Re: Sun Netra T1 amd "memory pickyness"

[Michael <macallan1888%gmail.com@localhost>]

Hello,

On Fri, 26 Sep 2025 15:20:35 +0200
Julian Coleman <jdc%coris.org.uk@localhost> wrote:

> > Mixed with bad lucks of a probable bad module, I would say the Kingston
> > Infineonn 7.5 is not a nice player.
> > It has chips HYB39S256400DT-7.5  
> 
> And this is strange, as that is similar to the chip that is shown on the
> genuine Sun part!  I wonder what the suffix "DT" versus "T" signifies.

Hmm, this is what lives in my v120, which should be fairly similar to the T1:

[     1.000000] spdmem0 at iic0 addr 0x54
[     1.000000] spdmem0: SDRAM (registered), data ECC, 512MB, 133MHz (PC-133)
[     1.000000] spdmem0: 13 rows, 11 cols, 1 banks, 4 banks/chip, 7.5ns cycle time
[     1.000000] spdmem0: tAA-tRCD-tRP-tRAS: 3-15-15-37
[     1.000000] spdmem0: voltage LvTTL (not 5V tolerant), refresh time 7.8us (self-refreshing)
.. four identical modules. Not sure if they're Sun branded, I never really messed with them.

> > Perhaps it makes to buy another DIMM off eBay and see if I have good luck. I
> > would say the board is still sound and it is a DIMM issue. Perhaps I can
> > read the chips in the picture and stay away from Infineon 7.5 :) Either
> > faster infineon or Samsung.  
> 
> I guess that the machines were tested with some subset of available chips
> and the Sun parts have only those chips.  They should all be compatible,
> but it seems from your experiments that they aren't.  That was why I thought
> that getting a set with the same chips/stick number as a known Sun part would
> work.

I had mixed luck with memory modules for my Blade 2500. In theory it
needs PC2100R/ECC, in reality it rejects anything with more than 2.0ns
CAS time, and there's a bunch of 2.5 ones out there, rarely marked as
such. It does run fine with a bunch of Samsung PC2700R/ECC modules.

I guess what I'm trying to say is, if the issue is something like that
then faster modules should have a better chance of working.

have fun
Michael