PCIE Bifurcation on Z170 chipset boards?

[BonzaiDuck]

Sometime within the last hour, I stumbled across this product or product-configuration in a casual web-search:

SuperMicro AOC-SLG3-2M2 PCIe NVMe Dual M.2 x8 Gen 3 Adapter Card with 960GB (2x480GB) Samsung PM953 Enterprise SSD

Motherboard compatibility information varies, and the board can be bought without NVME M.2 paired drives for under $50. As an x8 PCIE card, it requires PCIE bifurcation for x4/x4 use. There is a default RAID0 configuration for the board.

As to the bifurcation issue, promotional materials mention certain models of Dell workstations, or motherboards coded as "X10" or "X11" motherboards. As can be gleaned from the materials, any motherboard providing the PCIE bifurcation feature would work with the card.

In respect to a choice of operating systems, some requirement summary only referred to "Windows 2012 R2" [server] and some various released of Linux, etc. Other materials, possibly those linked above, refer to board operation in Windows 10 and Windows 7.

I cannot find anything so far about "PCIE bifurcation" features on a Z170 motherboard. Maybe I missed something in scanning the material -- can't say at the moment.

Does anyone know with certainty about this feature such that they could confirm or deny the feature on the Z170 boards (mine is Sabertooth Z170 S)? Why do I recall that slot specifications for mine or other boards refer to "X8/X4/X4" operation per having 1, 2 or 3 slots filled? Does this specification have anything to do with "PCIE bifurcation?"

Certainly someone knows something about this.

 

[XavierMace]

Last I heard for Z170 your only options for a board with bifurcation is the ASRock Z170 Fatal1ty ITX or Gigabyte GA-Z170N-WIFI (Rev 2.0 only). .

 

[BonzaiDuck]

Last I heard for Z170 your only options for a board with bifurcation is the ASRock Z170 Fatal1ty ITX or Gigabyte GA-Z170N-WIFI (Rev 2.0 only). .

Actually, those two were the indications I had before I posted here.

But it almost sounds to me as if this is a matter of BIOS implementation, since we're talking about the same chipset. Or is it? What do you think? Curious minds . . . want to know . . .

I can only say it would be nice if I could use that SuperMicro card. Would I give up using PrimoCache if it were an option? Probably not!

 

[XavierMace]

Yes, it's generally matter of BIOS implementation.

 

[BonzaiDuck]

Yes, it's generally matter of BIOS implementation.

Nice that you concur. Second opinions are always reassuring.

Assuming that it is just a matter of BIOS revision, I could hope for a positive development, after the exchange I had with ASUS tech support.

The rep pushed it to "the highest level" of their expertise, with two successive e-mail reassurances -- for which I'm still awaiting something more substantive.

I could hope that the heads-up would lead them to revise the Sabertooth Z170 BIOS so that I eventually get the PCIE bifurcation feature.

Hope springs eternal . . . . or something like that . . .

 

[thecoolnessrune]

There's also very specific noise and timing issues presented in PCI-E bifurcation that requires design consideration directly in the motherboard design. That's not to say it's not impossible to have on lots of boards, just that there's a lot of boards that would likely fail validation if attempted, which I say because it's a bit more than just a matter of BIOS Revision. It would need to be validated, and then if successful yeah, it's just a BIOS revision.

 

[BonzaiDuck]

There's also very specific noise and timing issues presented in PCI-E bifurcation that requires design consideration directly in the motherboard design. That's not to say it's not impossible to have on lots of boards, just that there's a lot of boards that would likely fail validation if attempted, which I say because it's a bit more than just a matter of BIOS Revision. It would need to be validated, and then if successful yeah, it's just a BIOS revision.

The ASUS folks, still promising to "get back," have yet to say anything one way or the other.

My BIOS revision was released three months after I built the machine, or in December 2016. There have been two revisions since -- the latest released this month of 2018.

The release notes on the two BIOS revisions cite "fixed RAID card issue" and "fixed Samsung M.2" or "Sandisk M.2" issues.

So I sent them another e-mail to get clarification. I'm just not going to update the BIOS unless it gives me what I'm looking for. You know the routine: make note of all custom settings [and overclock settings]; choose update method; restore to default speed and settings; update BIOS; restore overclock profiles etc.

And of course, why do it until (a) they confirm and (b) I've gone forward to buy the SuperMicro card? And why buy the card until they confirm?

 

[BonzaiDuck]

Wey-ull!! I'm building a system this year (2021) with a dated ASUS Z170-WS motherboard to potentially use three NVME drives: two SK Hynix 1TB and a single Samsung 970 EVO.

So I decided to revisit the issue which sparked this thread in the first place.

There is an ASUS list of motherboards supporting their Hyper M.2 PCIE card series -- devices that accommodate two or more NVME drives. One could assume that there are no less nor more limitations on any other multi-NVME card than on the ASUS products.

In any event, the Hyper M.2 cards are only supported on ASUS boards using the X299 chipset or later.

On the plus side for me, at least the workstation board has four PCIE x16 slots -- two of which will run as x16 if populated by x16 cards. I wasn't planning to do anything more with those slots in any event, but for my dGPU graphics card and storage.

I'd even forgotten that I posted this thread back in 2018. It turned up on one of the morning's web-searches, when I chose to re-investigate multi-NVME PCIE cards.