Supermicro X9DAi Weirdness

[Luke Boyce]

So the past several days have been frustrating for me. I switched to a PC build from working on Mac Pro's. Went with the Supermicro X9DAi as recommended multiple places. Unfortunately, it has not been a good experience and I want to see if anybody has any thoughts. First time we built the PC we through 128 gigs of Kingston 1600 RAM in there (as recommended by Jeff Kilgore) in all 16 slots, 8gb each. Computer would turn on but wouldn't boot into the BIOS. After many attempts we eventually narrowed it down by placing each RAM in one at a time in various slots. We finally figured out that any RAM placed in the top 2 slots of the bottom 4 DIMMS on the right CPU the computer wouldn't boot. Everytime we took them out, no problem. We tried multiple configurations with different sticks, both in and out. Same exact results, so we knew it wasn't our RAM. So we assumed that those slots were faulty and I returned the board and received the new one today. After rebuild: exact same problem on the exact same DIMM slots with a totally new board. What. The. Crap?!

Anybody care to enlighten me here? What am I doing wrong?

At the moment I'm running everything with those slots empty at 112gb of RAM (I know, poor me) and there's a few other strange things going on, but nothing too much. The main weirdness is that I have Windows running off of a brand new Kingston 240gb SSD. After installing Windows and updating everything, the computer is telling me that I have 220gb full on the SSD. All I have is Windows on it. What is going on there!? Things seem to be running okay outside of that, except for randomness here and there which I attribute to PC world.

But if anybody has any ideas regarding this, I'd love to hear 'em.

Thanks

[Joseph S.]

Luke, the reason your drive is full is because you installed the OS with all the RAM loaded, and Windows made a huge page file. I don't have my windows running, but right click my computer, and under advanced settings, just remove your page file (option for that). You'll need a reboot. If the page file is still there, just delete it. I forgot the exact detailed method, but that should do the trick.

BTW, I've used that board with 128GB and it works with all the slots. There is a list on supermicro for that model board telling you which memory part numbers are compatible. If it's not on that list, it may or may not work.

These are known to be certified 8GB modules:

MT36JSF1G72PZ-1G6M1
M393B1K70DH0-CK0

[Jeff Kilgroe]

Er... Did I recommend any Kingston RAM for this motherboard?

What specs are the RAM?

Do any/all modules fail, or the system fail, to post when installed in those specific DIMM slots? If so, that seems like a bad motherboard and I wouldn't waste my time trying to make it work by just not installing memory modules in slots that cause problems.

Installing Windows with all the RAM in place should have no bearing on page file size or disk consumption on a properly operational system. There's definitely something amiss here and based on the little bit of info posted so far, it seems like a defective motherboard and/or defective or incompatible RAM.

In my X9DAi system, I'm running Micron RAM. I have not tried any other RAM in this board, I know that Joseph used Samsung RAM in his. There was a lot of back and forth for a bit here in the forums because the RAM was (and still is) hard to come by at times. Last I recall talking about Kingston modules as a possibility was that they seem to align with what HP may be using in the Z620 / Z820 workstations. While kingston makes modules that have the proper specs on paper, SuperMicro has not added any Kingston modules to their tested memory module list.

[Joseph S.]

The pagefile is usually twice the amount of RAM. Happens on all installs. Even in linux the swap file is twice the amount of ram (normally).

Some people say leave at least 2-4GB swap.

This is the way to check and remove it:

http://windows7themes.net/how-to-del...windows-7.html

[Luke Boyce]

Yep, the pagefile thing fixed it perfectly. Thank you so much for that. Everything seems to be running wonderfully now. I've got the RAM in all of the slots except for the 2 and everything works, so I may just be going with the 112gb for now.

Jeff, yes you did mention that you were very sure that the RAM in your Z820 was the Kingston. The specs matched up and the price was right so I went for it. I had a gut feeling that it could very well be the issue, but can't test it unless I threw approved RAM in those slots. And yes, the system fails to post anytime ANY RAM is in those slots, no matter what. I suppose if I encounter any problems along the way that seem like they could be related, I may have to move to the approved RAM, but as of now the system is actually running quite well. RED footage plays beautifully at 4k and everything is FAST.

