spierk@turing.cs.rpi.edu (Kevin Spier) (07/08/89)
I am looking into purchasing an extended / expanded memory board for my Epson Equity III+, and I have no idea what speed RAM I need. The machine runs at 12MHz with 1 wait state. Can I use 80ns chips or do I need faster? And can I just buy 70ns chips anyway; ie. is better to have faster chips even if you don't need them or does it cause problems? Also can Windows 286 v2.11 make use of extended memory besides just using it for swapping??? Any help would be greatly appreciated. Thanks, Kevin L. Spier spierk@turing.cs.rpi.edu Kevin L. Spier spierk@turing.cs.rpi.edu
bobmon@iuvax.cs.indiana.edu (RAMontante) (07/09/89)
spierk@turing.cs.rpi.edu (Kevin Spier) <5876@rpi.edu> : -machine runs at 12MHz with 1 wait state. Can I use 80ns chips or do I -need faster? And can I just buy 70ns chips anyway; ie. is better to It occurs to me that the simplest, most reliable answer to this sort of question is to look a the memory chips you currently have, and get more of the same. My understanding of waitstates is (still) weak, but if it really means just a second clock cycle in memory accesses, then 12MHz/1ws would be equivalent to 6MHz/0ws and 120ns chips would be adequate (but what does the machine already have??). In any case, I can't imagine how faster-than-necessary chips would do you any good, as the machine will take as long as it takes to access memory regardless.
chasm@attctc.DALLAS.TX.US (Charles Marslett) (07/09/89)
In article <23080@iuvax.cs.indiana.edu>, bobmon@iuvax.cs.indiana.edu (RAMontante) writes: > My understanding of waitstates is (still) weak, but if it really means > just a second clock cycle in memory accesses, then 12MHz/1ws would > be equivalent to 6MHz/0ws and 120ns chips would be adequate (but what > does the machine already have??). In any case, I can't imagine how > faster-than-necessary chips would do you any good, as the machine will > take as long as it takes to access memory regardless. Actually, a normal (no-wait) memory access takes 2 wait states, so if you add one wait state, it takes 50% longer, rather than twice as long. This means that a 12 MHz 1 wait state machine actually has 250 ns to fetch a word from memory. The real problem in figuring out how fast the RAM has to be is the time it takes to decode addresses and enable buffers (all that stuff in the C&T or W/D chip sets). They usually take 75-100 ns, so add that to the access time and if it fits, it's probably OK. Another "feature" of RAM chips is they get slower when they have company (like people) -- RAM on a large memory board will be slower than the same RAM chips on a smaller memory board. =========================================================================== Charles Marslett STB Systems, Inc. <== Apply all standard disclaimers Wordmark Systems <== No disclaimers required -- that's just me chasm@attctc.dallas.tx.us
ralf@b.gp.cs.cmu.edu (Ralf Brown) (07/10/89)
In article <23080@iuvax.cs.indiana.edu> bobmon@iuvax.cs.indiana.edu (RAMontante) writes: }My understanding of waitstates is (still) weak, but if it really means }just a second clock cycle in memory accesses, then 12MHz/1ws would }be equivalent to 6MHz/0ws and 120ns chips would be adequate (but what Actually, a wait state means a *third* clock cycle. 8086's take four clocks per memory access, which is slow enough that I've never heard of a PC or XT with wait states (though I've seen a number advertised as having 0 wait states--some "feature"!). 80286's and 80386's take two clocks per memory access, and the new 486 chip has a burst mode that can load the internal cache at the rate of one memory read every clock cycle. -- {harvard,uunet,ucbvax}!b.gp.cs.cmu.edu!ralf -=-=- AT&T: (412)268-3053 (school) ARPA: RALF@CS.CMU.EDU |"The optimist is the kind of person who believes a FIDO: Ralf Brown 1:129/46 | housefly is looking for a way out."--Geo.J.Nathan BITnet: RALF%CS.CMU.EDU@CMUCCVMA -=-=-=-=-=- DISCLAIMER? I claimed something?
mlord@bnr-public.uucp (Mark Lord) (07/11/89)
In article <5876@rpi.edu> spierk@turing.cs.rpi.edu (Kevin Spier) writes: >I am looking into purchasing an extended / expanded memory board for >my Epson Equity III+, and I have no idea what speed RAM I need. The >machine runs at 12MHz with 1 wait state. Can I use 80ns chips or do I >need faster? And can I just buy 70ns chips anyway; ie. is better to >have faster chips even if you don't need them or does it cause >problems? Also can Windows 286 v2.11 make use of extended memory >besides just using it for swapping??? Any help would be greatly >appreciated. > >Thanks, >Kevin L. Spier >spierk@turing.cs.rpi.edu > >Kevin L. Spier >spierk@turing.cs.rpi.edu 80ns ought to be more than fast enough. A simple, non-interleaved non-cached 12-Mhz system deals with 83.3ns per cycle. At one wait state, a memory access is therefore allowed 83.3+83.3=166.6ns to complete. Dynamic rams require their "access time" (such as 80ns or 70ns in your query) to respond with data, plus an allowance for other circuitry to generate the chip selects and strobe the addresses. This extra overhead varies from machine to machine, and about 30ns is probably a reasonable allowance. 80ns+30ns = 110ns = plenty of time. BUT.. dynamic rams also have a requirement for resting between accesses (hard work..), typically slightly less than their access times. In this example, it might be 70ns or 60ns, although this depends on the manufacturer. When figuring out this tolerance, the 30ns overhead need not be considered in most cases, so as long as the time available (166.6ns) is more than the chip "access time" (80ns or 70ns) plus the resting time (70ns or 60ns), it will probably be okay (depending on other system variables as mentioned). In this case, 80+70ns or 70+60ns are both less than the 166.6 cycle time for the 12mhz+1ws machine, so 80ns rams are just peachy. Most of the 12Mhz systems I have seen use 100ns chips. Perhaps this is because an extra cycle is used by the processor (2 cycles for memory accesses?) allowing plenty of time. I don't have a h/w manual handy to see exactly how many cycles the 80286 really uses when going to memory, but the above calculations are worst case. -Mark
bobmon@iuvax.cs.indiana.edu (RAMontante) (07/11/89)
ralf@b.gp.cs.cmu.edu (Ralf Brown) <5456@pt.cs.cmu.edu> : - -Actually, a wait state means a *third* clock cycle. 8086's take four clocks -per memory access, which is slow enough that I've never heard of a PC or XT -with wait states (though I've seen a number advertised as having 0 wait -states--some "feature"!). 80286's and 80386's take two clocks per memory -access, and the new 486 chip has a burst mode that can load the internal cache -at the rate of one memory read every clock cycle. Aha. Thanks....BTW, my Zenith 158 (XT clone) is a 1-waitstate box, at 8MHz. I think that one of the changes made to its Z-159 successor was to go to no waitstates.