Re: RAM for digital server 5300

[Jay Estabrook <jay.estabrook@xxxxxx>]

Simon Tierney wrote:
> Thanks for replying on this elementary hardware question Jay, I think I'll
> risk trying the 512's because W. Bohrer has got them working and I just want
> to beef up the RAM, not max out of the 4Gb limit.

Someone actually has tried it? How many did he use, and how did the
memory get configured?

I don't think you can hurt the machine with them, but I'm quite
surprised that it was bootable with them in it; in fact, frankly
I'm amazed... ;-}

Is it that you can't get the 256MB DIMMs now, and only the 512's
are available? Otherwise, I'd think that two 256's would be cheaper
than 1 512...

Good luck, and please let me know how it works out.

--Jay++


> Thanks,
> 
> Simon
> 
> ----- Original Message ----- 
> From: "Jay Estabrook" <jay.estabrook@xxxxxx>
> To: "Linux on Alpha processors" <axp-list@xxxxxxxxxx>
> Sent: Wednesday, April 25, 2007 9:14 PM
> Subject: Re: RAM for digital server 5300
> 
> 
> Simon Tierney wrote:
>> I have one of these machines which has cards to take PC100 ECC SDRAM
>> modules, 1 card for each CPU.
>>
>> The Service and User manuals require replacement RAM  modules of 128 Mb
> and
>> 256 Mb (in pairs), but larger modules (512Mb) have become available since
>> these manuals were written.
>>
>> Has anyone any experience adding these larger modules?  Will the system be
>> able to address them, or should I stick to 256 modules?
> 
> I've no experience trying those 512MB DIMMs, but the docs do say that
> the memory limit is 4GB done by 8 pairs of 256MB DIMMs, so I really
> would NOT recommend trying anything larger. From the docs:
> 
> "The rules for addressing memory are as follows:
> 
>     * A memory pair consists of 2 DIMMs of the same size.
>     * Memory pairs in riser cards may be of different sizes.
>     * The memory pair in slot 0 must be the largest of all memory
>       pairs. Other memory pairs may be as large but none may be
>       larger.
>     * The physical starting address of each memory pair is N times 512MB
>       (2000000) where N is the slot number on the riser card.
>     * Memory addresses are contiguous within each module pair.
>     * If memory pairs do not completely fill the 512MB space provided,
>       memory “holes” occur in the physical address space.
>     * Software creates contiguous virtual memory even though physical
>       memory may not be contiguous."
> 
> So, it's really a hardware limitation in the way that memory DIMM pairs
> are mapped into the physical address space, that tops out at 4MB on the
> 5300.
> 
> Sorry.
> 
> --Jay++
> 


-- 
Jay A Estabrook                            HPTC - XC I & B
Hewlett-Packard Company - ZKO1-3/D-B.8     (603) 884-0301
110 Spit Brook Road, Nashua NH 03062       Jay.Estabrook@xxxxxx