RAM: Questions Answered
In our continuing Customer Questions Answered series, we're looking at RAM.
RAM is an incredibly nuanced hardware component. When you consider the variables such as Load-Reduced, ECC vs. non-ECC, and Registered RAM, it can become overwhelming to make sure you are taking everything into account. (And we didn't mention DDR3 vs. DDR4, Low Voltage (LV) RAM, and Rankings!)
While we have a wealth of resources in our blog on RAM (scroll to the bottom for a list of linked articles), we thought it would be helpful to share some of the common – as well as interesting – questions that our customers have asked us. You may be surprised how some obscure components can become incompatible when used in conjunction with the incorrect RAM!
Hello, Mike, and I hope all is well. I have a TechMikeNY R730 that I upgraded the RAM to modules with a speed of 2666MHz. The problem is I do not see the added speed and the machine still shows the RAM speed at 2400. Where did I go wrong?
Hi, thanks for reaching out.
Looking at your order history, your R730 has 2x Intel Xeon E5 v4 chips. The issue is the v4 processor RAM bus runs at a max of 2400. The 2666 RAM you installed will not provide any additional benefit. Unfortunately, processors that can run RAM at the full 2666MHz speed are incompatible with the Dell 13th Gen servers. That said, this RAM will run at the full 2666MHz speed in a Dell 14th Gen server. You can harvest out that RAM when you are ready to upgrade.
Hi Mike. I'm getting a memory error message on my R620, specifically on slot B8. Is the RAM faulty? Am I still under warranty to get a replacement module?
Hi, thanks for reaching out. Please attempt a couple of troubleshooting steps.
- First, remove the RAM and clean the slot with some compressed air.
- Then reseat the module and see if the error persists.
- If so, please move the module to slot B12 (it should be the next one over) and see if the error follows. We want to confirm that it is the module, not the slot.
Yes, your R620 is still under our one-year warranty. If the RAM module is faulty, we will ship a new one; if the slot is damaged, we can replace the server.
RESOLUTION FOOTNOTE: The issue for this customer was a bad RAM module. We shipped one out the next day, and he is up and running.
Hi Mike and team. We ordered an R820 from you guys, and it doesn't have the correct amount of RAM. My techs opened the service lid and found that it only has modules for 500GB instead of the 1TB of RAM we ordered. How can we address this?
The R820 (along with the R830) servers are quad-processor systems. There is a top mezzanine board with CPU 2 and 4 (and its RAM). Underneath that mezzanine board is the primary system board, where you can see CPU 1 and 2 with the associated RAM. The design of this model server tightly packs the components into a 2U frame, and when opening the service lid, it creates the illusion that the top mezzanine board is the main system board.
I confirmed with our order management tool that this configuration has all four processors installed. Likely, your techs aren't seeing/counting the RAM underneath the top mezzanine.
You can confirm the RAM installed by booting the system and viewing the specs.
Mike, I'm using your Configure-to-Order tool and building an R930 server. I don't see DDR3 RAM in the dropdown, but I am under the impression that the E7 V4 Intel Xeon processor family supports DDR3 RAM. Why isn't the DDR3 option available for this R930?
Hi, and thanks for your question. While the processor you list supports DDR3 RAM, the user manual of the Dell PowerEdge R930 does NOT support RRD3 RAM. This is true for all 13th Gen Dell servers – they are DDR4 RAM only, and the pins for a DDR3 module would not align with a DDR4 system board.
Hey Mike. We need your expert guidance! We installed an RMS-200* card into an R730xd. Before the server can complete its data migration, it loses power and reboots after about 15 minutes. The error we are receiving on reboot is: "UEFI0067: A PCIe link training failure is observed in PCIe Slot 2, and the link is disabled."
Any ideas? Is the RMS-200 card faulty?
Hi. While it's still possible that the card is faulty, it is more likely that model RMS-200 is not compatible with the server because it only works with machines with DDR3 RAM (you can view the RMS-200 specs here). Since the R730xd only takes DDR4 RAM, it will not work.
As for options, you can procure an R720xd or try an RMS-300, which is compatible with DDR4 RAM (which is in your R730xd).
*TECH FOOTNOTE: The RMS-200 card is a PCIe NVRAM Accelerator. NVRAM is a type of RAM that retains stored data even if the power is off. These devices are generally used for faster boot times and quicker access to data. As you can see from the above "compatibility pitfall," there are unique components and variables that you have to consider!
Our RAM Resources
For the past couple of years, we've been writing extensively on our blog about RAM.
- RAM Compatibility Checklist. What to Look for To Ensure You Are Getting Compatible RAM
- How To Upgrade RAM in a Server
- How to Identify Laptop & Desktop RAM Specs by the Module
- How to Identify Server & Workstation RAM Specs by the Module
Please feel free to use these resources!
Do you have any other questions on RAM compatibility, troubleshooting, or anything else? Leave it in the comments, and we will answer!
Don’t forget that some (all?) machines will not POST with a mix of LR-dimms and R-dimms! I have an r730 and an extra 256GB of ram that does not make it happy.
I just have it here, on my desk. Taunting me.
Great question. This probably will not cause any issues with virtualization, as ultimately, the CPUs will use RAM that hasn’t been allocated yet. Admins decide how much of the CPU and memory to use, but the server should have no trouble selecting where to use it from.
we have several Dell r430 of which we have purchased there, this is a general question about the R430, my question revolves around the fact the ram per cpu is different CPU1 has 8 dimms and CPU2 has 4 dimms, we are populating all slots with 32g chips so does this create some sort of issue in a virtualized environment considering how much ram per cpu as assigned to VM’s.