Dell PowerEdge R310 not starting – no power except for DRAC

Just ran into the same problem with an R310. Took quite a bit of digging but the link to the R310 Tech Guide helped. The 1GB UDIMM uses 128Mb x8 chips which is why it works. The Samsung sticks they keep sending you use 256Mb x8 chips and as you found in the Tech Guide the R310 won't work with 256Mb chips. I ran into the same problem using some memory we pulled out of an R710 server but they were 256Mb x4 chips so it was a double no-no.

Here is what I've found regarding memory. The actual DRAM chips on the memory module have differing numbers of lines. This is also referred to as the rank width, e.g., x4, x8, x16. The DRAM chips are also different sizes, e.g., 128Mb, 256Mb, 512Mb. The easiest way to explain ranks is to think of each rank as a virtual DIMM. So a single rank module is just that; a single memory module on a single physical DIMM. Dual rank memory actually has 2 memory modules on a single physical DIMM (it's deeper than that but this is the easiest way to visualize it).

So the bottom line is that the R310 will only accept DIMMs that are made out of 128Mbit or 1Gb/2Gb chips that are x8 devices. Or at least that's my story and I'm sticking with it until Dell tells me otherwise; e-mail to them is out and I'll let you know if they say something different.

Link to Hynix part: http://www.hynix.co.kr/datasheet/eng/module/details/module_20_HMT112U7BFR8C.jsp?searchType=2&menuNo=1&m=1&s=1&RK=20&RAM_NAME=

A link to an excellent briefing from Intel: http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CDsQFjAB&url=http%3A%2F%2Fdownload.intel.com%2Feducation%2Fhighered%2Fsignal%2FELCT865%2FClass2_21_lab_and_project_assignment3.ppt&ei=VKOiT-yKDoau9AS86730CA&usg=AFQjCNHOMlxRyI0C8F7GIdwMkU8h1ZOmyQ&sig2=nLNmplaHcJjbv7rktoqb2Q

Slide 13 is the key, but the entire package is very good depending on your level of geekdom :)

Update: Confirmed with Dell except that I was wrong about requiring 128Mb and smaller chips. The Foxhollow platform does support 1Gb and 2Gb chips. See here: http://cache-www.intel.com/cd/00/00/43/26/432686_432686.pdf

Well, you've actually got a lot of good information.

• First, you know that the problem doesn't affect a running system assuming it runs for extended periods of time without issue (e.g. weeks). You might double-check that with tools designed to stress your system (cpuburn or the like), but let's take it as read for now.
• Second, you know that the problem does prevent booting.
• Third, you know that the problem reliably goes away with time.

What that tells me is that there is something that is true immediately after the system shuts down that is not true 1-24 hours after shutdown. Also, that fact only matters when the system is booting.

My first thought is heat, but there are some tests that you don't mention doing:

• In case the problem is some electrical state remaining in the computer, I would try unplugging the power, and then holding the power button in for 10-20 seconds.
  • It takes a little while for the last remaining charge to bleed off, possibly 1-24 hours.
  • If you listen and watch very carefully, you can probably hear a fan start to spin up, or an LED light and then dim to off when you do this.
  • You're more likely to see the above if you do this with the cover off.
  • If the problem remains, it is not a simple case of some critical value in memory failing to clear.
• The other test I suggest is to boot the system to full operational status, and then immediately shut down (no more than 5 minutes on) and see if the system will restart successfully.
  • If it restarts, you know that the problem takes some time to build (my first thought would be heat at this point)
  • If it doesn't restart, you know that the issue is triggered almost immediately. You can restart at different points during the POST to narrow it down, but ...

I have seen systems that will not reliably recognize the RAM in the system. If this happens and results in no RAM... well, I would expect more booting to occur than you describe, but I wouldn't bet my life on it.