Now, let's say you have a WinXP machine, with 3GB of RAM (memory) installed.
That is, the maximum of memory XP 32-bit will acknowledge. Allright, so you
have your 3GB. And that is what Windows knows. Hence, it makes use of all
the 3GB, if necessary, for loading its many processes.
If now, you were to install a software, that defined a RamDrive, known from
the old DOS days, and which was a big deal back in the 90's; this software
would have to draw its 'diskspace' from the available system memory. That
is, if you defined a RamDrive of 500MB - which converts into just about
0.5GB - this will be subtracted from your 3GB available system memory;
leaving you with only 2.5GB for the system processes. In a sense, you cramp
the system processes, for the benefit of making room for some given process,
which half the time or more, won't really need the high speed of the RAM.
On top of that, we have the PageFile, or SwapFile of Windows. What is that?
Well, when the system processes currently running, are running out of memory
space, Windows will determine which processes are not high-priority, and
will get most or at least some of these processes' stuff off the memory, and
onto the hard disk, in a temporary file. We could in a short sense tell this
SwapFile to be much like a disk image of the contents that do not currently
fit into the available memory on the system. Agree, this turned quite
technical. And why would I ever bother bringing it out?
Let's follow the process for a moment. You are running a system, with some
flavor of Windows, and your screen reader. Whenever the screen reader is to
speak up, it goes to the hard disk, and picks up its predefined sounds, and
'replays' them through your speaker. Depending on the way the voice is
build, and whether you are running a SAPI voice, or some kind of dedicated
voice, it might be necessary for the voice to 'go get' its stuff from the
hard disk, to a higher or less high extend. The voice might want to load a
good amount of its predefined sound library into the memory of your
computer. If your computer happens to use 2GB of RAM for its basic
processes - that is, the processes that is not related to your voice, you
have 1GB left for your voice. Now, many of the SAPI voices might be asking
several hundreds of MB, and one GB could quickly be taken, if you were
running different voices and maybe switched between them frequently.
From time to time, your system needs a bit extra memory access. Maybe your
Email client right now is looking for new mail. That means it will need to
do more than just sit idle on your system. So, it loads a bulk of stuff into
memory. This maybe means, that you are running low on memory. Windows
quickly kicks in, and determines that right at this very second, you dont
use your screen reader voice, and some of the stuff for the voice, can
temporarily be moved out of memory, and 'Swapped' onto the SwapFile of your
hard disk. Once the Email check has been done, the Email client (Outlook or
whichever you are using), goes back to an idle state. That means, that now
it doesn't need much memory any longer, and Windows can swap off from
memory, most of the Email client's stuff; and rather reload your screen
reader and all it is in need of. This swapping can be quite extensive at
times, and if your system has little memory (like 1GB), and is running
modern-day software like Office programs, it wil have to be going on
continually - rather frequently moving things back and forth between your
hard disk and memory. Add on to this, that your running processes might want
access to the hard disk as well.
Keep hanging on to the example with your Email client. If there is new Email
on the net, the software will download this, and save it to the hard
disk.Since the Email client might have the highest priority at the moment,
the swapping of memory contents might be down-prioritized, and could cause
your screen reader not to get hold of its sounds, hence cause choppy speech;
due to the fact of the voice only sounding one piece of sound every fraction
of a second.
Then, back to the RamDrive issue. Now, imagine you had placed all your voice
on a RamDrive, that is, to simply load all of its sounds into RAM. This
would, as we have seen, cause the system memory to be reduced. A reduced
amount of memory, will of course, cause Windows do more excessive swapping;
even on the 'basic processes' of the system. You voice might further likely
be constructed the way, that it always will look in its installation folder,
for the sound-info needed. That is, if your voice is installed in
C:\Crystal, that is where it will look for all the info it needs. No matter
how much you copy it into a RamDrive, it will always look in its instalation
folder.
At the end of the day, you will find yourself with a reduced RAM for all the
basics of your Windows machine, and still having to lookup the voice sounds
on the hard disk all the time. Only, since the RAM is reduced, there is more
swapping, and that means a higher load on your hard disk.
