SMCI NVMe Rear Hot Swap Drive Kit - Integration Options?

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

svtkobra7

Active Member
Jan 2, 2017
362
87
28
[ See "Current Status" / "Potential Solution" - Succinctly: Given info I've gathered, I have concerns over whether I'll be able to connect the NVME Kit to my X9 board and would appreciate you input! Thx in advance :) ]

Objective
: Facilitate integration of additional NVMe in 2.5" format by installing MCP-220-82619-0N - NVMe Rear 2 x 2.5 Hot Swap Kit (hereinafter "NVMe Kit"). [ Reference the Spoiler - "Pics" - if interested in what the NVMe Kit looks like / includes ]

Integration: After I had sourced 2 x NVMe Kits, I started to consider how I would integrate into my chassis (here) and floated the idea of NVMe Kit => OCuLink Cable => PCIe OCuLink Adapter. My thinking = plenty of available PCIe and adding 2x per server is still less than half the cost of a AOC-SLG3-2E4.

Reported Connectivity Issues:
  • @Aestr was kind enough to report (here) that NVMe Kit => OCuLink to MiniSAS HD Cable => AOC-SLG3-2E4 = FAIL. He also tested a PCIe OCuLink Adapter = FAIL. Many thanks to him for his assistance! (I don't think such a solution will work as the NVMe Kit swap kit must have NVMe0 => NVMe0 + NVMe1 => NVMe1)
  • Additionally, SMCI tech support noted the NVMe Kit is only validated for a single system - 5029P-E1CTR12L - which is odd since the product description from resellers = "Supermicro 2.5x2 NVMe Drive Kit for 216B/826B/417B/846X/847B,RoHS" [ Reference the Spoiler - "SMCI Tech Support Replies" ]
Current Status:
  • Will integration be successful, and if so, how?
  • AFAIK, NVMe Kit => OCuLink to MiniSAS HD Cable => AOC-SLG3-2E4 should work ...
  • In absence of clarity here, which I'm hoping members much smarter than myself can add, I suppose my only option is to start buying stuff and then paying restock fees or reselling on ebay.
But I did come up with a bit of a hack that may work ...

Potential Solution
- Remove Backplane from existing non-NVME Rear Kit (12G version) [ Referece Figure 1 - "Current Backplane" ]
+ 1 x $5 Molex 4-pin Y-Cable (connect to existing Molex 4-pin cabled to existing non-NVME Rear Kit)
+ 2 x $27 = $54 Ableconn MiniSAS HD SFF-8643 PCIe Adapter
+ 2 x $15 = $30 Supermicro CBL-SAST-0957
= $89 Total [ Reference Figure 2 - "Potential Solution Illustration" ]

Pros = Enhanced drive cooling (backplane is highly restrictive to airflow). Economical.
Cons = Cabling mess. Not hot swap (no biggie). Loss of "stability" - with a "free floating" drive connector (not one secured to the chassis), drives are only held in place by the caddy hinge which locks onto the top of the drive cage.

To take it a step further, if one were able to figure out a way to secure the connector to the existing mount points for the backplane (via custom bracket or otherwise), you would be able to retain hot swap.

1 - Current Backplane--- 2 - Potential Solution Illustration

#1
Currently, MCP-220-82619-0N is only for our complete system which our lab has validated.
We do not recommend using it as building blocks.
#2
We have tested 5029P-E1CTR12L with MCP-220-82619-0N for rear NVMe drives.
It is recommended to order complete system with MCP-220-82619-0N.
 
  • Like
Reactions: croakz

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
Have you disproved the compatibility with the T card? That would enable OcuLink to OcuLink connection which might be less error prone
 

svtkobra7

Active Member
Jan 2, 2017
362
87
28
No I haven't proven any AOC to work or not (I wish I knew for certain my board couldn't bifurcate as that might provide direction). I suppose I need to investigate if I can purchase and return if they don't work?

I agree that taking an adapter "transform" out of the equation may assist; however, the T cards don't have a switch, which means I would be relying on unproven bifurcation (or not).

So, I think the only OCuLink AOC with PLX may be a better play = AOC-SLG3-4E2P but those are $200+ (and keep in mind I'm doing this 2x over).

Code:
MPN            Description                   Switch      Connectors
AOC-SLG3-2E4   6.4GB/s Gen-3 Int NVMe HBA    PLX PE8718  2 x SFF 8643
AOC-SLG3-2E4R  6.4GB/s Gen-3 Int NVMe HBA    No Switch   2 x SFF 8643
AOC-SLG3-2E4T  6.4GB/s Gen-3 Int NVMe HBA    No Switch   2 x OCuLink
AOC-SLG3-4E2P  6.4GB/s Gen-3 Int NVMe HBA    PLX PE9733  4 x OCuLink
AOC-SLG3-4E4R  12.8GB/s Gen-3 Int NVMe HBA   No Switch   4 x SFF 8643
AOC-SLG3-4E4T  4 Port NVMe PCIE X16 Retimer  No Switch   4 x OCuLink
 

svtkobra7

Active Member
Jan 2, 2017
362
87
28
So do you have enough x4 slots if you cant bifurcate?
Indeed ( total on board = 1 x PCIe3.0 x16 + 6 x PCI-E 3.0 x8 ), thus my proposal to "waste" 2 slots on that "economical" solution by removing the backplane from the existing 12G Rear Kit (essentially just turning it into a drive cage) and burn two of the x8 slots on adapters (1 per drive). The one thing I hate about that "solution" is that I lose hot swap which doesn't bother me so much, but it being a bit of a "hack" does (OCD). On the flipside, I can see those drives starving for airflow, so there is the advantage of more airflow with excess cabling / U.2 connectors properly cable managed. [not sure if this is where you were taking your question]

