ΙΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝ» Ί Ί Ί INTEL MATH COPROCESSOR SOFTWARE TROUBLESHOOTING Ί Ί Ί ΘΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΌ LOTUS 1-2-3 AND SYMPHONY @NA FUNCTION The @NA function in 1-2-3 version 2.01 and early versions of Symphony will not work correctly on a system with a 387DX installed. The problem has been corrected in 1-2-3 version 2.2. The phone number for Lotus Development is (617) 253-9150. GENERAL SOFTWARE USE OF MATH COPROCESSORS Not all programs make good use of MCPs. Scientific and CAD/Graphics programs use MCPs intensively for screen graphics and floating point calculations. Spreadsheets use coprocessors for exponents, internal rate of return, net present value, and certain arithmetic calculations. Basic math functions (add, subtract, multiply, and divide) are sped up minimally or not at all, depending on the program. -E+999 WHEN ENTERING NUMBERS You'll see this error if the math coprocessor switch on the system board is set incorrectly. 386 XENIX/UNIX AND THE 387DX When 386(TM) versions of Xenix or Unix run on a computer containing an Intel 387(TM) DX math coprocessor, the computer can lockup intermittently. Read FaxBack document #3054 for more detailed information on this problem. CADKEY CadKey released a 32-bit version of their 3D CAD software. This triggers the errata 21 problem when the customer asks the program to redraw images with arcs or text. CadKey has an article that talks about the Errata 21 problem as though it's a problem in the MCP instead of in the processor. So their tech support is telling customers to get their MCP replaced. Instead, the customer needs to call the manufacturer of the computer to deal with the 386 processor. CadKey tech support Voice ph: 203-647-0220 FAX: 203-649-8231 ΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝ February 25, 1991 Intel FaxBack # 3060 Page 1 ΙΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝ» Ί Ί Ί INTEL MATH COPROCESSOR SOFTWARE TROUBLESHOOTING Ί Ί Ί ΘΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝ (continued) ΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΌ LOCKUPS WITH CERTAIN CAD PROGRAMS An incorrect computer configuration can cause certain programs to lockup while others work fine. You must make sure that the computer's SETUP and/or switch settings correctly indicate the presence of a Math Coprocessor. (Field report) A report indicated that in a computer with the Intel Inboard 386/AT and 387DX Math Coprocessor, if Math Coprocessor switch was set wrong, Autocad ran just fine but Digicad didn't. LOTUS 1-2-3 MATH COPROCESSOR TEST Here's a quick and easy method for testing your math coprocessor using Lotus 1-2-3: 1. Enter this formula in cell A1: @exp(@rand) 2. Copy the formula to a large range with this command: /CA1..Z400 3. Press F9 and time how long it takes the spreadsheet to recalculate. In an 8 MHz computer, it should take about 1 minute without a math coprocessor and about 10 seconds with a math coprocessor. In a 4.77 MHz computer, it should take about 3 minutes to recalculate without a math coprocessor and about 27 seconds with a coprocessor. The difference in time is noticeable even without a stopwatch. 1-2-3: GARBAGE IN CELLS A defective math coprocessor can cause unpredictable values in number cells. Test your math coprocessor with Intel's CHKCOP program. Download the program with your modem from (503) 645-6275 or contact Intel Customer Support ΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝ February 25, 1991 Intel FaxBack # 3060 Page 2 ΙΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝ» Ί Ί Ί INTEL MATH COPROCESSOR SOFTWARE TROUBLESHOOTING Ί Ί Ί ΘΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝ (continued) ΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΌ SET 87 = NO (OR OFF) Some programs will respond to the command SET 87=NO. This DOS command creates an environment variable. It does not directly disable the math coprocessor. If a program checks the variable "87" and find it set to "NO" (or "OFF"), the program will use floating-point emulation subroutines instead of the math coprocessor. Not all programs will check for this variable so its usefulness is limited. Ask your software manufacturer for information on how their program responds (or doesn't respond) to this command. ΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝΝ February 25, 1991 Intel FaxBack # 3060 Page 3