Comparison on Performance ========================= Microsoft Corporation on October 10th published performance numbers comparing OS/2 Warp Beta II to a non-public Windows 95. As we understand, the performance numbers were pulled from many bulletin boards due to complaints pointing out glaring inaccuracies. The The document is still available on the Microsoft internet server. The performance numbers published by Microsoft are inaccurate and unreliable for the following reasons: 1. The percentage numbers were 100% too high; for instance, if Microsoft asserted that Windows 95 was 20% better than OS/2 Warp the Microsoft document calls Windows 95 120% faster. This error occurred 48 times in the Microsoft document. Even though this has been pointed out for weeks Microsoft still has not corrected it. 2. Our tests conclude that the Windows for Workgroups (WFW) and Windows 95 machines must have had a 32-bit Western Digital controller (or equivalent) for disk access, which Microsoft's 32-bit VFAT driver takes advantage of. This configuration is not the default or the representative config in the market. 3. Microsoft used OS/2 Warp Beta II which was tuned for 4MB as were Beta I and the released product. 4. OS/2 Warp uses different installation parameters when installing on 4MB machines. Microsoft we suspect installed OS/2 Warp on a machine "with greater than 4MB and then stripped memory to get down to a 4MB configuration." 5. The disk cache size for WFW was configured 4 times larger than the disk cache size for OS/2 Warp giving WFW a totally unfair advantage. We ran performance tests comparing OS/2 Warp's performance to Windows 3.1 and Windows for Workgroups. Since Windows 95 is still under Non Disclosure Agreement (NDA) we were unable to do any performance testing but we would be happy do so if Microsoft agrees to provide us with a copy. Our performance test found that in many cases OS/2 Warp does outperform WFW 3.11 and Windows 3.11 in 4Mb of memory. We used the Generally Available versions of OS/2 Warp, Windows for Workgroups 3.11 and Windows 3.11 using default configurations with the disk cache was set to 256Kb for all systems. All of the tests were run on the following machine: Machine Configuration ===================== Machine: IBM PS/Value Point Processor: 486DX CoProcessor: Installed Speed: 25Mhz Hard Disk: Maxtor 244 Mb IDE model 7245A System Memory: 4Mb External Cache: 256Kb Internal Cache: 8Kb Video Memory: 1024Kb Flash EEPROM Revision Level: L9ET30AU Several user scenarios designed to measure OS/2 Warp's performance against Windows 3.1 in a multiple application environment yield better performance for OS/2 Warp in 4Mb of memory: - Lotus 123 + MS Money + Winclock running concurrently -- OS/2 Warp is 7% faster than Windows 3.1 - Lotus 123 + Quicken + Winclock running concurrently -- OS/2 Warp is 11% faster than Windows 3.1 - Lotus 123 + Amipro + Winclock running concurrently -- OS/2 Warp is 7% faster than Windows 3.1 Printing in OS/2 Warp is significantly faster than in Windows 3.1 - Printing using Wordperfect for Windows 3.1 is approximately 40% faster under OS/2 Warp vs. Win 3.1 in a multitasking environment using the HP560C printer. (multitasking accomplished by downloading a file from CompuServe while printing.) - On an HP560C, printing using Wordperfect for Windows is 30% faster in OS/2 Warp. When comparing the products for read and write times you find that OS/2 Warp is significantly faster. We tested various record sizes in random and sequential format attached below are some of the results in Kilobytes per Second (Kbs). When reading 200 Byte record in a random read and a cache of 256K File Size 256K 1M 2M 4M 8M 16M 32M ====================================================== WFW 148 9 7 7 6 5 4 OS/2 Warp 291 12 11 10 8 8 8 ------------------------------------------------------ % Warp Faster 97% 33% 57% 43% 33% 60% 100% _______________________________________________________________________ When reading 2K Byte record in a random read and a cache of 256K File Size 256K 1M 2M 4M 8M 16M 32M ====================================================== WFW 1566 93 78 70 62 53 40 OS/2 Warp 2319 115 102 94 93 88 79 ------------------------------------------------------ % Warp Faster 48% 24% 31% 34% 50% 66% 97% _______________________________________________________________________ When reading a 512 Byte record in a sequential read and a cache of 256K File Size 256K 1M 2M 4M 8M 16M 32M ====================================================== WFW 563 641 654 643 638 631 622 OS/2 Warp 617 682 682 707 711 697 688 ------------------------------------------------------ % Warp Faster 10% 6% 4% 10% 11% 10% 11% _______________________________________________________________________ When reading a 4K Byte record in a sequential read and a cache of 256K File Size 256K 1M 2M 4M 8M 16M 32M ====================================================== WFW 592 710 714 714 707 693 691 OS/2 Warp 2072 862 878 492 891 883 888 ------------------------------------------------------ % Warp Faster 250% 21% 23% -45% 26% 27% 29% _______________________________________________________________________ When writing a 200 Byte record in a random write and a cache of 256K File Size 256K 1M 2M 4M 8M 16M 32M ====================================================== WFW 76 6 5 4 4 3 3 OS/2 Warp 184 8 8 7 6 5 6 ------------------------------------------------------ % Warp Faster 142% 33% 60% 75% 50% 67% 100% _______________________________________________________________________ When writing a 512 Byte record in a random write and a cache of 256K File Size 256K 1M 2M 4M 8M 16M 32M ====================================================== WFW 287 32 29 27 24 18 13 OS/2 Warp 306 39 37 35 33 33 33 ------------------------------------------------------ % Warp Faster 7% 22% 28% 30% 38% 83% 154% _______________________________________________________________________ When writing a 200 Byte record in a sequential write and a cache of 256K File Size 256K 1M 2M 4M 8M 16M 32M ====================================================== WFW 221 149 147 148 150 148 148 OS/2 Warp 314 314 221 314 310 302 301 ------------------------------------------------------ % Warp Faster 42% 111% 50% 112% 107% 104% 103% _______________________________________________________________________ When writing a 512 Byte record in a sequential write and a cache of 256K File Size 256K 1M 2M 4M 8M 16M 32M ====================================================== WFW 353 361 363 363 361 357 355 OS/2 Warp 517 620 623 373 622 623 359 ------------------------------------------------------ % Warp Faster 46% 72% 72% 3% 72% 75% 1% _______________________________________________________________________ To truly compare the performance of Windows for Workgroups to Warp in 4 Meg we suggest that you compare the time it takes to perform an everyday task on a Windows system using existing applications and a system that has been Warped using the Bonus Pak applications. Scenario: Fax a memo to a business associate and get back to doing something else. Warp Windows or Windows for Workgroups 3.11 ==== ====================================== 1. Open Address Book 1. Start Application 2. Drag business cards onto 2. Select File from pull down document menu 3. Drag document to Fax machine 3. Select Open 4. Click on OK for cover sheet 4. Select Document 5. Do other work 5. Click on OK 6. Wait for the document to open 7. Select File from pull down menu 8. Select Printer Setup 9. Select the Fax printer 10. Click on OK 11. Select File from pull down menu 12. Select Print 13. Click on OK 14. Key in phone number and cover sheet information 15. Click on OK 16. Wait until faxing is over 17. Close application 18. Do other work We have presented performance data where OS/2 Warp performs better than WFW and Windows 3.11. We understand that it is possible to create other operating environment to achieve performance results that are desired. The operating system you choose really does make a difference. We hope that you choose the reliable, stable, proven operating system that protects your existing investments and gives you the ability to exploit future technology.