Conference noted::ibmpc-95

    Be really really sure you can install SIMMs NOT in pairs. Check your
    motherboard manual. PCI based systems typically need the paired SIMMs.
    Someone has the technical reason for that? Is it because of 64-bit
    memory access?
    
    I have an ASUS PCI based motherboard and it clearly states that I can
    only install pairs of SIMMs. The pairs have to be absolutely identical
    (size, parity, ns, etc.).
    
    Guenther

re .1

>    Be really really sure you can install SIMMs NOT in pairs. Check your
>    motherboard manual. PCI based systems typically need the paired SIMMs.
    
this is VLB, and the docs clearly say pairs not needed.
this system was typically sold with 1 4MB SIMM. i upgraded to 2.
The doc has a table: 4, 8, 12,  up to 64MB and which SIMM to put in
which socket. 20MB is 4MB in 0, 16MB in 1.  48MB is 3 16MBs.

    Re .0:
    
    Chuck - Does your system have an ATI video card?  The Graphics
    Ultra Pro cards had a memory aperture setting which defined the
    base address of a region used to communication between the video
    card and the system.  You're supposed to set this to an address
    which is beyond the amount of physical memory in the system; e.g.,
    if you had 8 Mb, you'd set the aperture to (at least) 8 Mb.  If
    your system came from Gateway with the aperture set up that way,
    adding additional memory would cause all sorts of wonderful
    interference when you did anything using the video card - like
    what you're seeing.
    
    You have to run the INSTALL utility from ATI to change the base
    address of the aperture.  To be safe, I simply set mine to 100 Mb
    which was well out of the way for conflict.
    
    Your system does not need SIMMs to be installed in pairs; that's
    only for (most, but not all) Pentium-based systems.  Also, be aware
    that you MUST install memory in a very specific order (i.e., the
    smaller SIMMs should be in the first banks) - the order is listed
    in your motherboard manual.
    

	This is a 486, not a Pentium. One SIMM at a time is fine.

	What speed is the new SIMM, same or faster than the old SIMMs? Try
	the 16MB SIMM by itself and see how things work. If you still have
	problems, you could try playing with the BIOS settings wrt memory.
	The SIMM may have flaky bits, did you buy it new? Last but not
	least, as mentioned previously, some systems just have a problem
	when you mix memory types/sizes, even if the configuration is
	supposed to be valid.

Larry

re:
      <<< Note 2947.3 by STAR::DZIEDZIC "Tony Dziedzic - DTN 381-2438" >>>
                 -< ATI video card memory aperture conflict? >-

>    Chuck - Does your system have an ATI video card?  The Graphics
>    Ultra Pro cards had a memory aperture setting which defined the
>    base address of a region used to communication between the video
>    card and the system.  You're supposed to set this to an address
>    which is beyond the amount of physical memory in the system; e.g.,

Thanks!  I bet that is it.  I'll try it first.
I have an ATI Mach32 (or similar name).


>    You have to run the INSTALL utility from ATI to change the base
>    address of the aperture.  To be safe, I simply set mine to 100 Mb
>    which was well out of the way for conflict.

I like that idea, too.  No use solving the same problem again if I
ever add memory again.
    
>    that you MUST install memory in a very specific order 

the list of where to put what size SIMM always had all the 4MB SIMMS
in lower numbered sockets than any of the 16MB SIMMS. of course 16MB
can be realized in two ways.  they never said why to put the little
ones first.  i wondered if there was an operational reason, or if
they were just telling us not to remove and replace unless we had to.

thanks again. i'll report back ASAP.    

re
>       <<< Note 2947.4 by COOKIE::FROEHLIN "Let's RAID the Internet!" >>>
>                -< Does the manual list a 24 MB configuration? >-

yes. 24 is 4+4+16 in sockets 0, 1, and 2.
that was how I had them.

i was going to try permuting them to see if that provided any clues,
but i think i will not try that experiment.

	My system, which is a 486 system, DOES require that SIMMs be installed
in matched pairs.  It's an ASUS SP3 board.

re:                 <<< Note 2947.5 by TLE::INGRAM "oops" >>>


>	What speed is the new SIMM, same or faster than the old SIMMs? Try
>	the 16MB SIMM by itself and see how things work. If you still have
>	problems, you could try playing with the BIOS settings wrt memory.
>	The SIMM may have flaky bits, did you buy it new? Last but not
>	least, as mentioned previously, some systems just have a problem
>	when you mix memory types/sizes, even if the configuration is
>	supposed to be valid.

i did not notice the speed.  i'll check. perhaps i'll have to ask in
the "what chip is this" note, but i can probably tell. 
i'll try the 16MB SIMM alone. there are no waitstate settings in the
bios, so i expected the system to take as many as needed.
flaky bits is a possibility, but i don't think it is likely.
the croaking behavior is very consistent.
ditto for timing. playing music does not make more applications croak.

"unknown incompatibility" is still a good possibility.
the SIMM came with no packaging from Speedware. He will give a refund
if it does not work.  It was probably used, but he said tested and
guaranteed.

Tony's suggestion in .3 seems likely to be right.  I'll try it first,
then gather the speed info and try the 16MB alone if I have to.

