Transceiver, DAC, AOC FAQs

Speed & Form Factor

Q1: Are 1G SFP transceivers supported?

No. The SFP56 and QSFP56 ports on the 50G and 200G NICs of NetScaler MPX/SDX 9200 and 17000 appliances do not negotiate to 1G SGMII or 1000BASE-X (Ethernet over fiber). The minimum supported port speed is 10G.

Q2: Which 10G transceivers are supported?

NetScaler-qualified 10GbE SFP+ optics are supported on both the MPX/SDX 9200 and MPX/SDX 17000 platforms. This includes SR (multimode, OM3, up to 300m) and LR (single-mode, up to 10km) transceivers. For more information, see the Transceiver, DACs, AOCs’ matrix.

Q3: Which 25G transceivers are supported?

NetScaler-qualified 25GbE SFP28 optics are supported on both the MPX/SDX 9200 and MPX/SDX 17000 platforms. Supported options include SR (multimode, OM3 up to 70m / OM4 up to 100m) and LR (single-mode, up to 10km) transceivers. For more information, see the Transceiver, DACs, AOCs’ matrix](/en-us/netscaler-hardware-platforms/mpx/mpx-9200-17000-master-matrix).

Q4: Is 50G supported?

Yes. NetScaler-qualified 50GbE SFP56 SR and LR optics, as well as 50G SFP56 Direct-attach copper (DAC) cables, are supported on both the MPX/SDX 9200 and MPX/SDX 17000 platforms. For more information, see the Transceiver, DACs, AOCs’ matrix](/en-us/netscaler-hardware-platforms/mpx/mpx-9200-17000-master-matrix).

Q5: Are 40G QSFP+ transceivers supported?

Yes, only on the MPX and SDX 17000 platform. The Citrix-qualified options are:

• 9530411 (Citrix MPN) (40GbE QSFP+ BiDi SR LC optical): Ensure that the same NetScaler-branded module is used on both ends of the connection. Mixing this transceiver with a Cisco-branded 40G-BiDi module on the switch side can cause link flapping.

• 9530513 (Citrix MPN) (40GbE QSFP+ DAC, 5 m passive copper): A DAC alternative to BiDi optics, ideal for short-reach 40G links.

Q6: Are 100G/40G QSFP28 transceivers supported?

Yes, only on the MPX and SDX 17000 platform. The Citrix-qualified options are:

• 9530410 (Citrix MPN): 100/40GbE Dual-Rate QSFP28 SR4 (MPO-12 MMF)

• 9530413 (Citrix MPN): 100GbE QSFP28 LR4 (LC, SMF, 10 km)

• 9530517 (Citrix MPN): 100GbE QSFP28 AOC (20 m)

Q7: Is 200G supported?

Yes, only on the QSFP56 ports of the dual-port 200G NICs on the MPX/SDX 17000 platform. The Citrix-qualified options are:

• 9530409 (Citrix MPN): 200GbE QSFP56 SR4 (MPO-12 MMF)

• 9530514 (Citrix MPN): 200GbE QSFP56 to QSFP56 Passive Twinax DAC (2.5 m)

• 9530522 (Citrix MPN): 200GbE QSFP56 AOC (20 m)

Q8: Can I use a QSFP28 adapter at 25G or 10G with regular SFP optics on the MPX/SDX 17000 200G QSFP56 ports?

Yes, by using the Citrix adapter (Citrix MPN: 9530416). This is a QSFP28-to-SFP28 mechanical adapter based on the Mellanox/Nvidia MAM1Q00A-QSA28 design.

The adapter slots into a 200G QSFP56 port on the MPX/SDX 17000 platform and accepts any qualified Citrix-branded SFP28 (25G) or SFP+ (10G) optic. This is particularly useful when the appliance’s native SFP56 ports are fully utilized, allowing customers to provision an additional 25G or 10G uplink from an available QSFP56 port.

Auto-Negotiation, Speed Forcing & FEC

Q1: Do ports auto-negotiate speed with these transceivers?

It depends on the link speed and the transceiver type:

Q2: What FEC mode should I configure on the peer switch?

Match FEC to the link speed and media. FEC mismatch is the #1 cause of link-up-but-no-traffic or high-CRC-error symptoms.

Q3: Why does not my 40G link come up with auto-negotiation?

Whether a 40G link comes up on its own depends on the type of cable or transceiver used, not on the NIC card itself. The behavior differs by cable type:

• 40G copper cable (passive DAC): The link comes up at 40G automatically. No configuration is needed on either end.

• 40G fiber transceiver (BiDi): These run at a fixed speed and do not support auto-negotiation. If a port is left set to auto-negotiate, it does not find a matching partner and the link does not come up. To fix this, set the port to a fixed 40G speed on whichever device defaults to auto-negotiation, typically the Cisco peer switch (set “speed 40000” on the Nexus peer switch).

• 100G transceiver (QSFP28) used at 40G: With auto-negotiation enabled, the link comes up at its native 100G speed, not 40G. To run it at 40G, manually set a fixed 40G speed on both ends.

Mixing & Compatibility

Q1: Can I use third-party optics or DAC/AOC cables?

Only Citrix-qualified, Citrix-branded MPNs (953xxxx series) are supported. Using non-qualified modules may cause link bring-up issues, FEC mismatch, or DDM read failures, and may not be covered under standard support terms.

Q2: Can I mix SR and LR (or different vendor optics) on each end of the link?

Both ends of a link must use compatible optics, which means same wavelength, same compliance code (for example, SR4 to SR4, LR4 to LR4), and matching connector or fiber type. Mixing SR with LR will not establish a link. For 40G BiDi optical, both ends must be the same Citrix MPN 9530411.

