- Messages sorted by: [ date ][ thread ][ subject ][ author ]
- Next message: mcrawfor: "Pronunciation Difficulties"
- Previous message: Gary: "Re: Query from a co-worker"
advantage of the obvious new capabilities connected with PCI, however I did
notice an uncharacteristically favorable reference to the Fiji card on
Dragon's hardware compatibility list. They referred to it, if memory serves
(their relevant page on tech support isn't responding at the moment, so I
will paraphrase here), as an exceptionally good card. And then I noticed
on the Turtle Beach site that they call it a sound capture card and didn't
mention anything about 64, 128, or 256 voices or other typical details
concerning output that are of no help for VR. I hope somebody will compare
it with a good PCI card and tell us about their conclusions.
-----Original Message-----
From: Jim Green <jgreen@@robotvision.com>
To: voice-users@@voicerecognition.com <voice-users@voicerecognition.com>
Date: Tuesday, August 04, 1998 1:32 AM
Subject: Re: Sound cards - PCI vs. ISA
>You correctly stated that good frequency response and low noise can improve
>VR accuracy. Those advantages could outweigh the down side of using up CPU
>cycles inputting high bandwidth digital sound input over an ISA bus. The
>point I was trying to make is, if you have a marginal CPU for NatSpk 3.0
>with Best Match, a PCI sound card will allow the maximum number of CPU
>cycles to go to NatSpk rather than using them to run the sound card.
>Aren't there some good PCI sound cards? I have a Creative Labs / Ensoniq,
>and it seems to work fine. (Although I haven't done any A/B comparison
>tests.)
>
>------------------------------------------------
>At 09:50 PM 8/3/98 -0500, Holland St.John wrote:
>>The Fiji is an ISA card, not PCI. I have not personally tried enough
>>cards to say from experience whether PCI vs ISA makes a difference in VR
>>use on typical machines, but reading the comments of many other users
>>over the past year, particularly those who have used the various Turtle
>>Beach cards with similar specs, but some PCI and some ISA, leads me to
>>believe that the ISA vs PCI interface is not significant in VR accuracy.
>>Other aspects of sound card design are far more important.
>>
>>Holland
>>
>>
>>Jim Green wrote:
>>>
>>> There seems to be some confusion as to the merits of PCI vs. ISA sound
>>> cards. Let me try to clarify this. Holland correctly states below that
a
>>> PCI card puts less load on your CPU than an ISA card. But he didn't say
>>> why. (He also recommended a particular sound card as being good for
NatSpk
>>> because it has "flat frequency response and low noise in the input
>>> circuits" which will make it easier for NatSpk to "decide" which words
to
>>> display. This is also very true.)
>>>
>>> The ISA bus was the original I/O bus released by IBM in 1985 with the
"AT"
>>> computer, with a 8 MHz 80286. It was good for I/O devices with very low
>>> bandwidth such as slow printers, 300 baud modems, etc. The ISA bus is
very
>>> inefficient because most transfers to and from boards are "programmed
I/O",
>>> that is, a number of CPU cycles are required to transfer one byte from
the
>>> board to the CPU. With high bandwidth boards, such as digital sound
input
>>> cards, a significant fraction of the CPU can be consumed just inputting
the
>>> digitized sound to RAM. Also, digitized sound input can consume the
whole
>>> ISA bus bandwidth.
>>>
>>> The PCI bus is a recent Intel invention to overcome the bandwidth and
CPU
>>> intensive limitations of the ISA bus. It uses its own controller to
unload
>>> the CPU. Its bandwidth is very high and growing.
>>>
>>> If you have a very fast CPU (like a P-II 400) the amount of CPU stolen
by
>>> an ISA bus sound card will be a small fraction of available CPU cycles.
As
>>> your CPU gets slower, however, a larger and larger percentage of your
CPU
>>> will be consumed just inputting sound card data. Thus, contrary to what
>>> was said here by another person, the slower your CPU, the more you will
>>> benefit from a PCI sound card. Of course, this is "all other things
being
>>> equal", such as "frequency response and noise in the input circuits" as
>>> Holland mentioned.
>>>
>>> ---------------------------------------
>>> At 11:57 AM 8/3/98 -0500, Holland St.John wrote:
>>> >Jim,
>>> >
>>> >The Fiji is a good card because of flat frequency response and low
noise
>>> >in the input circuits. There is no extra processing on the card that
>>> >helps with recognition (there are lots of other extra chips that help
>>> >with music production, mixing, etc, but those are irrelevant to VR).
>>> >
>>> >The Fiji will take some of the load off the CPU because the CPU has
less
>>> >noise to deal with, and therefore Nat and the CPU can "decide" easier
>>> >which words to display.
>>> >
>>> >Holland
>>> >
>>> >
>>> >Jim Walsh wrote:
>>> >>
>>> >> I am a bit confused. I just bought a PII 400 w/256 ECC SDRAM with a
>>> >> 7,200 EIDE Seagate drive. I am considering buying the TB Fiji card.
My
>>> >> thinking is that the Fiji has a built in processor which may assist
the
>>> >> PII in some of the processing. Will the extra processor be a benefit?
>>> >> What about the MMX instructions? Anyone have a clue?
>>> >>
>>> >> Robinson, Robert wrote:
>>> >> >
>>> >> > Our experience with both NS (not including BestMatch) and VV98 is
that
>>> you
>>> >> > can run the software with a P 200 with 128MB RAM, but it is slow. A
>>> Pentium
>>> >> > II 400MHz with 256MB RAM does much better but CPU utilization
during
>>> speech
>>> >> > processing is still "100%".
>>> >> > Robert Robinson
>>> >
>>
>#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~#
>James E. Green, Ph.D. # Jim@@RobotVision.com # 703-757-5140
> Robot Vision Corporation
> Computer Vision System Design & Implementation
>9978 Blackberry Lane # 6 Mountain Laurel Ranch
>Great Falls, VA 22066 # P. O. Box 569, Leakey, TX 78873
> http://www.next-wave.com/jim.htm
>
- Next message: mcrawfor: "Pronunciation Difficulties"
- Previous message: Gary: "Re: Query from a co-worker"
 |  |