albert@endor.uucp (David Albert) (12/25/90)
I'm running MS-DOS 4.01 on a 25Mhz AMI '386. When I run the Landmark cpu-speed test (the one with the bar that graphically compares the speed of the processor to the speed of a 4.77Mhz PC and a 6Mhz AT) the value I get depends on how many TSRs are loaded -- but not monotonically. The value I get is either 25 (approx) or 17 (approx) and changes from one to the other as I load TSRs. Some TSRs cause the change, some don't. Any idea what's going on? ----------------------------------------------__------------------------------ David Albert | / ) / UUCP: ...!harvard!albert | / / __. , ___o __/ INTERNET: albert@harvard.edu |/__/__(_/|__\/ <___(_/_
schuster@cup.portal.com (Michael Alan Schuster) (12/25/90)
>I'm running MS-DOS 4.01 on a 25Mhz AMI '386. When I run the Landmark >cpu-speed test (the one with the bar that graphically compares the >speed of the processor to the speed of a 4.77Mhz PC and a 6Mhz AT) >the value I get depends on how many TSRs are loaded -- but not >monotonically. The value I get is either 25 (approx) or 17 (approx) >and changes from one to the other as I load TSRs. Some TSRs cause >the change, some don't. Any idea what's going on? This sounds like a bug in the Landmark code that was fixed in version 1.1x; systems using page-interleave memory would execute the benchmark at different speeds depending on WHEREin physical memory the code was actually loaded. (i.e. on a bank boundary or not). Welcome to the world of benchmarks. Is this what you bought your computer to run? :-)
silver@xrtll.uucp (Hi Ho Silver) (12/26/90)
It probably depends on what the TSRs do. I've found that Landmark doesn't notice any difference if I put an empty loop into the clock tick to slow the machine down; it seems to slow down its timer as well. What TSRs are you using, and which ones cause a slowdown? -- __ __ _ | ...!nexus.yorku.edu!xrtll!silver | always (__ | | | | |_ |_) >----------------------------------< searching __) | |_ \/ |__ | \ | if you don't like my posts, type | for _____________________/ find / -print|xargs cat|compress | SNTF
dave@westmark.WESTMARK.COM (Dave Levenson) (12/30/90)
In article <5157@husc6.harvard.edu>, albert@endor.uucp (David Albert) writes: > I'm running MS-DOS 4.01 on a 25Mhz AMI '386. When I run the Landmark > cpu-speed test (the one with the bar that graphically compares the > speed of the processor to the speed of a 4.77Mhz PC and a 6Mhz AT) > the value I get depends on how many TSRs are loaded -- but not > monotonically. The value I get is either 25 (approx) or 17 (approx) > and changes from one to the other as I load TSRs. Some TSRs cause > the change, some don't. Any idea what's going on? I would guess that one or more of your TSRs is hooking the clock interrupt, and stealing some of your CPU time on every tick. I think this interrupt occurs about 18,000 times per second, so even if the TSR is only looking at its work queue and deciding to do nothing, it does a lot of that. I'm not familiar with the internals of the Landmark test, but it may use the clock-tick itself, and there may be some contention for it. -- Dave Levenson Internet: dave@westmark.com Westmark, Inc. UUCP: {uunet | rutgers | att}!westmark!dave Warren, NJ, USA AT&T Mail: !westmark!dave [The Man in the Mooney] Voice: 908 647 0900 Fax: 908 647 6857