New RAM == unstable 1

[tanderso]

I've seen posts that have been somewhat similar to my situation, but I haven't seen a convincing response yet. Here's my situation... I have a Soyo Dragon+ MoBo w/ Athlon 1600+ w/ 2x 256M PC2100 DIMMs. One DIMM is original to the PC, the other is a replacement that was sent from the manufacturer when the other original developed errors. I don't know the brands on either, but they are both 256M PC2100 DDR, as that is written on both. Now, I've received a PNY 512M PC2100 DDR DIMM for Christmas and added it to my system for a total of 1G of memory.

I immediately noticed several applications crashing mysteriously, including Mozilla, some of my taskbar apps, Tux Racer, and even some of the daemons on boot. Log files indicate segmentation faults, etc., and provide stack traces. I ran some memory testing software, and they also failed with memory errors. I'm running Mandrake 9.0, and when I installed the OS, I provided for larger than 3G of memory, so I don't think it's an addressing problem.

Ok, so I try some simple troubleshooting. First, I underclock everything. I rearrange the DIMMs in the slots... still errors. I remove the new DIMM and try the original two alone... all fine in that configuration. I then remove the original ones and use only the new one... all good again! Then I tried using one each of the 256 with the 512... again, it works fine! So, my only problem seems to be when I use all three DIMMs at the same time. Does anyone know what might be the problem? Does this seem like a software problem or a hardware problem?
Sincerely,

Tom Anderson
CEO, Order amid Chaos, Inc.

http://www.oac-design.com

 

[cdogg]

It's still unclear as to whether it's software or hardware. I see you were able to get the 512MB stick working with one 256MB stick. Are you saying it becomes unstable if you leave the 2 sticks where they're at and add the other 256MB stick in the 3rd slot?

The next step would be to upgrade/flash the BIOS to ensure the most up-to-date compatibility:

http://www.soyousa.com/downloads/

Also, I'm curious about the power supply. Perhaps there is some kind of inconsistency there. To be honest, it definitely seems to be a hardware problem since the software under Mandrake runs fine in a certain config.


~cdogg

"The secret to creativity is knowing how to hide your sources."
- A. Einstein

 

[gargouille]

Am I to understand correctly that you ran some memory diagnostics and some of the memory failed? (although it would be difficult
Did you run the tests again with the new population scheme?
(am unclear about what "still errors" in the last paragraph of your post means) errors in the log files or in the memory tests?

 

[gargouille]

Well, that's a perfect non sequitur.
the second sentence should say:
(although it would be difficult to tell which module until
more experimentation is done, as in trying 1 stick at a time.)
We're entirely confident the new stick is the correct ram?
You say you underclocked the ram in the last experiments.
Was it overclocked before?
When you tested one of the old and the new one were you using different slots.
If I were in your shoes I'd try 1 stick at a time, run the memory diagnostics and rely on the results.

 

[scottyjohn]

Im a bit of a novice but, in researching some mobos for a new machine Im building, I notice that some have 4 dimm slots but only 2 of these are usable at a time when using PC2100/PC2700 memory. Could this be the source of your trouble? Im not sure what symptoms using more than 2 would display but I thought Id suggest it anyway? John
ski_69@hotmail.com

 

[tanderso]

1) Yes, it becomes unstable after adding the last stick of memory, no matter which two are currently installed.

2) The BIOS upgrades do not address anything concerning memory. Also, at boot, the BIOS reads the full 1G and passes it in it's test. I'd rather not make too many changes to the otherwise stable system or else risk not knowing what is producing some new error.

3) The power supply is 350W, powering an Athlon 1600+, two HDDs, a Zip, a Floppy, and a DVD. Considering the removable media aren't in use and one of the HDDs is generally powered down, the power supply is more than adequate. Adding an additional 256M DIMM isn't any additional strain than spinning up the DVD, which never causes errors.

4) When I say there are errors, I mean that programs terminate with segmentation faults, stack traces, etc., sometimes at bootup, often in other software I run, and always in the memory test programs. When I'm running only two DIMMs (either two), the system is solid as a rock and the memtest programs report zero errors. Thus I am fairly certain that none of the RAM contains errors, but only cause problems when all three are present.

5) All three DIMMs are PC2100 (266MHz DDR), which is the correct RAM for my MoBo (SY-K7V Soyo Dragon Plus!).

6) I had my MoBo overclocked to approximately a 1700+ before adding the new RAM. Although it was negligible, I've now underclocked it to be certain that it wasn't causing any problems. Additionally, I've set all of the memory timings to their most conservative to minimally stress the RAM. The DIMMs all function perfectly either alone or with any one other DIMM, overclocked or underclocked, in any slot combination. Adding the third produces the errors.

7) The SY-K7V has three DIMM slots, which the manual says may all be populated at the same time for up to 3G of total memory.

8) I've made an additional observation... my system monitor utility under Gnome only recognizes about 850M when I have the full 1G installed.
Sincerely,

Tom Anderson
CEO, Order amid Chaos, Inc.

http://www.oac-design.com

 

[SoulAssassin]

The only thing I can think of (since you've covered everything I would have done) is that the third DIMM slot is bad. Figuring in that any combination of any three pieces used cause the problem, the RAM is obviously not bad but that third slot can very well be bad. I'd call Soyo to talk to them but try to return it and get a new board. That shouldn't happen with your RAM.

 

[gargouille]

Forgive my persistence with the pessimism, but are we sure the 3rd (new) stick (even tho it's PC2100) is not ECC or some other? or is otherwise mismatched to your system (registered)?
I know not what Gnome is...assume 3rd party googaw...but from what it says the memory density may be wrong and its only seeing half of one stick (the 512) or none of one (the 256).
I still would run a memory diagnostic...
but that's just me.

 

[jonhiker]

I'd run the memory diagnostic as suggested by gargouille.

However, it sounds like a bad 3rd RAM slot. Especially if you can make 2 RAM chips in various combos work.

 

[mainegeek]

Are you sure your mobo will handle over 768mb?
Maybe try 3 256mb?

 

[icrf]

I'd go with a faulty motherboard. I've read that some boards, thought they have three or four slots, are not stable when fully populated. Yours should be listed here (second from bottom?):

http://www.anandtech.com/mb/showdoc.html?i=1578&p=24

They reported stable with three of three slots filled. ----------------------------------------------------------------------------------
...but I'm just a C man trying to see the light

 

[tanderso]

I've determined that none of the slots are bad, as I've successfully run diagnostics with memory in each of the slots in multiple combinations.

I've made a docmem boot disk, and the memory passes the tests with the full 1G. Therefore, it must be a linux problem rather than a hardware problem.

I've downloaded the latest kernel, loaded the Mandrake default config, and modified it so that the "high memory" value is 4G rather than "off" as default. I also changed the processor to Athlon and removed some extraneous modules, but that's beside the point. I've just finished compiling, and I'll let you know if it's fixed after reboot.
Sincerely,

Tom Anderson
CEO, Order amid Chaos, Inc.

http://www.oac-design.com

 

[tanderso]

I'm moving this thread over to thread619-436793 in the linux forum, as I'm fairly certain this is not a hardware problem. It seems that Mandrake defaults its kernel to high memory = off, which only works for systems using 960M or less. Please go to thread619-436793 if you want to follow this and help further. Thanks!
Sincerely,

Tom Anderson
CEO, Order amid Chaos, Inc.

http://www.oac-design.com

 

[gargouille]

Thank you, Tom, for your patience and feedback.
Now that we know it's not a hardware problem!