dhale@csm9a.UUCP (Dave Hale) (09/07/89)
I recently purchased 4 1-megabyte, 70 nanosec, low-profile SIMMs from Technology Works (1-800-622-2210) for $114 each plus $10 shipping, for a total of $466. I bought these SIMMs intending to upgrade a couple of Macintoshes, but decided instead to upgrade the file and print server in our small cluster of NeXTs from 8 to 12 MB. These SIMMs seem to work just fine, and they cost a lot less than $1500, which is the price I was recently quoted from NeXT for a 4 MB upgrade. Has anyone else out there upgraded a NeXT with third-party memory? From my Mac experience, about the only thing I knew to be careful of was the speed of the memory. 70 nanosec is faster than the 100 nanosec chips that were shipped with our NeXTs. I also know that some SIMMs have 8 chips and some have 9 (1 extra for parity). The NeXT (and Mac) SIMMs have 8 chips. But are there any other potential problems? Incidentally, the NeXT screwdriver is quite useful as a tool for removing SIMMs. First, poke the end of the screwdriver into the hole at the end of the SIMM, pushing the little plastic hook out the other side of the hole. Then, with the screwdriver still in the hole, pry (gently) upward, using the nearest empty SIMM socket as a fulcrum. There are probably more appropriate tools for the job, but this worked nicely. Of course, if you are simply adding SIMMs, then you should not have to remove anything. But before adding the new SIMMs, I wanted to be sure that the pin layout on the new SIMMs was the same as that on the old SIMMs. So, I removed one of the old ones to get a good look at it. -- Dave Hale dhale@csm9a.colorado.edu (303) 273-3408 Department of Geophysics, Colorado School of Mines, Golden, CO 80401
cs141043@brunix (Ronald Antony) (09/09/89)
Is there any way to speed up a NeXT using faster chips? (less wait states?) Ron
c162-de@zooey.Berkeley.EDU (David Navas) (09/10/89)
In article <1875@csm9a.UUCP> dhale@csm9a.UUCP (Dave Hale) writes: >... But are there any other potential problems? Now, I don't claim to be an expert on the topic, but I read once that the NeXT machine uses nybble-mode RAMs. Mightn't this pose some serious hazzards to (at the very least) the speed of your machine?? >-- >Dave Hale dhale@csm9a.colorado.edu (303) 273-3408 >Department of Geophysics, Colorado School of Mines, Golden, CO 80401 David Navas (415) 643-2486 A complete nobody who can be found at c162-de@zooey.Berkeley.Edu and flamed at /dev/null
phd_ivo@gsbacd.uchicago.edu (09/13/89)
>In article <1875@csm9a.UUCP> dhale@csm9a.UUCP (Dave Hale) writes: >>... But are there any other potential problems? > >Now, I don't claim to be an expert on the topic, but I read once that the NeXT >machine uses nybble-mode RAMs. Mightn't this pose some serious hazzards to >(at the very least) the speed of your machine?? Is there a final resolution of the answer to this problem? I would buy a $400 memory upgrade if I knew it was the right memory. Has someone looked at the installed memory, and compared it to this memory (preferably someone who knows the difference between a nybble-mode and standard RAM) ? Information would be appreciated. /ivo phd_ivo@gsbacd.uchicago.edu ps: I am always curious to hear about the release date of 1.0.
dhale@csm9a.UUCP (Dave Hale) (09/20/89)
Awhile back, I posted a message stating that 1-megabyte SIMMs ( 8 70 nanosec chips per SIMM) originally purchased for a Mac from Technology Works seemed to be working fine in a NeXT. I received one email message from someone who was using SIMMs intended for a SUN 3 with "no problem". Since then, I have done a little more testing. At bootup, my NeXT reports that both the original and new SIMMs are "page mode" SIMMs, and that I have 12MB (which equals the new 4 MB plus the original 8 MB). In short, I have encountered no problems after about 2 weeks of use. I also wrote a small test program that repeatedly fills a large array. I have found that the largest size that will stay "in core" is 3 MB. The "ps" command reports that this is 25% of the available physical memory. I can get two such processes running in core at once (for a total of 6 MB), but thats it. A third such process causes lots of swapping. Does anyone know how to get a SINGLE large process running that uses more than 3 MB? (Or, in general, how do I use more than 25% of physical memory in a single process. On an 8 MB machine the limit seems to be 2 MB.) The "limit" command in csh seems to have no effect. -- Dave Hale dhale@csm9a.colorado.edu (303) 273-3408 Department of Geophysics, Colorado School of Mines, Golden, CO 80401
avie@wb1.cs.cmu.edu (Avadis Tevanian) (09/20/89)
In article <1895@csm9a.UUCP> dhale@csm9a.UUCP (Dave Hale) writes: >I also wrote a small test program that repeatedly fills a large array. >I have found that the largest size that will stay "in core" is 3 MB. >The "ps" command reports that this is 25% of the available physical memory. >I can get two such processes running in core at once (for a total of 6 MB), >but thats it. A third such process causes lots of swapping. > A single program can use all of memory if nothing else is running. However, the virtual memory system detects sequential access to memory and efficiently schedules pages to be paged out. This can result is ps listings that appear to be smaller than one might expect, although your pages are probably still in RAM. Note that swapping doesn't actually occur until you add the third process. I assume that you test is sequentially filling memory, you might try making it perform random access to see what happens. -- Avadis Tevanian, Jr. (Avie) Manager, Systems Software NeXT, Inc. avie@cs.cmu.edu or avie@NeXT.com