Decision making for older desktop

[timelord989]

Same system discussed here (didn't want to necro-bump + things have changed):http://www.tomshardware.com/answers/id-3355967/end-desktop-life-decisions.html

Got this used Dell Precision T7400 for $200 this Spring, and it's been working fine except about once a month it crashes and refuses to boot until I remove one of the memory sticks. So, I've gone from 16 GB RAM down to 12 theoretical (only 4GB recognized, probably due to the stick order).

Nothing else has broken, but this system uses DDR2 FB-DIMM RAM (e.g. https://www.amazon.com/Komputerbay-PC2-6400F-Buffered-Heatspreaders-Unbuffered/dp/B005HIWD6E/ref=sr_1_19?s=electronics&ie=UTF8&qid=1501347191&sr=1-19&keywords=ddr2+fb-dimm+800+mhz), so I'm not sure whether it's worth replacing.

The options as I see them:

• 
  Replace the RAM: Desktop works well enough for me, but this runs the risk of needing to replace the RAM again (perhaps bad mobo?), leading to ~$140/year upkeep.
  
  Buy/Build new Desktop, reusing SSD's & GPU: Higher initial cost but less likely to need more money/year. Not willing to go the used route again since this causes more problems than the savings justify IMO
  
  Sell all Desktop parts: I'm using my laptop increasingly more often than my desktop, so in theory I could just connect the laptop to my monitors (3 1080p possible? not for gaming) and not use a desktop.

Have I missed any options, and does anyone have opinions on which could be better?

EDIT: When it crashed today, I was able to get it to boot by taking one of the sticks and moving it to another slot (not just removing it like in the past). Could that mean it's the ram slots that are breaking (not the actual sticks), or is that unlikely? I still have all the original sticks.

 

[william p]

FBDIMMs have a memory controller on each module. They setup their own bus. This memory is very cheap used. You might replace the whole set.

 

[timelord989]

FBDIMMs have a memory controller on each module. They setup their own bus. This memory is very cheap used. You might replace the whole set.

That's interesting; does it mean that the sticks are what break (instead of the slot)? I guess it managed to reboot by ignoring the bad stick I moved to a different slot then.

I'm not sure I want to just replace the RAM since I don't know why they are randomly breaking. I doubt it's power surges since no other devices in the house are harmed (and this desktop has done this in two different cities). The sticks break in both idle and stress situations (I stopped BOINC'ing after the first stick broke).

If I could find out exactly what is happening I could better decide what to do next.

And about used memory: how can I avoid buying "lemon" equipment used? Nothing about the craig ad seemed odd; it seemed like a great deal! ($200 for 2x4 cores, 16 GB ram, 2x2TB SAS drives; only downside is old age - 2008 I think).

 

[william p]

Every time you reboot the sticks "talk" to each other and set up a memory bus with channels equal to the number of FBDIMMs in that bank. So a 4 channel bus is possible, but it also produces high latency. The extra logic chips make them run hot. If a stick fails then the bus collapses. The first stick is the "controller" that assigns data to the others. It may be that the chips get hot and one fails. You replace a chip but they've cooled down and run a while then fail again. This complexity could make it hard to tell which module is causing the problem.

 

[timelord989]

Every time you reboot the sticks "talk" to each other and set up a memory bus with channels equal to the number of FBDIMMs in that bank. So a 4 channel bus is possible, but it also produces high latency. The extra logic chips make them run hot. If a stick fails then the bus collapses. The first stick is the "controller" that assigns data to the others. It may be that the chips get hot and one fails. You replace a chip but they've cooled down and run a while then fail again. This complexity could make it hard to tell which module is causing the problem.

So you're saying I could try putting all the sticks back again and they might work? I'll try that next time I have access to the computer (next weekend)

 

[william p]

You might try replacing the modules in the first socket. It does much more work controlling the bus and runs much hotter. About 30W.

 

[timelord989]

You might try replacing the modules in the first socket. It does much more work controlling the bus and runs much hotter. About 30W.

My config before the crash was RR|RR|RR|__, and then I fixed it by moving the 6th stick: RR|RR|R_|R_. This reduced the "available" memory from 12GB to 4GB though.

 

[william p]

It looks like you made 1 channel 4GB so the other can't have more. I would go to Dell support and look at the memory configuration chart there. I still see 6 sticks there.

 

[timelord989]

You might try replacing the modules in the first socket. It does much more work controlling the bus and runs much hotter. About 30W.

I finally got 6 of the 8 sticks to work, tripling my usable RAM up to 12! Last week I had 4 sticks installed but only 4 GB usable ram. After reading the fine manual, I found out that they need to be in pairs, so that meant only 2 of the sticks were installed right.

Removing all but one and putting them back in carefully made it work. For all I know I only have one bad stick (all 8 = no boot + beeps, 7 would cause a memory error and/or not register the memory I think)

I'm marking this as the solution since you motivated me to try plugging/unplugging them.

Oh, and you're right that the sticks run hot; I almost burned my fingers on the first stick!