Port-sparc64 archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Sun Netra T1 amd "memory pickyness"



Hi,

I wanted to upgrade my Sun Netra T1 200 memory. I bought some PC133 ECC BUF RAM on ebay from different sources and found it difficult to assemble a stable system.

The Netra started with two banks, in Bank 0 and Bank 1 of assorted RAM. I t is:
Apacer Module with Infineon 7
Kinston Module  with Samsung TC75

This configuration (even if 7 vs 7.5) works quite well: OPB passes, system boots, works, compiles....


First I bought 2 Kingston PRç133/512 DIMMS, they have Infineon 7.5 chips. Put them in... system recognizes them but fails OPB memory test with possible error.
Nice, I thought... got a lemon? lets buy another pair.

I got a pair of old HP modules:

HP PC 512MB/133MHZ  REG SDRAM. Thet have Samsung TC75 chips on it.

The Two HP modules plus the old Apacer + Kingston-Samsung. This worked well... for one day compiling and network usage, then system locked up. It did pass OBP and diagnostics, but halts on NetBSD boot loader. Wow! Probably something fried.

Tried a couple of combinations, system fails... so I did one thing.

First test:  ALL 6 DIMMS putting only one in Bank 0 and see if it boots, one after the other.

One of the DIMMs casues a lot of numbers appear on serial console, like some testing? All 5 other dimms get detected 512MB and NetBSD boots to prompt. I suppose one of the two HP modules really did "fail" under load... bad luck? Well, used stuff on ebay.
Put it aside and mark it as possibly bad.

Second Test: 2x Kingston Infineon (Bank 0, Bank 1): detected, but boot fails with:

Loading netbsd: 12135360+511216+333496 [742800+488129]=0xefb578

Watchdog Reset
Externally Initiated Reset

Single tested they did both work!

Third Test: Kingston-Samsung Bank 0, HP "good" in Bank 1
OBP passes, NetBSD boots fine.


Forth Test: Kingston-Samsung Bank 2, HP "good" in Bank 3
OBP passes, NetBSD boots fine.

From this I would say that All banks 4 banks work.

Firth Test: Apacer Bank 0, Kingston-Samsung Bank 2, HP "good" in Bank 3
OBP passes, NetBSD boots fine. 1.5 Memory seen

Now remember Test 2 failing and let's do

Sixth Test:
Apacer Bank 0 - Kingston Infineon "A" Bank 1
OBP passes, NetBSD boots fine.

Seventh Test
Apacer Bank 0 - Kingston Infineon "B" Bank 1
OBP passes, NetBSD boots fine.

So Kingstion Infineon is incompatible with two modules? Both Kingstion Infineon modules (didn't mark them A B) work fine as second one

Eigth Test:
Kingston Samsung Bank 0 - Kingston Infineon Bank 1
OBP passes, NetBSD boots fine.

Ninth Test:
Kingston Infineon Bank 0 A - Kingston Samsung Bank 1 -
Watchdog Reset ! Fail!

Tenth Test:
Kingston Infineon Bank 0 A (Just removed the other one)
Works! so it is not that one... it is together that doesn't work.

Ninth Test:
Kingston Infineon Bank 0 B - Kingston Samsung Bank 1
Works!


This is Close to absurd! Would guess Kinkstion Infineon A is somehow bad in that it works only alone and not with others!

So really bad luck? Two DIMM couples one has a bad module, one has a nasty module?

Tenth Test:
Kingston Infineon B Bank 0
Kingston Samsung Bank 1
Apacer Bank 2
HP "good" bank 3

Starts boot, but....
[   1.0000000] atabus0 at aceride0psycho0: uncorrectable DMA error AFAR 21fc010 AFSR 0x480000ff40800000<BLK,S_DRD,P_DRD>
Stopped in pid 0.0 (system) at  netbsd:cpu_Debugger+0x4:        nop
db{0}>

Sounds again some nasty memory error!

So bad luck or is the Netra T1 picky about memory combinations.... since modules are buffered they should play well together, shouldn't they?
Any other has a similar experience?

Can't cobble together a combination to 2GB

I wonder if my questionable modules are really bad. I suppose the one that doesn't work single is, most probably,
Should I get another module ??



Riccardo




Home | Main Index | Thread Index | Old Index