Adventures in Server Land and Lab Rebuilds

OK, so short story:
If you have a PowerEdge 1950, and buy a 64GB RAM kit and shortly after upgrading you get either E1420 CPU Bus Perr errors (especially when loading ESXi), go into the Memory section of the BIOS and turn on Mirroring.

Now, the long story:

Alright, so this made me a bit irritated. Dell has 3 Generations of PowerEdge 1950, the Gen 1 Supports up to 32GB RAM, Gen 2.... eh, no real good documentation on that, and the Glorious Gen 3 which will allow this old server to go up to 64GB of DDR2 RAM. Beautiful for a home lab right?

Well, be careful about which version you have. If you're lucky and have a Dell branded one, you can look at the rack ear on the left hand side and it'll either be blank, or have a circle with an II (Gen 2) or III (Gen 3 obviously). Unfortunately, I don't. I have a pair of salvaged EMC RecoveryPoint servers which were 1950's rebranded. Yes I have service tags, no I don't have the markings. I check the website with the service tag, no bueno. I finally get in touch with someone in sales and they say, "Oh yeah! That's a Gen 3 and supports 64GB". Sounds reat right? I go out and buy 2x64GB kits for my pair, Install the ram in the first one.

Boots fine. Memory check took a while, so I went into the BIOS and disable it. The RAM counted up correctly (well, enough... shows 63.75GB of RAM, figured some was shared for video). Start to load ESXi. Locked up. Error on the LCD shows E1420 CPU Bus Perr. Looking that up shows a CPU Bus Parity error. Might as well be Greek to me. Ok, maybe my old server has something gone bad from moving, or worn out, or even maybe the bad RAM (the though did cross my mind, even though there were no memory errors indicated).

I get frustrated and call it a weekend. Hurricane hits. Tuesday finally rolls around and I figure "Hey, you have a second server, why not try that?". Dust off the server, install the RAM, power it up. Everything looks good. Windows 2003 loaded without a problem (well, slightly slow, but hey, it's Windows). No errors, no mus... no fuss. Since I'm running Standard, it only showed 32GB.

Alright - Let's load VMware!

Boot to the CD. Get to loading the modules. CPU Bus Perr. Now, I'll be quite honest, I cursed. Repeatedly. LOUDLY. Both servers! I figure, let me try loading updated firmware on anything I can. Start to load the firmware and before it even begins inventorying it locks up again. Yes, profanity ensued as I'm thinking about what the best retribution for this insolence would be.

I considered pulling half the RAM to test, but if the issue is the amount of ram, is the 32GB limit based on all banks being filled AND a maximum of 4GB modules, or any combination therein? Not really a good way to find out and with Dell no longer providing free tech support, it's time to tinker.

As I was poking around in the BIOS, I stumble into the memory section and saw something about redundant memory (my memory is a bit fuzzy on this, but you get the idea - no pun intended). I noticed an option for Mirrored RAM. Hey, let's try that!

Reboot, and the server now shows 32GB of RAM. Loaded the firmware disk and to my amazement, it runs faster and doesn't lock up. So there you have it. If you can't find the definite generation of your old 1950 (time to upgrade maybe? same for me), then if you buy the 64GB kit and get the E1420 CPU Bus Perr, then simply turn on Memory Mirroring in the CPU. You'll get redundant RAM modules AND you don't need to discard any of that precious RAM.

Stay Tuned, more mishaps to come!