Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!samsung!umich!sharkey!cfctech!teemc!ka3ovk!raysnec!shwake From: shwake@raysnec.UUCP (Ray Shwake) Newsgroups: comp.sys.ibm.pc Subject: Re: Performance of 386SX Keywords: performance, cache Message-ID: <48@raysnec.UUCP> Date: 5 Jun 90 16:02:28 GMT References: Reply-To: shwake@raysnec.UUCP (Ray Shwake) Distribution: na Organization: IRS - ACI Project Office Lines: 15 In article carpio@questor.wimsey.bc.ca (antonio carpio) writes: > MIPS 1.91 > Coretest - Data Transfer Rate 460 Kb/sec; Ave. seek 18.9; Track-Track Seek >Seek 6.2; Overall performance 5.654. Speed 18.8 Mz. Your results are typical for the class of machine under question, but I wouldn't take the Data Transfer Rate too seriously. Since you don't specify the block size used by Coretest, I assume you used the default (typically 32K or 64K); most real applications use far smaller block sizes (2K or less) with much less throughput. Coretest also reads only a single data stream many times, thus compounding the error. As far as increasing performance, adding software disk cache can often increase performance in real terms. I've used the smartdrv cache included with Windows to good effect, but many flavors of disk cache exist.