Thanks for the help!!

[Jeff Kilgroe]

The pagefile is usually twice the amount of RAM. Happens on all installs. Even in linux the swap file is twice the amount of ram (normally).

Yeah, I suppose it does do that initially. I guess I haven't run a small enough SSD for the OS in relation to the amount of RAM to encounter problems. Once loaded, I go through and tweak everything anyway...


Luke, as for the RAM issue in the non-working slots. This most certainly should not be happening. It could be due to a RAM incompatibility, but also highly likely that it is indeed a defective motherboard. I know it's not the easiest thing to diagnose without sourcing more components to test with. I would contact SuperMicro about it and see if they can assist. But it's definitely something I think you should diagnose in case the motherboard is in fact a dud. It may start with a couple DIMM sockets not working properly, could lead to memory corruption and therefore data corruption on disks or other problems down the road if there are electrical problems.

You are running 1600MHz RAM at 1.5V, correct? If you're running 1.35V modules, you will have a mis-match.

[Eric Z]

Luke,
Here are your RAM options on this X9DAi (Intel C602 chipset) with the RAM modules you've purchased:


16x 8GB 1600MHz Dual-Rank(x8) NON-ECC UDIMM modules, 1.5V
16x 8GB 1600MHz Dual-Rank(x8) ECC UDIMM modules, 1.5V


16x 8GB 1600MHz Dual-Rank(x8) RDIMM modules, 1.5V
16x 8GB 1600MHz Single-Rank(x4) RDIMM modules, 1.5V
16x 8GB 1600MHz Dual-Rank(x4) RDIMM modules, 1.5V


Other Important Notes and Restrictions:
• All RAM modules should be of the same type, same speed and same frequency.
Mixing of RDIMMs (registered), UDIMMs (unbuffered) or LRDIMMs is not allowed.

• Do not install both ECC and Non-ECC memory modules on the same motherboard.

• Using DDR3 DIMMs with different operating frequencies is not allowed. All channels
in a system will run at the lowest common frequency.

• Make sure the Memory Mode in the BIOS is set to Independent, which makes all RAM available to the OS.

Since you've already replaced your Motherboard, and experienced the same exact problem, we can probably rule-out the motherboard as the cause.
And you've stated that you've tested all modules separately, so we can also rule-out defective RAM sticks.
I highly suggest you check the model # on each of the RAM modules, and check they are all exactly the same. Sometimes they differ by a single character, which makes all the difference.
As a past server builder, I can tell you I've had my fare share of RAM mismatches, and only by inspecting them by model #, I could assure proper compatibility.

Good luck with your new PC system, I'm sure it's super-fast and rewarding (once everything works as expected).

[Mark Toia]

PC's are so scary

[Jeff Kilgroe]

PC's are so scary

What's scary about PC's is that many people build their own to "save money". If building a PC saves you money, then you obviously have a lot of free time on your hands....

I must have been really tired when making my last couple posts to this thread, I didn't notice when I read the first post that Luke had already swapped motherboards and still has the same memory issue. It's almost guaranteed to be incompatible or mis-matched RAM. I would double check that the RAM specs are indeed correct for the board. At 128GB capacity, they must be ECC Registered DIMMs, if they're non-ECC, you can also run into issues that may account for the issues you have with all the slots populated. Technically the board and chipset do support 128GB (or less) for non-ECC UDIMMs, but not if dispersed across all 16 sockets. Not as far as I'm aware, anyway. All things considered, you don't want to be running non-ECC, unbuffered RAM in these systems anyway.

According to your description of which DIMM sockets don't work, that's CPU1-D2 and CPU1-D1. Sounds to me like the RAM is mismatched or out of spec where the motherboard won't enable quad-channel mode. There is no quad-channel ability possible with those DIMMs removed.

One other possibility... Did you buy the DIMMs with the thermal sensor? I'm thinking of another possibility here, not sure if it will mean anything, but just thinking out loud in a way...

[Luke Boyce]

This is the RAM I purchased: http://www.newegg.com/Product/Produc...82E16820239015

It is 1.5v Registered ECC RAM. When I get back in the office I'll have to check them to look for any irregularity.