Well, I hope I did not get too techy here. It all cooks down to the fact,
that the more you use your machine's memory for, the more it will have to
swap. And swapping is extremely slow. First, the hard disk works only at a
fraction of the speed of your RAM, and secondly, Windows will have to use
some CPU resources to operate, and determine, the swapping process. The only
real solution, is to give the system as much memory as possible to play
with, so as to minimize the swapping activity. A RamDrive would do exactly
the opposite, and therefore is not being encouraged under Windows systems.
Only time you could have a benefit of a RamDrive, is if you are doing a lot
of temporary saving. I.e with databases or similar. Yet, such software is
already constructed to give you as much for a speed as possible. And again,
you always will have to weigh the benefit of working quickly in RAM, long as
your software stays fully loaded there, as to the drawback of more excessive
swapping due to reduced RAM for the rest of the processes currently running.
For most users, a RamDrive is not even to be considered. Lovely if there had
been a benefit of it, but unfortunately no!
Lastly, you would have to deal with the fact, that the software that defined
your RamDrive, and kept it running all the time, would be at least one more
process, that would need a tiny bit of memory, and a certain amount of CPU.
All in all, you might end up with a harder load on your system as a total;
causing all the gained speed of the RamDrive over the Hard disk, to be
disappearing in overloading your system's resources.
Now, of course, either WE itself, or the voice on its part, could have been
constructed to load as much as posible into RAM, thereby reducing hard disk
usage. Again, that is a lovely dream; unfortunately not the best in real
life. If WE loaded all the voices, and all it ever would need for set files
ETC into RAM at start up, there would be little RAM left for the rest of
Windows, Office, AntiMalware and so forth. And, from what you learned above,
where would that take us? -
Yes, you got it!
We would be left with far more swapping...
The last drawback of swapping, that I will let you take to heart, is the
fact that you physically are moving electronic signals from the RAM, and
onto the hard disk, and back again. Even in a laptop, where the physical
distance between the two components is limited, it still has to be moved
out of one piece of electronic, and into another, and on and off from a
magnetic-based medium. All of this moving, does every so often, cause loss
of data. We well know, from this list, that W.E tends to stop speaking in
the middle of your work. I am not claiming to have the answer in all cases,
but chances are, that some of the speech loss, is due to data loss in the
swapping processes of a highly loaded system. Just the fact, that increasing
your system memory might cause such speech loss to minimize, should stand
for its own proof.
Hope all of this has clarified up at least some of the questions you might
have had.
----- Original Message -----
From: "Lloyd Rasmussen" <[email protected]>
To: <[email protected]>
Sent: Saturday, May 07, 2011 9:48 PM
Subject: RE: speech breaking up and distorting
I don't really know what I'm talking about. But I'll talk anyway. Access
to a hard disk is still not as fast as it is to RAM. But Windows can
access
so much more RAM than DOS could, that I think software can be optimized to
put into memory the things it needs the most. In DOS, before extended
memory was well understood, programs were more disk-intensive and could be
speeded up by converting some memory above the 640K limit into a RAM
drive.
I never hear of anyone promoting the use of a RAM drive in Windows, so
there
must be reasons why it is better to let Windows and your software manage
their memory usage. I would say, though, that if you are trying to use a
concatenative synthesizer and you only have 1 GB of system RAM, it will
sometimes be problematic, and would improve somewhat with 2 or more GB.
You can get a little insight into what is happening if you look at Task
Manager, and set it up so that it shows all running processes. The
keyboard
shortcut for this is ctrl-shift-Escape, and you read some of the output
with
your mouse pointer. Check the status line at the bottom, which tells you
how many processes are running and how much memory and processor
percentage
they are using in total. Use F5 often to refresh the display. Running
processes is the second tab of this tabbed dialog box. Look at this
information from time to time and you can tell when a process has gotten
stuck, when your antivirus program is working hard, etc. You can Google
for
the names of some of the running processes and get some information about
them. But the pages which come up offer to run optimization software for
you. I would not fall for such offers.
