jep@fantasci.UUCP (Joseph E Poplawski) (01/05/89)
How bad of a speed difference would my UNIX V.3 for the 80386 system incur if I added 120ns chips instead of the 100ns chips it already has 2 meg of? The system is a personal system with no more than 3-5 users on at the extreme most. The reason I am considering the slower chip is mainly the price. Any advice or suggestions would be greatly appreciated. -Jo ------------------------------------------------------------------------------- | Joseph E Poplawski (Jo) US Mail: 1621 Jackson Street | | Cinnaminson NJ 08077 | | UUCP:..!rutgers!rochester!moscom!telesci!fantasci!jep | | ..!princeton!telesci!fantasci!jep | | ..!pyrnj!telesci!fantasci!jep Phone: +1 609 786-8099 home | -------------------------------------------------------------------------------
terry@eecea.eece.ksu.edu (Terry Hull) (01/06/89)
In article <377@fantasci.UUCP> jep@fantasci.UUCP (Joseph E Poplawski) writes: >How bad of a speed difference would my UNIX V.3 for the 80386 system incur if >I added 120ns chips instead of the 100ns chips it already has 2 meg of? The >system is a personal system with no more than 3-5 users on at the >extreme most. First, system load does not have much to do with memory speed requirements. The processor runs full speed all the time whether it is waiting for keyboard input or moving a 1MB array in memmory. Memory is not a good place to economize when you are using UNIX. For some reason, UNIX tends to bring out memory errors in a system faster than other OSs like MSDOS. You do not mention what machine you have, but in general: 16 MHZ 100ns 20 80ns 25 60ns I got lucky when I purchased my Inboard/386 AT, it is a 16MHZ 386, but it uses a 64K static cache and will run with 120ns chips. Remember, if the machine would run reliably with the slower chips, the manufacturer would have used them to save money. This might not be true if a manufacturer got a particularly good deal on faster chips though. You will know if it is not working when your machine dies with panic messages. -- Terry Hull Department of Electrical and Computer Engineering Kansas State University INTERNET: terry@eecea.eece.ksu.edu Manhattan, KS 66502 UUCP: rutgers!ksuvax1!eecea!terry
debra@alice.UUCP (Paul De Bra) (01/07/89)
In article <377@fantasci.UUCP> jep@fantasci.UUCP (Joseph E Poplawski) writes: >How bad of a speed difference would my UNIX V.3 for the 80386 system incur if >I added 120ns chips instead of the 100ns chips it already has 2 meg of? The >system is a personal system with no more than 3-5 users on at the extreme most. > >The reason I am considering the slower chip is mainly the price. > There is no magic in the PC and AT compatibles (including 80386's) to detect what speed the memory can handle. There are 2 possibilities: either your system doesn't really need 100ns chips and will run just as fast with 120ns chips, or else the system will not run at all with 120ns chips (it will give parity errors at least). Considering the price of memory chips it is very unlikely that you got the machine with 100ns chips when it only needs 120ns. The only way to make your system work with slower chips is to lower the clock-frequency (like from 20Mhz to 16Mhz). What ram-chips your system needs is very motherboard-dependent. Some boards use very fast rams (60ns) and no cache memory, others use slow ram (up to 120ns) and compensate with cache memory. In any case the hardware expects a minimal speed of ram which is really necessary at the highest clock rate. Paul. -- ------------------------------------------------------ |debra@research.att.com | uunet!research!debra | ------------------------------------------------------
sl@van-bc.UUCP (pri=-10 Stuart Lynne) (01/07/89)
In article <8687@alice.UUCP> debra@alice.UUCP () writes: >In article <377@fantasci.UUCP> jep@fantasci.UUCP (Joseph E Poplawski) writes: >>How bad of a speed difference would my UNIX V.3 for the 80386 system incur if >>I added 120ns chips instead of the 100ns chips it already has 2 meg of? The >>system is a personal system with no more than 3-5 users on at the extreme most. >> >>The reason I am considering the slower chip is mainly the price. >> > >There is no magic in the PC and AT compatibles (including 80386's) to detect >what speed the memory can handle. There are 2 possibilities: either your >system doesn't really need 100ns chips and will run just as fast with 120ns >chips, or else the system will not run at all with 120ns chips (it will give >parity errors at least). Considering the price of memory chips it is very >unlikely that you got the machine with 100ns chips when it only needs 120ns. > >The only way to make your system work with slower chips is to lower the >clock-frequency (like from 20Mhz to 16Mhz). > >What ram-chips your system needs is very motherboard-dependent. Some boards >use very fast rams (60ns) and no cache memory, others use slow ram (up to >120ns) and compensate with cache memory. In any case the hardware expects >a minimal speed of ram which is really necessary at the highest clock rate. Actually the speed of your RAM is dependant on the clock-rate *AND* the number of wait states. Generally speaking you can use slower memory chips with high speed machines by adding wait states. Depending on your machine you will have one or more of the following possibilities: - the new chips work great (try and run a good memory test though) - they don't work If they don't work: - reduce the clock rate - add global wait states - add wait states for the new memory only The last is best, but probably not supported on too many 386 systems. The actual impact lowering the clock-rate or adding wait states will have is dependant on the system: - does it have four way interleave - does it have a cache The bottom line, is RTFM. Find out what options you have in terms of setting your system up, can you: - modify clock-rate - add wait states - add wait states for specific memory only BTW these are issues that are best dealt with *before* you buy the system, unless you are never going to be expanding the system after you buy it (or are willing to return to the vendor for all upgrades). -- Stuart.Lynne@wimsey.bc.ca {ubc-cs,uunet}!van-bc!sl Vancouver,BC,604-937-7532
cww@ndmath.UUCP (Clarence W. Wilkerson) (01/07/89)
That might allow slower RAM to be used.
gwyn@smoke.BRL.MIL (Doug Gwyn ) (01/07/89)
In article <377@fantasci.UUCP> jep@fantasci.UUCP (Joseph E Poplawski) writes: >How bad of a speed difference would my UNIX V.3 for the 80386 system incur if >I added 120ns chips instead of the 100ns chips it already has 2 meg of? It will either work at precisely the same speed or not work at all.
tom@dvnspc1.UUCP (Tom Albrecht) (01/11/89)
In article <377@fantasci.UUCP>, jep@fantasci.UUCP (Joseph E Poplawski) writes: > How bad of a speed difference would my UNIX V.3 for the 80386 system incur if > I added 120ns chips instead of the 100ns chips it already has 2 meg of? The > system is a personal system with no more than 3-5 users on at the extreme most. > > The reason I am considering the slower chip is mainly the price. > > Any advice or suggestions would be greatly appreciated. > I was running Unix V.3 on a 16MHz 386 system for some time. When I encountered a problem that the system wouldn't boot, I sent it in for repair. The guys at the depot replaced my 16MHz board with a 20MHz one. For a while I continued to use my old 4MB memory expansion board that was rated for 16MHz. I never encountered any problems with the slower memory, although I did eventually get new memory. Hope it helps. -- Tom Albrecht