Error in server memory update

from the message, riser A dimm 2 is having a problem/may be bad. my suggestion would be to reseat or swap that dimm with a new one. You could also install 1 or 2 dimm per proc and see if the bootup is correct, then add dimms to each proc and see where the message comes back. 

Rey
#Iwork4Dell

Hi,

Yea, as suggested from Rey, you can try either to swap or reseat the memory DIMM to check if the error occurs on the slot or the memory. 

Also, could you confirm the memory population? You mentioned you have 4 risers with 4 CPU, which means you have Riser A,C,E,G. 16GB and 8GB are in which slots? According to the installation guide, 16GB should be in Slot 1,2,5,6 and 8GB be in 3,4,7,8. If the 16GB are in DIMM 1 and 2 slot, please swap with other riser DIMM 1 and 2. 

Hello everyone. I swapped the memory between slots on the same riser and also placed it on another riser. Whenever memory was placed in a new appropriate slot on the riser, the server detected the problem at boot. When I replaced the memory with another one, the problem was solved. However, the problem memory was new and therefore had never been used before.

According to the description on the server, I used slots 1, 2, 3 and 4 for 16 GB memories and I used slots 5, 6, 7 and 8 for 8 GB memories. See description:

After detecting the faulty memory, I have a new challenge. I used the server described precisely to test the 16 GB memories purchased. I have four Dell PowerEdge R910 servers, with the same configuration described above: 

- It has 4 Intel Xeon E7-4850 processors;
- 4 risers;
- 4 x 8 GB memory modules per riser. Therefore, there is a total of 128 GB of RAM.

Recently, according to the money we had, we bought 4 more risers for each server and 8 x 16 GB of RAM for each server. The challenge now is how to distribute memory between servers in a satisfactory way without losing performance.

Initially I thought about keeping the configuration of the 4 old risers unchanged and adding 2 x 16 GB of RAM to each new riser. This way, each server would have:

4 old risers x (4 x 8 GB) = 4 x 32 = 128
4 new risers x (2 x 16 GB) = 4 x 32 = 128
Total: 256 GB

However, after studying a little more due to the memory error I described, I understood that the solution I thought of may not have good performance. Can you guys guide me on this?

Hi,

Based on the memory installation guideline page 95: https://dl.dell.com/manuals/all-products/esuprt_ser_stor_net/esuprt_poweredge/poweredge-r910_owner's-manual_en-us.pdf the slot installation seems to be incorrect. As mentioned, 16GB should be in slot 1,2,5,6 instead of 1,2,3,4.

If you have installed 4 more risers, which make total of 8 risers, you can refer to the guide for installation allocation. All risers' population must be identical. You will not be able to insert addition 4 risers differently than the existing 4 risers. Do let us know how many total memories do you have and intend to install. We'll try to figure out what's best on how to install. 

@DELL-Joey C The image above is of the server cover. I understand that larger memories should be placed in the white latch slots. In each riser I placed 4 16GB memories (slots 1, 2, 3 and 4) and kept the 4 8GB memories (slots 5, 6, 7 and 8).

Calculation: (4 x 8GB) + (4 x 16GB) = 32 GB + 64 GB = 96 GB per Riser. 4 risers x 96 GB = 384 GB of RAM on the server.

When starting the server, the total amount of recognized memory matches the calculation. Look:

Since the total amount of memory is recognized, I am assuming that the memories were placed in the correct slots. 

Hi,

Alright, since the memory population on the existing riser is showing expected memory total size, you may need to populate the exact same on the new risers, with reference of the user guide page 92: General Memory Module Installation Guidelines > All populated memory risers must have identical configurations.

@DELL-Joey C I'm not sure if it was clear in the previous posts. We have 4 servers organized in pairs within pools. The challenge now is to divide the 16 new risers and the 32 newly acquired 16 GB modules among the 4 servers. I didn't know that each riser should have the same amount of memory.  Consider the following cases to try to solve the challenge.

Case 1: In the first case, I keep the 8 GB memory modules and add two more 16 GB modules.

Case 2: In the second case, I move the 8 GB memory modules to the second server and install the 16 modules of 16 GB in the first server.

In both cases presented above, I achieve 256 GB in each server. 

In the third case, 4 new risers are added to each server. We divide the existing 8 GB memory modules among the risers. In this model, we assume a new purchase of 16 GB memory modules. I consider this case an evolution of the two previous cases.

Question: In the three cases above, I put the 16 GB memories in the white lock slots. Now, when I don't have 16 GB memories in some of these slots, should they be left empty or should I put 8 GB memories?

Hello,

The key here is all risers must be identically populated.

If you don’t have enough 8G DIMM to populate all risers identically then leave it out of all.

You can try with the different configurations and see if you get memory errors on post and then adjust your population.

It shouldn't hurt anything to experiment.