Even if I have only two or three applications running, under XP I often
have
45 to 55 processes running, and there is not much you can or should do
about
that. Under Windows 7 I may have 70 processes running, but I have 4 GB of
RAM in those systems instead of the 1 GB in this one or the 2 GB at work.
Optimize with care!
Lloyd Rasmussen, W3IUU, Kensington, Maryland
Home: http://lras.home.sprynet.com
Work: http://www.loc.gov/nls
http://facebook.com/lloyd.rasmussen
-----Original Message-----
From: Joe Lanier [mailto:[email protected]]
Sent: Saturday, May 07, 2011 1:53 PM
To: [email protected]; [email protected]
Subject: RE: speech breaking up and distorting
Lloyd,
In the good ol days, we use to have what was called a ram disk. I am
wondering if that would help. I know that disk access has become much
faster, but, does it meet or exceed the i/o rates possible with physical
memory? Just a thought.
Warm Regards.. Joe Lanier
-----Original Message-----
From: Lloyd Rasmussen [mailto:[email protected]]
Sent: Saturday, May 07, 2011 9:54 AM
To: [email protected]
Subject: RE: speech breaking up and distorting
I think that David's advice was pretty good. The AT&T, Ivona, Neospeech
and
other modern SAPI voices are concatenative synthesizers. They need to
throw
more than a hundred megabytes of data around in order to string speech
units
together. When you alter their speed or pitch, they have to work a
little
harder, still. DecTalk and Eloquence speech will always be a lot more
responsive because they compute speech parameters and generate audio from
scratch, rather than having to look up the speech units in a database.
Microsoft Sam, Mike, Ana and Mary fall in between on the processor
utilization scale, I think.
I don't know whether any of our technology is taking advantage of multi-
core
processors or not. But disk and memory access may still be a bottleneck.
If you want to run these high-end synthesizers, you'll have to limit what
other stuff is running simultaneously.
Lloyd Rasmussen, Kensington, Maryland
Home: http://lras.home.sprynet.com
Work: http://www.loc.gov/nls
> -----Original Message-----
> From: Jenine Stanley [mailto:[email protected]]
> Sent: Friday, May 06, 2011 8:32 AM
> To: [email protected]
> Subject: speech breaking up and distorting
>
> I have 2 computers and they both are having an odd issue lately with
> their synthesizers. Both run WE 7.5
>
> Desktop runs Windows7, with Ivona Kendra speech synthesizer.
>
> From time to time the speech will break up, like an old radio that's
> poorly tuned. Sometimes also when adjusting the pitch, which I like at
> 2, the speech will become distorted as if it's too loud though the
> volume hasn't changed at all. The breaking up tends to fix itself but
> the distortion requires a complete reboot of the computer, not just
> shutting down WE and restarting.
>
> Laptop is running Windows XP with the AT&T Crystal synthesizer. I have
> WE start prior to log in and nothing has changed with the DECTalk
> speech. The last several days, when I boot up, and the speech changes
> over to Crystal, the voice is very distorted as if through a bad AM
> radio speaker. This is continuous, unlike the desktop computer issue.
> The only way to fix it is to exit WE and restart it. Then the speech
> returns to normal.
>
> Any idea what might be going on here?
If you reply to this message it will be delivered to the original sender
only. If your reply would benefit others on the list and your message is
related to GW Micro, then please consider sending your message to
[email protected] so the entire list will receive it.
GW-Info messages are archived at http://www.gwmicro.com/gwinfo. You can
manage your list subscription at http://www.gwmicro.com/listserv.
If you reply to this message it will be delivered to the original sender only.
If your reply would benefit others on the list and your message is related to
GW Micro, then please consider sending your message to [email protected] so
the entire list will receive it.
GW-Info messages are archived at http://www.gwmicro.com/gwinfo. You can manage
your list subscription at http://www.gwmicro.com/listserv.
