Please keep us informed regarding the performance which you experience, Jamie.
- Jonathan [ still running on Win95 w/single CPU ]
At 08:56 AM 3/11/98 -0500, Joseph S. Wisniewski wrote:
>I've used a machine that has dual hypersparc RISC processors and an array
>of four C50 floating point DSP chips. Figure between 1 and 1.5 billion
>instructions / second.
>
>Stephanie Sennif at MIT had a recognizer that needed 6 UNIX workstations
>just for the front end.
>
>And the all time winner: a system at AT&T that ran on a computer they
>call "Aspen" which has an array of 128 processors.
>
>Now, for what you're doing, I don't believe any of the commercially
>availiable Windows recognizers (Dragon, IBM, or L&H Kurzweil) are
>multithreaded, and none will take advantage of the extra processor.
>
>Now, UNIX stuff has been multithreaded and multiprocessored for so many
>years that all the major UNIX recognizers (HARK, Nuance, PureSpeech, etc)
>would take full advantage of such a system.
>
>Apparently, whatever your building is multithreaded.
>
>On Wed, 11 Mar 1998, Jamie Ferguson wrote:
>
>> Has anyone tried using voice recognition on a machine with multiple
>> processors? I'm about to experiment with a 200MhZ Pentium with 2
>> processors and 512MB of RAM.
>>
>> (There actually is a reason for such a machine -- I develop software that
>> takes a VERY long time to build, and 2 processors significantly reduces
>> the amount of time.)
>>
>> FYI I use DragonDictate 3 and NaturallySpeaking Deluxe.
>>
>> Jamie
>>
>
> Joseph S. Wisniewski | Views expressed are my own, and don't reflect
> Ford Motor Company | those of the Ford Motor Co. or affiliates.
> Project Sapphire | LeMans, Daytona, Bonneville, and Sebring are
> jwisniew@@ford.com | just races, won by people driving Ford cars!
>
>
>
Jonathan Epstein Jonathan_Epstein@@nih.gov
Unit on Biologic Computation (301)402-4563
Office of the Scientific Director c/o Bldg 31, Room 2A52
Nat. Institute of Child Health and Development 31 Center Drive
National Institutes of Health 20892
![]() |