CPU 24-Hour Burn-In Test Program v.1.22 by Peregrine Graphics Systems ======================================= (c) 1991 William Ip .September.26.1991. Introduction ============ CPU24 is a "industrial grade" burn-in test program for use in testing the ability of system-boards to withstand heavier-than-normal operations in order to locate any possible manufacturing/engineering defects. The testing involves all aspects of system-board resources, from CPU, FPU, BIOS ROM, system cache, RAM, and even video. The program has been named CPU24.EX! so that you must read this file before using CPU24. As a result, you now have the obligation to read the following instructions, warning, notice and disclaimer. To setup CPU24 for its executable, type INSTALL and the Enter/Return key. Then, type CPU24 to run. NOTE: This text file, CPU24.TXT MUST be distributed with the CPU24 program. Proper implementation of month change has been done implemented except for the last day of February on a leap year. As a result, using CPU24 on February 29th may yield and improper end time. This limitation will be addressed way before the next leap year rolls around and is probably not a cause for concern. Registered users will certainly be the first to know of the newest revision (more on registration later). Future revisions ================ - more involvment of video sub-system in testing - implement algorithm to detect leap year for February 29th - implementation of floppy/hard drive testing, either integrated with or as a companion module, to CPU24 - user-selectable time frame for testing - implement extended memory fills and flushes - implementation with C++ for higher efficiency program coding Revision History ================ 1.22 - Released: September.28.1991 - minor cosmetic changes and maitenance release 1.21 - Released: July.22.1991 - reorganized display screen - minor code changes to streamline execution - [F10] key offers on-line information screen 1.20 - Released: July.01.1991 - will now change last day of month properly (except February 29th on leap years) - implemented a standardized floating point equation based on x^x^x where x is a floating point value between 1 and 7 - enhanced, polished and streamlined internal coding to minimize program code bottlenecks. - increased memory block fills and flushes from 32K to 256K 1.10 - Released: June.21.1991 - bug fix: v.1.00a tended to stop after 1 second before 10 hours due to a misplaced data type - implemented full IEEE standard double precision floating point calculations - removed introductory text from executable - replaced 72-hour executable version with a batch file - implemented filling/flushing of 32K memory blocks 1.00a - Released: June.01.1991 - implementation of 24 hour time frame - addition of a 72-hour executable version for manufacturing use - minor equation alteration to prevent illegal values 1.00 - Released: February.14.1991 - implementation of full double precision floating point calculation - first full-scale distribution 0.01 - Released: December.01.1990 - beta version - minor bug fix 0.00 - Released: November.01.1990 - beta version - initial skeleton framework What CPU24 Is All About ======================= This program's purpose is simple and straight-forward: to burn-in system- boards and their components. This program works on the premise that most computer system problems arise as a result of: 1. poor engineering design, or 2. manufacturing defects. A number of problems may arise due to electrical problems. These problems may come about as a result of thunderstorms, brownouts, blackouts, surges, improper ground lines in the outlet or just plain electrical/mechanical wear and tear. These latter are not considered manufacturing defects or engineering problems since they occur as a fact of life apart from your computer. It is important to differentiate these problems from the first two since most warranties are only on the defectiveness of the manufacturing process and not on the result of environmental exposure. CPU24 performs tests in passes, each pass calculating 8 IEEE-compliant floating point operations of a complex equation, with an additional 8 additional calculations, 4 to calculate timing, 2 to handle counters, and 2 to write the first 8 calculations into a 256K memory block. Each 256K memory block is first filled, then its data is flushed out and refilled again. As a result, this test will use all system-board resources: CPU, FPU, RAM, BIOS ROM, system cache, chipset modules and even video. This type of intensive testing will bring the system-board components to maximum operating temperature and subject them to higher-than-normal operational use. Any engineering and/or manufacturing flaws will certainly show up during this exercise, usually by hanging the system or spontaneously resetting the system. Some manufacturers install CPUs which are improperly rated (ie. 286-10 CPUs boosted to 12MHz, or 386-25 CPUs to 33MHz, by means of a faster crystal). This test will almost always reveal the shaky reliability of such boards, causing them to hang/reboot about 60%-80% of the time. As a note, even if you have such a 'souped-up' system within the lucky 20%, I don't find it morally or ethically acceptable for retailers/manufacturers to sell such units. Hoping your 'souped-up' system-board falls within 20% is not a chance I'd risk. From personal experience, I've notice two types of such 'souped-up' boards. The first are the standard boosts with faster crystals. The boards usually have massive heat sinks on the CPUs and sometimes even on chipsets to try cooling these components down. The second type are called 'Factory Pre- Sorted' CPUs. These essentially faster chips which fail the CPU manufacturer's (usually Intel, Harris, AMD or Siemens) stringent quality control tests but are still much better than the normal properly rated CPUs. Thus, in the vast majority of standard usage, these 'Factory Pre-Sorted' CPUs tend to not give many problems. The most approriate time to use this program would be right after you bring your brand-new machine home or to the office from the store, even before you take the time to pack away the boxes. Prepare a batch file to run CPU24 for 3 days (72-hours). One such batch file is included, called CPU72.BAT. Sure, you won't have use of your brand new system for 3 days, but I'm sure you'll agree that 3 days now would be better than 1 week in the shop 6 months down the line! If a system passes without a hitch, you can be darn sure it will keep running. You should also examine if your case and power-supply provides adequate air ventilation/circulation to cool your system. Overheating may cause problems in the future. Again, it is better to catch the problems now, rather than later. If you do find that your system hangs or