I'll report back ASAP.  thanks again for the suggestions.

    re: .0
    
    This sounds like it might be a refresh problem with the 16MB SIMM. Many
    older 486 chipsets won't work correctly with certain newer type DRAMs
    (eg: 2Mx8, 1Mx16, etc). 
    
    How many DRAMs are on this SIMM (including both sides - if both sides
    are populated)?

re .8:
    
�	My system, which is a 486 system, DOES require that SIMMs be installed
�in matched pairs.  It's an ASUS SP3 board.
    
    But .0's board is VLB, yours is PCI, which usually do require pairs.
    

>   But .0's board is VLB, yours is PCI, which usually do require pairs.

Sticking my neck out a bit ...

Presumably (?) this difference has nothing to do with VLB vs. PCI per se,
but rather with the different chipsets used.

    re: .12 by WRKSYS::INGRAHAM
    
    >> Sticking my neck out a bit ...
    >> Presumably (?) this difference has nothing to do with VLB vs. PCI per
    >> se, but rather with the different chipsets used.
    
    Gosh I hope so!
    
    /Bill

    re .12: Quite possible - it just so happens that PCI boards usually use
    chipsets which require SIMMs in pairs, even if 486 based.
    

re several:

last night i tried the 16MB simm alone. the system would not boot.
no beeps.

the 16MB simm is labelled MH4M33SAJ   1435A00
there are 16 chips on one side and 17 on the other,
marked M5M44100AJ    141SF0H-8
(at least most of them looked alike. i did not check each chip for identical
labelling.)

i'm guessing the -8 means 80ns. i did not check the "what chip is this" 
note yet.  the gw2k docs say 70ns parity simms are needed.
there is no field in the bios to specify the speed or to set number of
wait states.  bios is  phoenix technologies 4.03 00.

the 4mb simms are labelled 0016   94-VO     HB56013    6BZ-7A    9301USA
the 8 chips on one side and 4 on the other are labelled
JAPAN A013      9301 2N0     HM5I 4400 A57

the memory aperture story is more complicated.  months ago, i lost disk C
to a viris.  when i was rebuilding, i discovered the ATI disks from GW2K
were corrupt.  i had faithfully copied them when i bought the system and
stored the copies and the originals separately.  the copies were bad,
but they were an accurate copy of the originals.  i got new drivers from
the ati web site. last night i could not find INSTALL.  another visit to
www.atitech.com and i was all set.  

install revealed that the memory aperture was enabled. there was no way
i could find to ask where or say where.  the gw2k docs are unclear.
in one place they say you can set the address, resolution 1M.  in another
they say it is automatically 2M past the end of physical memory.  in a
third, they say memory aperture does not work if there is more than
12MB on an ISA system.  mine is VLB, but it is ISA in lots of other ways.

anyway, since the only control i had was to disable memory aperture,
disable it i did.  the system booted, but only saw 16MB of the 24MB.
when i tried one of the programs that failed before, it croaked.

unless someone has another good idea, i'll return the 16MB simm today.
the gw2k string includes several reports of using nonparity memory
in systems that gw2k claimed needed parity memory.  at the compusa
memory sale last weekend, ther was over a 50% premium for parity.
i'll keep looking, or maybe try the nonparity memory.

thanks again for all the suggestions.  at least i got back a useful
utility program that i had not realized was missing.
notes and noters are great.

>there are 16 chips on one side and 17 on the other,

	33 chips? If all the chips are indeed the same, it sounds like you've
	got a SIMM with word parity (32 bit word + 1 parity bit (4MBx33))
	instead of byte parity (four 8 bit bytes + 1 parity bit for each byte
	(4MBx36)). This type of SIMM will work in some Sun systems, but won't
	work as a parity SIMM in an IBM PC compatible which requires byte
	parity. It also doesn't help that your documentation says to use 70ns
	memory and this is an 80 ns part.

	Exchange it or get your money back.

Larry

Here is an update on this story.

I exchanged the 33 chip 80ns simm for a 12 chip 60ns simm.
The croaking stopped.

All's well?  Not quite. With the system configured as 16 or 4+16 or
4+4+16 MB, only 16MB was seen. Motherboard switches were set properly.
I hoped it was an incompatibility between the 4 & 16.
I told the guy at Speedware I'd buy whatever memory we could make work.
At his shop, various combinations of 0 or more 4mb and 0 or more 16 mb
showed that only 16mb was visible.

So now I begin the work of deciding if I should replace the mother board
and CPU, or get a new system. But first, I'm going to take a close look at 
the switch pack that sets the memory size. Perhaps a stuck at 0 or stuck at
1 error could account for the experimental results.

I finally found the memory aperture setting in another screen of the
ATI INSTALL utility. It was set to 124M, so that was not the problem.
Memory aperture disabled or enabled gave the same results.

At first, when I realized I had old 80ns memory, I was ticked at Speedware.
Then I realized if it had worked, I'd have been content. I ended up with
4 4MB simms, so I got some money back.  The exchange was painless, and
he tried hard to find a memory configuration that worked. I'll seriously
consider Speedware for the upgrade or new system.

Thanks again to all for the help.