Here is what the seller claimed:
Sequential is rated at 3.6GB/s read and 3.2GB/s write sequential. In Dual Port Mode which the drive is in, speeds are at 1.8GB/s read and 1.6GB/s write sequential.
If we're looking at just the drive, dual ports provide increased performance. Even with a drive that allows a single port configuration (as opposed to simply not connecting the second port) the speed will likely always be limited compared to dual port operation. edit: This appears to be true only for dual port SAS drives and not NVME which allow x4 on a single port for full throughput when in single port mode.
The decrease in speed you see is caused by using an adapter that only supports one of the 2 ports. Which makes sense as you're essentially limiting output to a single port on a dual port drive. In which case, it would seem the adapter is the problem.
All consumer level adapters should pick up port A, and there are adapters to pick up just Port B instead. However, I know of no adapters that pick up both ports aside from expensive multiport capable backplanes and nvme switches.
Someone experienced with these drives might be of better assistance than me. If you can't put the drive into single port mode and you can't live with 1.8GB/s, current options are either spend the big bucks for a multiport setup or look for single port drives with the throughput you need.
Edit:
This may be of limited help to you but I found it interesting. A cursory search leads me to believe that enabling/disabling dual port is actually dictated by the host controller through the DualPortEn# pin on the connector.
For non-enterprise U.2 (SFF-8639) connectors DualPortEn# is on E25. Page 22 of
https://www.intel.com/content/dam/s...600_Series_PCIe_Product_Specification-004.pdf
DualPortEn# pin should be left un-connected or un-driven by the system to enable single port operation with all 4 lanes. If un-connected, P3600 will pull it high. However, if the pin is asserted by the system (driven low by storage backplane), then P3600 will be configured as x2 lanes.
If the drive follows the Enterprise PCIe (SFF-8639) connector standard then DualPortEn# is pin
E39. From SSD Spec page 14
http://www.ssdformfactor.org/docs/SSD_Form_Factor_Version1_00.pdf
2.4 DualPortEn# - PCIe Dual Port Enable
The Enterprise PCIe SSD can be configured to train as either a single x4 controller or dual x2. The
mechanism for enabling dual port operation is DualPortEn# (pin E39). DualPortEn# is pulled high internal
to the Enterprise PCIe SSD. If DualPortEn# is left open then the PCIe is configured as Single x4 port. If
DualPortEn is pulled low by the system (driven low or grounded by backplane), then the Enterprise PCIe
SSD has dual Ports enabled. DualPortEn# is part of the drive type determination shown in Figure 6.
Secondary resource asserting the same (page 5):
https://www.flashmemorysummit.com/E...oceedings/2018/20180809_TEST-301A-1_Nadig.pdf
If I'm reading this correctly, covering this pin would leave the drive in single x4 mode similar to how everyone tapes the B pins on perc h310 to get around the boot interrupt. I don't have a U.2 drive to verify the connector aligns with the diagram or the feasibility of covering the pin. Maybe someone else wants to give this a try?
FYI according to samsung this holds true for NGSFF ssd as well, except DualPortEn# on pin 46
https://www.samsung.com/semiconductor/global.semi.static/Whitepaper_Samsung_NGSFF_SSD_1809.pdf