Relative to:
All Versions
Recommended System Requirements

Updated 7/30/03
Definitions  "Minimum Recommended System Requirements" vs. "Minimum System Requirements"

The Minimum Recommended System Requirements are the minimum system requirements required in order for Speech Recognition software to function with the accuracy and speed necessary for productive computer operation consistent with a business environment.  Recommended System Requirements are identified by people working in the field and are based on their experience.  Because of the trial and error nature of determining Recommended System Requirements, different sources are likely to have different Recommended System Requirements.  Occasionally manufactures will include Minimum Recommended System Requirements or (Recommended System Requirements) on the box in addition to the Minimum System Requirements.  Increased system resources will generally result in faster performance and higher accuracy rates.

The Minimum System Requirements are the system requirements necessary to run the software without crashing.  These are the requirements that are generally listed on the box.  Computers that just meet the Minimum System Requirements are likely to run slower and to have lower accuracy than the  software is capable providing.


Minimum Recommended
Requirements
CPU and RAM
NaturallySpeaking Version 7.0
  • CPU Pentium 800 mhz or faster, full Pentium, not a Celeron (1)(2)
  • 384 MB RAM,  512 MB for Windows NT, XP, or Power Users (3)
NaturallySpeaking Versions 6.0/6.1
  • Pentium 500 mhz or faster, full Pentium, not a Celeron (1)(2)
  • 256 MB RAM,  384-512 MB for Windows NT or Power Users (3)
NaturallySpeaking Versions 3-5
  • Pentium 250 mhz or faster, full Pentium, not a Celeron (1)(2)
  • 256 MB RAM,  384-512 MB for Windows NT or Power Users (3)

All Versions
  • A quality (not expensive) sound card that provides clear recordings at 11,025 Hz, 16 bit, mono. (4)
  • 500 MB free Hard Disk space prior to installing NaturallySpeaking.
  • CD ROM (required for installation only)
  • Quality microphone.  The microphones that come in the box may not be of the highest mechanical quality and comfort but should provide good speech recognition results.
  • Latest versions of Windows 98, ME, NT 4, 2000, or XP


Notes
  1. Early Celeron processors had little or no L2 cache.  L2 Cache is used by programs handling large amounts of information such as NaturallySpeaking.  Small or no L2 cache can result in lower performance and accuracy.  Some new Celeron processors do have L2 cache and may work fine.  The processor should have at least 128 KB (not MB) of L2 Cache.  Very fast processors, 1+ Ghz, can sometimes compensate for smaller amounts of L2 Cache.
  2. NaturallySpeaking will run on other processors such as AMD, etc.  However, software manufacturers tend to design software for Pentium processors and then consider fixing any problems encountered with other processors.  Consequently, the possibility of problems is higher with other processors. On the other hand, other processors can provide substantial saving on the cost of a computer and many users are very happy with them. In the end, its a risk vs. gain choice.
  3. NaturallySpeaking and one other program, such as MS Word, should work fine with 256 MB RAM.  HOWEVER, an inadequate amount of RAM will cause lower accuracy and drastically slower performance.  Unfortunately, it is not always easy to determine how much RAM a computer is using and where it went. A professional technician may be required to identify and resolve RAM issues. Increased RAM requirements can be caused by:
    • New Operating Systems such as Windows XP
    • Running multiple programs at the same time such as MS Word, Excel, Internet Explorer and others along with the Printer Drivers, Network Drivers, Anti virus Software, etc.
    • Running very large programs and/or Graphics programs such as AutoCAD.
    • Some computers have unknown background programs or utilities that take up LARGE amounts of RAM.
  4. Sound cards built into the Motherboard are getting much better but have a history of providing recordings with digital noise and hiss that provide poor recording quality and therefor poor speech recognition performance.
©1995 - 2003 EXAQ