reboots, especially if you run the test 3 or 4 times, you can be sure there is something wrong with the system. This test has been tested on a wide range of systems and on a system-board with good components, you can run this test for weeks on end without causing any crash. After your initial test, you may give your system the occasional run-through whenever you have the time to spare. However, when the final week of your warranty is approaching, run the test another 72-hours, just to make sure you can get any potential problems fixed before warranty expires. Any system that runs for a year or more will continue to run. This program can also be used to benchmark various configurations. CPU24 depends on the configuration of the system-board in general in completing each pass it cycles through. Naturally, systems-boards with FPUs and system caches will certainly increase their ability to do more calculations in a time frame. Please note that CPU24 is not designed for true benchmarking as no provisions have been implemented for actually assessing the fine differences between CPU types. As a result, '386sx-16 and 286-16 CPUs may sometimes offer the similar results, especially if no math co-processor is installed. CPU24 is designed to test the ability of a system-board to do heavy computational work over a longer-than-average time period. Most people interpret benchmarks improperly. Some people (retailers, manufacturers, and buyers alike) throw around Norton SI or Landmark CPU speed ratings left, right and center. One very classic case is the use of the Landmark CPU Speed Test on the Intel 486 class system-boards. Some retailers enjoy quoting seemingly spectacular speed ratings with version 0.99 of the Landmark CPU Speed Test, with figures going up to 155MHz on 486-33MHz systems. I'd like to inform everyone now that the version 0.99 Landmark does not have any provisions for CPUs as high as a 386-33, let alone the 486-33. The proper version of Landmark is version 2.00 which gives most 486-33 systems a speed of 115MHz. In fact, when you run version 0.99 on a 386-33, the speed bar wraps around the screen several times. On a 486-33, version 0.99 hangs, since its algorithms to calculate a speed rating is simply blown away by the 486-33's response. The 155MHz reported speed is simply the last figure version 0.99 prints out before it dies. The same applies to Norton SI and a host of other such benchmark programs. Versions are very important and you must make sure a particular benchmark offers the correct results. A benchmark reported by Norton SI version 3.0 is not the same as a Norton SI version 4.5 or 5.0. And of course, a Norton SI speed of 33MHz is not the same as a Landmark 0.99 33MHz or a Power Meter 33MHz. The same holds true for CPU24. A benchmark must be compared against two systems, each with only one factor different, whether it be CPU speed or video display type. When using CPU24 to benchmark CPUs, you should have two machines with the same configuration exactly (hard drive type, video type, RAM) except for the system-board and CPU itself. WARNING ======= The testing performed will run your system-board to its maximum capacity. Although it has yet to occur, the high temperatures generated by your system- board during the tests may cause damage to improperly designed components. SYSTEM-BOARD COMPONENTS WHICH ARE OF ANY HALF-DECENT QUALITY AND PROPERLY DESIGNED FOR THEIR RATED OPERATION WILL NOT BE DAMAGED FROM THIS TEST. Many components will generate extreme heat which may cause severe burns if in contact with bare skin. Do not touch components during testing. (I found out the hard way and nursed a number of blisters for some days.) You are urged to read the intructions and disclaimer in the introduction to CPU24 before running the test program. REGISTRATION ============ If you like this program and find it of much use in maintaining the working operation of your computer, or find that it helped you locate a fault in your system, please send $10 (cash/money-order/no cheques) to register your copy. In addition to acknowledging the usefulness of this program, the registration fee entitles you to future versions and/or burn-in softwares produced by Peregrine Graphics Systems. If you intend to use this program to burn-in systems which you will resell, you MUST register. Comments and suggestions are highly desired. Send registration and/or comments to: William Ip Box 428, Station W Toronto, Canada, M6M 5C1 I can also be reached on Compuserve at 70670,737, or Canada Remote Systems 416-798-7730 (2400 baud) or 416-798-7733 (14,400 USR dual). Registered versions include a serial number displayed during the testing. IMPORTANT NOTICE & DISCLAIMER ============================= Peregrine Graphics Systems, as the manufacturer/producer of this program, offers no guarantees, explicit or implied, with regards to the fitness or usefulness of this program except for the following: this software, CPU24, has been written and fully tested on systems manufactured by: Abax Communications Group 2091 Dufferin St. Toronto, Canada, M6E 3R3 (416) 658-3989, (416) 658-3992 FAX CPU24 was designed primarily for burn-in testing and a method of quality control of Abax computer systems. Peregrine Graphics Systems shall not be responsible for any damages caused by the use, or misuse, of this software. Abax Communications Group has no involvment with Peregrine Graphics Systems or any software produced by the latter except for providing the time, place and equipment. As a result, Abax Communications Group shall not be subject to any judicial claims made against Peregrine Graphics Systems. All software program code is proprietary to Peregrine Graphics Systems, with the exception of the Microsoft QuickBASIC 4.5 Run-Time library which is registered to Microsoft Corp. No modifications of this software by any means is allowed. Distribution of CPU24 must be with this file (CPU24.TXT), the install file (INSTALL.EXE), the data file (CPU24.DTA) and the 72-hour batch files (CPU72.BAT). Distribution CANNOT include the CPU24 with the .EXE extension. With these provisions, execution of the CPU24 program indicates that you are bound by the rules listed above governing use of this software. Thanks to: ========== Microsoft Corp. for QuickBASIC 4.5 compiler, which beats the heck out of GWBASIC, and easily out-guns any C environment. Maybe C code runs faster, but I don't think CPU24 requires speed. I'll probably port it to C++ as a programming exercise/challenge for myself (Why'd I use QuickBASIC? Easy to use, quick to write with. Took me 2 hours to code the entire basic skeleton of CPU24 v.0.00) Abax Communications Group for use of their facilities and equipment (never mind that I'm the Operations Manager there...) Rebekka for keeping me company...