carter@PORTIA.STANFORD.EDU ("Thomas J. Carter") (08/08/89)
Not too long ago I bought 4 1meg SIMMs which I was told were 80ns. Do I just have to take this on faith or is there some way to check? How do I know someone didn't give me 150ns SIMMs either accidentally or because they figured I'd never know? Tom Carter carter@portia.stanford.edu
adam@mit-amt.MEDIA.MIT.EDU (Adam Glass) (08/08/89)
carter@PORTIA.STANFORD.EDU (Thomas Carter) writes: > Not too long ago I bought 4 1meg SIMMs which I was told were 80ns. Good for you! Isn't lots of memory just grand? > Do I just have to take this on faith or is there some way to check? Oh, no, no... you can check... > How do I know someone didn't give me 150ns SIMMs either accidentally > or because they figured I'd never know? Funny you should ask... Look at the individual memory modules (chips) on the SIMM boards. Up in the top right corner of the chips there should be a speed rating. "-15" means 150ns. "-12" means 120ns. "-8", unsurprisingly is 80ns. These methods of marking have been consistent with all of the SIMMs that I have seen, though I am not certain that yours will have the same markings (but more likely than not, I think). I honestly don't think you'll notice a large difference between 150ns and 120ns. Those are the maximum amounts of time that the chip could take to return the info. Most chips will perform below (i.e., faster than) their reported speeds. So what's a few billionths of a second between friends? Adam =============================================================================== || Adam Glass, NeXT hacker for the slave-drivers at the MIT Media Laboratory || || #include <std_dsclmr.h> || || %% "But Westley, what about the R.O.U.S.es?" %% || || %% "Rodents Of Unusual Size? I don't think they exist..." %% || || EMail to: adam@media-lab.media.mit.edu OR adamg@athena.mit.edu || ===============================================================================
bayes@hpislx.HP.COM (Scott Bayes) (08/09/89)
[...] >> How do I know someone didn't give me 150ns SIMMs either accidentally >> or because they figured I'd never know? > >Funny you should ask... > Look at the individual memory modules (chips) on the SIMM boards. Up >in the top right corner of the chips there should be a speed rating. "-15" >means 150ns. "-12" means 120ns. "-8", unsurprisingly is 80ns. These methods >of marking have been consistent with all of the SIMMs that I have seen, though >I am not certain that yours will have the same markings (but more likely than >not, I think). I honestly don't think you'll notice a large difference between >150ns and 120ns. Those are the maximum amounts of time that the chip could >take to return the info. Most chips will perform below (i.e., faster than) >their reported speeds. So what's a few billionths of a second between friends? > The problem is, if the SIMMs are not fast enough for the bus they're running on, you will have unreliable or just plain don't-work memory. It's no problem if the SIMMs are faster than the bus (e.g. require 120 nsec, have 80 nsec SIMMs), but watch out if they're slower. If the memory is rated faster than you need, you don't see any speedup or any other benefit on that computer. You only benefit when you move the SIMMs to a faster machine. Given the way technology moves and prices (usually) drop, most of us will probably never move our RAM to a faster machine--new RAM will be bigger, cheaper and faster than the RAM we bought 2 or 3 years ago. The 1MB RAMs I got just 3 years ago for my HP machines are useless or ridiculous on my new machines. 4 MB (or often 8 MB) is the minimum step size now, in general, and adding 1 MB is no more useful than throwing a cup of water in the ocean, even were it compatible. It's often not worth buying faster RAM. So don't spend extra on "faster" SIMMs, unless you're pretty sure you're gonna' put them in a faster machine SOON. If that faster machine is in the far future (> a year or two), bigger, fast RAM will probably be nearly as cheap as the price difference you'd pay now anyway. >Adam > >=============================================================================== >|| Adam Glass, NeXT hacker for the slave-drivers at the MIT Media Laboratory || >|| #include <std_dsclmr.h> || >|| %% "But Westley, what about the R.O.U.S.es?" %% || >|| %% "Rodents Of Unusual Size? I don't think they exist..." %% || >|| EMail to: adam@media-lab.media.mit.edu OR adamg@athena.mit.edu || >=============================================================================== Re memory speeds: on a CPU like the 680x0 family, memory read access works roughly this way (minimal flames please)--the CPU asks to read a datum, the memory system has so many nanoseconds to put the datum on the wires, then the CPU reads what's on the wires, regardless of whether the memory system actually put it out there in time or not (the above discussion is very simplified, and doesn't include perturbations like the bar-DTACK signal, DMA access, etc.). So if the memory is too slow, you may get random garbage coming back. If it's excessively fast, it doesn't really matter, as the CPU doesn't know that, and does the same speed access, no matter the SIMM speed. Write works in somewhat similar fashion. Scott Bayes "down near the metal"
stevel@eleazar.dartmouth.edu (Steve Ligett) (08/22/89)
In article (well, I don't know what article really, it arrived and disappeared while I was on vacation) someone writes: >> How do I know someone didn't give me 150ns SIMMs either accidentally >> or because they figured I'd never know? > >Funny you should ask... > Look at the individual memory modules (chips) on the SIMM boards. Up >in the top right corner of the chips there should be a speed rating. "-15" >means 150ns. "-12" means 120ns. "-8", unsurprisingly is 80ns. These methods >of marking have been consistent with all of the SIMMs that I have seen, though >I am not certain that yours will have the same markings (but more likely than >not, I think). A better method is to check the back of the SIMM for a -15 or -12 or whatever. Manufacturers such as TI and Toshiba speed-test the SIMMs after assembling. A SIMM made of all 120 ns chips won't necessarily meet 120 ns specs. I've seen SIMMs with all 120 ns chips marked -15 on the back (and had them fail in testing that required 120 ns SIMMS), and I've seen SIMMs made with a mix of speed chips. A lot of "garage" operations (like mine), don't speed-test the SIMMs they make. I just buy faster chips than are needed (nowadays I only buy 80 ns chips). On a Mac Plus, 200 ns SIMMs would be fast enough. Steve Ligett steve.ligett@dartmouth.edu or (decvax harvard linus true)!dartvax!steve.ligett