Community
cancel
Showing results for 
Search instead for 
Did you mean: 
wwelc1
New Contributor II
2,168 Views

Optane Memory VROC fails

Using Supermicro X11SPM-TF, scalable Xeon Silver 4110, and VROCSTNMOD dongle, I have been trying VROC RAID0 with four Optane Memory 32GB on ASUS Hyper M.2 X16 card. Neither Intel nor Supermicro nor ASUS supports such a use, I know, but Optane Memory is all but an SSD of Intel brand.

In the BIOS, I set PCIe bifurcation to x4x4x4x4, enable VMD and VROC with VMD Technology (5.3.0.1052), but only one of the drives is listed under VMD Controllers. This one can be set as a spare of VROC. Other three are shown as NVMe drives and accessible as such in Windows 10. Thus, at least PCIe bifurcation and NVMe driver must be working.

Swapping positions of Optane Memories shows that the one furthest away from the PCIe slot is recognised as VMD, therefore it is not a hardware problem of Optane Memories.

The VROC standard dongle makes no difference. One Optane Memory can be recognised as VMD whether or not the dongle is there.

I am not sure if the problem lies with ASUS Hyper M.2 X16 card or Supermicro X11SPM-TF. So, is there anything I can do other than contacting ASUS, Supermicro, and buying expensive Intel SSDs?

Thanx

0 Kudos
6 Replies
idata
Community Manager
205 Views

Hello Mr. Sorrow.

 

We have received your message, we appreciate sharing your setup. We would like to point out a few things:

 

 

-According to our documentation Intel® Virtual RAID on CPU (Intel® VROC) is compatible only with Professional and Data Centers SSDS. (We are currently investigating if some consumer SSDS are also compatible).

 

 

-Concerning the Intel® Optane™ memory module, we would like to inform you that is not supported and may not work as intended with VROC, since the intended use of the Optane its for cache acceleration purposes, not for storage(although you are trying a RAID 0). In order to create a RAID 0 Array you will need a minimum of 2 SSD drives.

 

 

-Concerning your doubt about only one PCIE lane being recognized: The best way is to confirm with https://webpr3.supermicro.com/SupportPortal/IssueMgt/FileIssue?location=US Supermicro support directly, since from our experience with other customers using motherboards from the same family, they have been told by Supermicro that a customizable BIOS is needed and by default only one single PCIe slot can be addressed via VROC. We strongly recommend you to confirm this fact with Supermicro and let us know your experience as well.

 

- You proceed well with the basic configuration ( PCIe bifurcation to x4x4x4x4, enable VMD and VROC with VMD Technology)

 

 

Finally, We will like to share some documentation in case you want to consult:

 

 

https://www.intel.com/content/www/us/en/architecture-and-technology/optane-memory-faq.html -Optane FAQ:

 

https://www.intel.com/content/www/us/en/support/articles/000024550/memory-and-storage.html -VROC FAQ

 

We will keep a followup on your thread in order to dive in deeper about possible suggestions upon your setup. Please share any information that you consider pertinent to this case.

 

Best Regards,

 

Andrés.
idata
Community Manager
205 Views

Hello Mr. Sorrow.

 

We are following up on your case. Please let us know if you need anything else.

 

Regards

 

Andrés.
wwelc1
New Contributor II
205 Views

Hi Andrés

Thank you for your continued interest. I have been communicating with Supermicro support. They suggested me using their add-on card AOC-SLG3-4E4T, that is a quad m.2 card with a PLX PCIe switch. Under this setting, PCIe bifurcation is carried out by the add-on card and not by the m/b. You need to set BIOS PCIe bifurcation to Auto mode and not to x4x4x4x4, since the latter will cause confusion. Supermicro BIOS connects only one PCIe lane to VMD controller, and that is why only one drive on the ASUS card is recognised under VMD controller. ASUS BIOS can connect four m/b-bifurcated lanes to VMD.

Using BIOS PCIe bifurcation x4x4x4x4 mode, Windows can happily recognise four drives on ASUS card as NVMe devices, and you can set up sowtware RAID.

If only someone could mod Supermicro BIOS and ASUS BIOS together.

idata
Community Manager
205 Views

Hello Mr. Sorrow.

 

Thank you very much for the updates and your response. We will take note in what Supermicro is suggesting, and we appreciate for sharing their feedback. We understand the limitations on your setup, but at the same time we are available for any doubt in the process you might have. Let us know if you need anything else.

 

Best Regards

 

Andrés C.
wwelc1
New Contributor II
205 Views

Hi Andrés,

I thought someone in Intel might have a better idea why Supermicro m/b fails to connect bifurcated drives to Volume Management Device controller. orz

The only difference between ASUS Hyper M.2 x16 and AOC-SLG3-4E4T is how PCIe slot is splitted. Supermicro m/b is of cource capable of bifurcation but somehow Supermicro has left that simple system incompatible with VMD. Instead, they made a system that requires PLX PCIe switch for the bifurcated drives to be recognised by VMD controller. That is puzzling.

 

I think it is a bug in the VMD driver, because X299 family that spear VMD has no problem with ASUS Hyper M.2 x16 in constructing VROC.

Curiously, Supermicro m/b PCIe bifurcation works flawlessly with ASUS Hyper M.2 x16, that is, I can assemble Windows software RAID0 using four Optane drives on ASUS Hyper M.2 x16, with VMD "DISABLED".

You experience a problem only when you try to connect bifurcated slots to VMD controller. Otherwise, there is absolutely no hardware issue. If it is a bug in the VMD driver, why not avoid it altogether? I do not need VMD since I do not want to light up LED of drives.

Please give me VROC without VMD just like in X299 system. Or please give me a choice to include/exclude VMD in VROC.

Thanx

idata
Community Manager
205 Views

Hello Mr.Sorrow,

 

Once again we appreciate your time and concern for sharing your experience and data. We will share this with the engineering team and let you know its resolution.

 

Best Regards,

 

Andrés.
Reply