memory for an LX164

[Davis Johnson]

I've got an AplhaStation 200 4/166 here (yes, I know) with 416 MB of RAM.

Every place I look reports 384 MB as the maximum ammount of RAM, for 
example:
http://h18002.www1.hp.com/alphaserver/workstations/retired/a200_specs.html

It has three pairs of 72 pin sim sockets, so 384mb is with 64mb Sims.

The way 416mb happened is I got the machine for cheap with 32MB. That 
obviously wasn't enough. Lots of memory vendors were more than willing 
to sell me memory guaranteed to work for more than the machine was 
worth. So I hung out on eBay, buying likly looking memory (parity, odd 
number of identical looking chips, lots of squinting at bad photos). I 
got a few 64mb Sims and a couple of 128mb sims. Stuffed them in and they 
work no problems.

"show memory" looks somthing like this (pardon typos, no easy cut and 
paste):

 >>>show memory
     416 Meg of System Memory
     Bank 0 = 256 Mbytes(128 MB Per Simm) Starting at 0x0
     Bank 1 = 128 Mbytes(64 MB Per Simm) Starting at 0x10000000
     Bank 2 = 32 Mbytes(16 MB Per Simm) Starting at 0x180000000

I don't think it is either unusual or guaranteed for memory devices 
larger than were originaly supported by the manufacturer to work. 
Sometimes higher density devices were not commonly available for testing 
and certification when the spec sheet was written. Sometimes there is a 
marketing reason. Sometimes there is actualy a hardware reason.

I think I got lucky. I also think it is a kind of chance I'd be willing 
to take again. Weigh the risks (possible but unlikly mother board 
damage, wasted money on Simms that won't work) against the benefits. Pay 
your money and take your chances -or- play it safe.

Michael Kronsteiner wrote:

>is that also true for AlphaPC164SX Mainboards? Im currently maxing out on
>one board with 512 MB Samsung RAMS (4x128MB) i dont *need* more ram, but
>alpha is a memory hungry platform - 256mb (in another machine) is just a
>little too few ...
>
>brgds & thanks in advance
>
>Mike
>
>
>
>  
>
>>--- Ursprüngliche Nachricht ---
>>Von: Juhana Helovuo <juhe at iki.fi>
>>An: axp-list at redhat.com
>>Betreff: Re: memory for an LX164
>>Datum: Wed, 11 Jan 2006 13:28:23 +0200
>>
>>Maurice Hilarius wrote:
>> >Ted Goodridge, Jr wrote:
>> >
>> >> You sure 256 sticks will work? My (albiet old copy of) manual shows
>> >> max  memory of 512 megs, 4x128 chips.
>> >>
>> >> I just wnat to be sure before I buy.
>> >>
>> >
>> >
>> >256MB modules work.
>> >The manual was written before 256MB modules were commonly available.
>>
>>[I am now replying to a rather ancient thread, but I ran across the same 
>>problem.]
>>
>>
>>Maurice is right in the fact that 256MB unbuffered ECC SDRAM DIMMs work 
>>in a 164LX motherboard. An 164LX data sheet by Samsung lists support for 
>>256MB DIMMs also.
>>
>>However, not all 256MB DIMMs work, even if they satisfy the spec above.
>>
>>For example, Kingston KVR133X72C3/256 (PC133) does NOT work, it is not 
>>recognized by the motherboard at all. I tested this myself.
>>
>>On the other hand, Kingston KVR100X72C2/256 (PC100) seems to work fine. 
>>I tested this myself also.
>>
>>The Kingston part numbers would lead one to think that the main 
>>differences between these modules are the maximum supported frequency 
>>(133 MHz vs. 100 MHz) and minimum CAS latency (3 clocks vs. 2 clocks), 
>>but since the 164LX uses a frequency less than 100 MHz (66?), and both 
>>of these modules should reach CL=2 at 100 MHz or less, the differences 
>>do not seem significant.
>>
>>The important difference seems to be the internal organization of the 
>>memory modules. The organization of the DIMMs is visible to the memory 
>>controller, as it has to generate memory addressing signals according to 
>>the structure of the DIMMs.
>>
>>The Kingston PC133 module is built from 9 chips of 256 Mbits each, with 
>>an internal organization of 32M x 8 bits each. It seems that the 21174 
>>memory controller on the 164LX board does not support this configuration.
>>
>>The Kingston PC100 module is built from 18 chips of 128 Mbits each, with 
>>an internal organization of 16M x 8 bits each. This is supported by the 
>>21174, although some older docs do not mention it.
>>
>>
>>I did not successfully test other brands of memory, but I believe they 
>>should be ok as long as they have a structure similar to the Kingston 
>>PC100 part mentioned above.
>>
>>
>>Juhana Helovuo
>>
>>
>>_______________________________________________
>>axp-list mailing list
>>axp-list at redhat.com
>>https://www.redhat.com/mailman/listinfo/axp-list
>>
>>    
>>
>
>  
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/axp-list/attachments/20060111/1f3d5a6e/attachment.htm>