or a DIrectshow audio codec that communicates with it. I haven't heard of any successes.

POD100 consumes 38% of the CPU for the same conditions.

Cyrix 5x86-133/4x consumes 55% of the CPU.

I am not sure how I arrived at 77.5% CPU load on a Cyrix 5x86-100. Upon retesting, I get 68%.

Interesting. I remember having Windows 98 on my Dell CPi, and Desperados was laggy. Later I upgraded to 2k, and it was smooth. I presumed WinNT obviously had a performance increase.. I guess that was not it at all? 😕

(I know this is about winamp, not Desperados, but similar I suppose)

i would provide my results that i got recently.

testing platform:
umc8881 mainboard, looks like a hybrid of adi tk8881, soyo syl8884 and pcchips m921.
amd5x86-133 at stock speed.
512kb onboard l2 cache at 2-1-1-1.
64mb fp ram at 0wait.
video card: riva128zx.
sound card: diamond ess1968 with generic ess driver.
windows95C(IE4 not installed)
directx 7.0
winamp 2.95

i tested 128kbps and 192bps mp3 samples. with waveout plugin, cpu usage is 77-79% playing 128kbps ones, and 83-85% playing 192kbps ones. i'll try playing 256kbps later but i guess it won't go up to 100%.
directsound output yields 3-4% higher cpu usage than waveout. this is in total contradiction with my results in win2000, where directsound output uses significantly less cpu than waveout.

Your results indicate that the X5-133 has slightly worse FPU than the Cyrix 5x86-100, which is in line with the results from the ultimate 486 benchmarks. I'm not sure if you using a different Winamp version plays any factor here, but I expected X5-133 FPU results to be a little worse, like 84-87% usage for 128 kbps.

wasnt threre an mp3 player that did without floating point? could be interesting on a 486 imho