Or perhaps maybe you are suggesting that I could go the AOC-SLG3-2E4R / AOC-SLG3-2E4T route? Meaning if I can't go x4x4, then just buy another adapter, such that I have one per card. But I think there is something wonky about that NVMe Rear Kit and the drives have to be connected in the proper order (suggesting this due to @Aestr 's connectivity experience), i.e. NVMe0 => NVMe0 + NVMe1 => NVMe1 to an NVMe backplane or AOC (if you look closely at the pics in the first thread - second spoiler - you can see the backplane notes "0" vs "1").​

The rockstars out there would 3D print a "bracket" with OD = existing backplane + add holes for airflow and 3 mounting screws (to drive cage)+ figure out a way to secure U.2 female / U.2 male connector to it to create a clean solution, and retain hot swap. I lack that skill set. :(

 

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
Or perhaps maybe you are suggesting that I could go the AOC-SLG3-2E4R / AOC-SLG3-2E4T route? Meaning if I can't go x4x4, then just buy another adapter, such that I have one per card. But I think there is something wonky about that NVMe Rear Kit and the drives have to be connected in the proper order (suggesting this due to @Aestr 's connectivity experience), i.e. NVMe0 => NVMe0 + NVMe1 => NVMe1 to an NVMe backplane or AOC (if you look closely at the pics in the first thread - second spoiler - you can see the backplane notes "0" vs "1").​
Yes thats what i was suggesting as it seemed the cleanest.
I don't see where having a port labeled induces wonkyness though:p He just said that the 8463 cable was acting up which might be due to the fact that it might be directional, i thought i heard that before.
/me likes OcuLink, but no idea whether that card works with you board.
 

svtkobra7

Active Member
Jan 2, 2017
362
87
28
What's up my swizzle ... Let me help you out with that and de-swizzle some infos for you ...
Yes thats what i was suggesting as it seemed the cleanest.
I don't see where having a port labeled induces wonkyness though:p
  • Seems like a reasonable starting point (fingers crossed).
  • By wonky, I meant that I don't see a reason why he couldn't connect the NVMe rear kit direct to the AOC, rather he only obtained connectivity via connection to the NVMe backplane. Why? (I don't know)
  • But the port numbering does matter unless my brain is swizzled ... (out of scope scenarios)
With one end of the cable connected to CH1 of the add-on card, connect the other end to NVMe#0 (JSM3) of the BPN-SAS3-116A-N2 backplane. With one end of the cable connected to CH2 of the add-on card, connect the other end to NVMe#1 (JSM4) of the BPN-SAS3-116A-N2 backplane. You will hear a click if the cable is connected properly. [ Reference AOC-SLG3-2E4 User Guide, page 3-4 ]

"x4 NVMe drive connect requires a special cable" "Cable re-orders lanes and effectively de-swizzles the adapter layout"
also note the "P0" & "P1" labeling [ Reference this deck on the 9400 Tri-Mode , page 24 ]

Disclaimer I have no idea what a U.2 enabler cable is / what is meant by de-swizzle (although I'm going to start using that word as it sounds cool).
He just said that the 8463 cable was acting up which might be due to the fact that it might be directional, i thought i heard that before.
/me likes OcuLink, but no idea whether that card works with you board.
  • Cables are definitely directional (as noted in my prior post) as noted in the AOC user manuals (images in spoiler)
  • SFF-8643 or SFF-8611, its all the same by the time it hits PCIe, right? (with no native OCuLink connectivity on X9 boards)

 

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
Yes the 9400 needs a special cable (i.e. not the regular one included in u2 nvme drives).

This does not apply to the 4T, or i have coincidentally bought the correct (b/c it was the cheapest) cable - i doubt that.
Also on the 4R, the regular U2 cables worked.

That leaves the backplane which uses oculink so not sure the 9400 reference matches or the same logic applies or if that is specific to the 8643 which are also supposed to carry SAS signals...
 

svtkobra7

Active Member
Jan 2, 2017
362
87
28
Yes the 9400 needs a special cable (i.e. not the regular one included in u2 nvme drives).

This does not apply to the 4T, or i have coincidentally bought the correct (b/c it was the cheapest) cable - i doubt that.
Also on the 4R, the regular U2 cables worked.

That leaves the backplane which uses oculink so not sure the 9400 reference matches or the same logic applies or if that is specific to the 8643 which are also supposed to carry SAS signals...
  • I think the user guide / intended use case = (1) Rear Kit => MiniSAS HD backplane + (2) AOC => MiniSAS HD backplane, where the "source" and "target" are more clear (to me at least).
  • When attempting to wire up a Rear Kit to AOC, any clue which is the "source" and "target"?


 

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
Not sure where the issue is, but from reading (without having tested) it seems pretty clear-

Rear Kit is the Backplane/Target, so gets a black MiniSASHD; the AOC is the Source and gets the green Oculink.
Now why the pic has both black and green and not a miniSASHD thats probably an oversight. As I understand it the upper right OcuLink connector would have to be a mini SAS HD connector to match the description