KDB Systems

With the pub­lic beta of Win­dows 7 in full swing many peo­ple are talk­ing about per­for­mance and com­par­ing dif­fer­ent ver­sions of Win­dows. I see many posts in forums and on news­groups exclaim­ing Win­dows 7 boots x sec­onds faster than Vista. They care­fully mea­sure how long XP, Vista, and Win­dows 7 take to boot or shut­down. Oth­ers mea­sure how much RAM each OS uses when at idle. Some peo­ple run bench­mark soft­ware com­par­ing var­i­ous OS’s. There are web sites ded­i­cated to per­for­mance with tips on which ser­vices and sched­uled tasks can be dis­abled to improve performance.

Most users are more con­cerned with per­ceived per­for­mance rather than actual per­for­mance. If I click on some­thing is there a pause before some­thing hap­pens? If that pause is longer than x (I don’t know what x is but I sus­pect it’s less than a sec­ond) the com­puter or appli­ca­tion is per­ceived as slow. If it’s faster than x then the com­puter or appli­ca­tion is per­ceived as fast. There isn’t really any in between. There is no per­cep­tion of medium per­for­mance for most peo­ple. It’s either accept­able or too slow. Most cur­rent oper­at­ing sys­tems take all this into account and are opti­mized to give a good user expe­ri­ence. Some­times this per­ceived bet­ter per­for­mance comes at the expense of actual per­for­mance. The oper­at­ing sys­tem is doing things in the back­ground like index­ing files, opti­miz­ing the file sys­tem, pre-caching disk sec­tors, and more. These back­ground tasks may cause bench­marks to run slower. Some peo­ple jump on this and dis­able these back­ground tasks then proudly post bench­marks prov­ing how much faster their com­puter is. The prob­lem is that dis­abling these back­ground tasks quite often makes the com­puter less opti­mized for the user expe­ri­ence. Pro­grams may actu­ally run slightly faster but load­ing the pro­gram or loading/saving files from within the pro­gram take longer. Find­ing the email you sent to Joe Smith about next week’s hockey game takes impos­si­bly long as you have to man­u­ally open each email. Over time Win­dows slows down because the disk is fragmented.

When tun­ing or mea­sur­ing com­puter per­for­mance you have to take many things into con­sid­er­a­tion. It’s very sim­i­lar to a car. Most of us don’t want to drive a souped up hotrod that’s tem­pera­men­tal and needs con­stant atten­tion. Most us want a car that starts up when we turn the key. The heat or the air con­di­tion­ing comes on quickly not sev­eral miles down the road. We want power locks, win­dows and seats. We want com­fort. It’s the same with com­put­ers. There are enthu­si­asts who enjoy eak­ing out every mil­lisec­ond of per­for­mance and don’t care about the com­forts or ease of use. Unfor­tu­nately many peo­ple lis­ten to their advice and think that if they apply the same tricks their com­puter will be faster. It will, but the catch-22 is that their day to day com­put­ing may actu­ally seem slower.