Q3: How do I choose SR vs LR?

• SR (Short Reach): Multimode fiber (OM3 70m / OM4 100m typical for SR4 at 100G+; OM3 300m / OM4 400m for 10G SR). Use within a single rack or row.

• LR (Long Reach): Single-mode fiber (OS2), typically up to 10 km. Use for inter-row, inter-DC, or campus links.

Always confirm the qualified reach in the Transceiver, DACs, AOCs’ matrix](/en-us/netscaler-hardware-platforms/mpx/mpx-9200-17000-master-matrix). Actual reach depends on the optic and the fiber grade.

Q4: What MPO polarity do I need for SR4 modules (40G/100G/200G)?

All Citrix-qualified MPO-12 SR4 modules (Citrix MPN 9530409, 9530410) require Type B (key-up to key-down) MPO-12 trunk cables for correct fiber polarity. Type A or Type C trunks result in no-link or a one-way link.

Verification & Troubleshooting

Q1: How do I verify that the transceiver is detected and the link is up?

• Check the NetScaler host interface state: link, speed, FEC.

• Check the peer switch interface: link, speed, FEC, error counters.

• If the link is down: reseat the module, verify cable polarity (LC duplex or MPO Type B), confirm both sides match in speed and FEC.

Q2: Can I use a Cisco Nexus N9K with non-Cisco optics?

Yes, with the service unsupported-transceiver command enabled in the global configuration of the Nexus switch. Without this command, the Nexus N9K may refuse to bring up the port even if the optic is electrically compatible. Per-lane DDM may be reported as N/A on the N9K for non-Cisco QSFP modules. This is cosmetic, not a fault.

Q3: My 9530413 (Citrix MPN) 100G LR4 link is not coming up over OS2 fiber to a Mellanox SN3700 or Cisco N9K switch. The transceivers are seated and powered. What should I check first?

The 9530413 (Citrix MPN) samples should auto-negotiate to 100G LR4 over OS2 single-mode fiber without any manual speed forcing. If the link is not coming up, the most common root cause is a fiber-side issue (dirty or damaged LC connector, or a bad/broken patch cable), not the transceiver or NIC.

Diagnostic indicator: Check switch-side per-lane RX power. If all four lanes report approximately -40 dBm (the optical floor / “no light”), this is not a transceiver fault. It means that no optical signal is reaching the switch receiver. The link never reaches PHY training under these conditions, which is why auto-negotiation appears to fail.

Example symptom on a Mellanox SN3700:

Rx Power (per channel):
    Ch1 Rx Power : 0.00010 mW / -40.00000 dBm   ← floor / no light
    Ch2 Rx Power : 0.00010 mW / -40.00000 dBm   ← floor / no light
    Ch3 Rx Power : 0.00010 mW / -40.00000 dBm   ← floor / no light
    Ch4 Rx Power : 0.00010 mW / -40.00000 dBm   ← floor / no light
    Ch1–Ch4 Rx Power Alarm : Rx Pwr Too Low

Meanwhile, TX power on both host and switch reads normal (~1.5–1.7 mW / ~2 dBm per channel) with no Tx alarms and the lasers are healthy. The optical path between them is broken.

Resolution steps:

1. Replace the OS2 LC-to-LC duplex single-mode patch cable with a known-good fiber.

2. Reseat both transceivers (host side and switch side) after fiber replacement.

3. Verify LC polarity and confirm that the duplex connectors are not crossed or mis-keyed.

4. Re-check switch-side RX power. After fiber replacement, RX power on all four channels should return to a normal range (typically +1 to +3 dBm), all “Rx Pwr Too Low” alarms should clear, and the link should come up at 100Gx4, auto-negotiation, without manual configuration.

Example after fiber replacement (Mellanox SN3700):

Eth1/31  Operational state: Up / Speed: 100Gx4 / Negotiation: Auto
Rx Power (per channel) → alarms now NONE
TX Power (per channel) → alarms NONE

Q4: What is the QSFP28-to-SFP28 adapter 9530416 (Citrix MPN) and when do I use it?

This QSFP28-to-SFP28 mechanical adapter (based on the Mellanox MAM1Q00A-QSA28) fits a 1-lane SFP-class optic into a 4-lane QSFP56 port on the MPX/SDX 17000. With this adapter installed, the 200G QSFP56 port can operate at lower link speeds, specifically 25G with an SFP28 optic (for example, 9530402 SR), or 10G with an SFP+ optic (for example, 9530403 SR), when the host NIC and peer switch are configured to match.

Q5: When would I use the QSFP28-to-SFP28 adapter 9530416 (Citrix MPN) instead of just using an SFP56 port directly?

Use the adapter when:

• All 8x SFP56 ports on the MPX/SDX 17000 are already in use and you need an additional 25G or 10G uplink.

• A deployment design specifically requires a 25G or 10G link from a QSFP56 port (for example, to match a particular switch panel layout).

For a fresh deployment, prefer the native SFP56 ports on the MPX/SDX 9200 and 17000 for 10G / 25G / 50G optics as they require no adapter and have the simplest interop path.

Q6: Is the QSFP28-to-SFP28 adapter 9530416 (Citrix MPN) supported on the MPX/SDX 9200?

No. The MPX/SDX 9200 has no QSFP56 ports. The 9530416 (Citrix MPN) is only applicable to the MPX/SDX 17000.

Q7: Does the QSFP28-to-SFP28 adapter 9530416 (Citrix MPN) add latency or signal loss?

No. The 9530416 (Citrix MPN) is a passive mechanical adapter. It physically routes a single lane between the SFP28/SFP+ optic and one lane of the QSFP56 port with no active components or retiming. Signal integrity and latency are the same as a native SFP socket.