Yup, works here.
There are some differences in the implementation though.
The ones I've got, are AC coupled into the onboard sound.
You plug amplified computer speakers to the green LineOut
on the motherboard. If you weren't using Front Panel Audio,
and the sound was AC'97 based, then you'd also have to check
that the two jumpers are on that 2x5 FP_Audio header. I don't
know which side of the header pins, the AC coupling is placed.
But it would be another thing to check (signal continuity).
If you had a PCI sound card, you'd still need to plug into the
green motherboard Line Out, to hear the message. A PCI sound
card has no way to tap into those sound signals.
It's also possible that at least one motherboard, used the case SPKR
for output.
Another reason for failure of Vocal Post, is if the Vocal Post
serial EEPROM got erased. Winbond provided an editor program,
where the user could re-program the vocal post messages. It
would take a long time (30 minutes?) to reprogram the chip.
On motherboards that lacked BIOS controls for the feature,
simply erasing the messages you didn't want to hear, was an option.
The 8 pin DIP serial EEPROM can also be unplugged, which is
another quick way to shut the thing up.
The Vocal POST on some motherboards, cannot be disabled. Later,
BIOS settings (sometimes more than one setting) were available.
Both my Vocal Post motherboards have BIOS controls.
Those settings allow "announcement" messages, such as "now booting".
Or to disable the thing entirely. I don't remember the exact wording
on the setting, but it allowed changes like that.
Two of the error messages are timer based. One of those allows
an error message to be emitted, if the CPU crashes before executing
any BIOS code. And one message, is controlled by a wire to the CPU
socket - if no CPU is plugged into the socket, that triggers a
specific message. Other messages, late in BIOS POST, are controlled
by BIOS code. And the BIOS code has to write the number of the desired
message, to get the message to be played. But a few of them are
"hardware" based. For example, if your motherboard crashes during
the BIOS quick memory test, one of the timers picks that up.
I learned these things, by reading the Winbond source files they
thoughtfully left in the Winbond editor package folder. It made things
a lot clearer. The thing is a two chip solution, a Winbond controller
(independent of the motherboard CPU) and the serial EEPROM. And what
is loaded into the serial EEPROM, is both a "program" as well as a
series of "voice samples" played at around 4KHz or so. I think
there is some way to adjust the bandwidth on the thing, at the expense
of making the (uncompressed) voice samples larger. In some cases,
you can barely fit them all in the EEPROM (so recording an extremely
long message with the Editor program, won't work out very well).
It was never designed to be high fidelity, which anyone who
has heard it can attest to
Paul
