SM X11DPi-NT, Xeon Gold5115, Kingston 2666 stable at 2133

[Kiss92]

Hi all,

I have almost the same problem than this thread : https://forums.servethehome.com/ind...l-i-kingston-2666-only-running-at-2400.21778/

The Config :

- X11DPi-NT
- 2 x XEON GOLD 5115
- 4x KSM26RD8/16HAI (Kingston 16GB Reg ECC 2666) recommended on Kingston configurator website
- Adaptec SmartRaid 3154-16i
- Chenbro RM23812 E3RP8

So Xeon Gold 5115 accepts 2400Mhz Memory but if i set the speed at 2400Mhz in the Bios, system Freeze (need to hard reboot) when i run some memory benchmark for exemple. I have to set at 2133Mhz in the BIOS so the system is stable.

So wtf? This Xeon very sensitive lol?

If someone has an idea im aware

Thank you

 

[MiniKnight]

It should work at 2400. Kingston memory isn't used by tier 1 manufacturers for a reason. Supermicro doesn't validate Kingston because their modules have issues. Micron, Samsung, Hynix.

 

[Kiss92]

However this modules has Hynix Dram chip?

By the way, if i set auto speed, system Freeze immediatly ^^

 

[MiniKnight]

Kingston has a lot of issues with Supermicro.

 

[William]

I used to battle memory issues long time ago, mostly for single socket enthusiast board.
I learned the hard way, purchase kits, find out they do not work.
Now I only use recommended sticks from the manufactures website.

Here is the one for your board.
Product Resources | Support - Super Micro Computer, Inc.

These things also take BIOS support, without that its a crap shoot.

If everything is new I used recommended parts (memory etc.).
If I have used stuff laying around and it works... Win.

This is not a Kingston issue, its all brands really. If the manufacture has not qualified the kits there is no BIOS support.

Also... check the manual and see what it says about the memory population rules, what happens to RAM speed when you have all slots populated, it usually drops on these boards.

 

[kapone]

A few things to check:

- Does the RAM even have a 2400mhz JEDEC profile?
- If so, what voltage does it require at that speed? The board may not be willing to give that voltage at that speed.
- Any "CPU/Memory bridge settings/ratios that you can play with?

 

[Kiss92]

Thanks for your answers,

I will buy new set of Samsung memory, probably :
64GB Samsung M386A8K40BM1 DDR4-2400 regECC

It's a very strange, very weird situation :

I am running 1 pass MemTest86 just for see quickly if a stick is broken. Because freeze is random, after the Bios Post when Windows is starting, hop, freeze! It's crazy first time i see that.

During i write, i see that MemTest is blocked at test2. I will try to check if it's not the memory slot who are defective.

 

[Kiss92]

Edit, let see the picture :

Actually i keep doing Memtest, benchmark for see if system is stable.

So, Memory Controller, bad slots ? It seems RMA request

 

[MiniKnight]

I still think this is RAM as the issue not the motherboard

 

[William]

I still think this is RAM as the issue not the motherboard

Agreed.
These kits are not specified for this board, memtest will tell you nothing, except these memory sticks are not qualified for this board.

You certainly can RMA the board but you are looking at possibly the same issues you are having now.

email Kingston and ask them if these kits will work on this motherboard.

 

[Kiss92]

You can see here, Kingston recommendation for this Motherboard : Recherche de mémoires

It's single not in kit.

You probably right but why all memory stick works everywere else slots without trouble?

 

[Kiss92]

Some progress :

So, i have repopulated all DIMM like as at the beginning. I have note the Freeze as soon as Windows is starting.

But this time i have retry to swap the CPU with all dimm populated, and .. it works, unbelievable.

And in CPU-Z, i see Memory in Dual Channel.

Next Step ? : ok i will buy some new memory (samsung) and replace + Re-swap CPU.

But wtf, it is possible that one CPU has something wrong ? and why no freeze ?

 

[alex_stief]

Not sure if I completely understood the issues at hand, but I think you should follow the recommended way (see manual) of populating DIMMs. 4 DIMMs all on one side of the CPU is not ideal. At least use blue slots first, they all belong to an independent memory channel. The black slots do not and were just added for convenience of users who need more than 12 DIMMs total can trade in some performance and compatibility for it. The manual should have more precise instructions on which slots to populate first.
Forget about CPU-Z for anything beyond single-CPU quad-channel memory platforms. It usually just shows bogus values.

 

[Kiss92]

Alex,

I have x4 16GB DiMM (specified in my first post) which are connected like recommanded : x2 for CPU1 (A1-D1) and x2 for CPU2 (A1-D1) , blue slot near the black. If you place wrong, you have a message in Bios.

Still stable since the last message with the x4 DIMM. All app's work fine for the moment.

 

[Kiss92]

For the records, the problem came from the Motherboard, managed and solved by RMA. Now all is working fine (with Kingston memory )

This is the answer of Supermicro for initiate RMA :

"I couldn't tell with certainty what's happening here, but if I had to guess, there is perhaps a problem with the primary socket connecting the DIMM1D DIMM slot. It could be that in combination with a CPU with a slightly worse memory controller (manufactoring differences between CPUs), if a pin in the socket has slightly less pressure, and thus a slightly higher resistance, might just be the point of working or not working.

It is presumably a board related issue in any case, you can request RMA for this and see whether they can solve this. But if you do, make sure to clearly describe the issue and you should probably include the 2 CPU's in the RMA (also indicate that in the request) as it's possible our RMA tests with a different CPU and then won't see the issue. You may also, as extra, refer to this case."