100GbE - QSFP28, Breakout Cabling and the PCIe Trap

Quick answer

A two-machine 100GbE link costs $837.97 today - two QSFP28 cards and one QSFP28 DAC, no switch. A switched four-host setup runs $3026.92 ($756.73/host). Breaking one 100G port to four 25G hosts instead costs $1163.84 total ($290.96/host, origin card shared) - usually the cheaper path per host.

Prices updated · 26 100GbE-speed listings tracked

Share:WhatsAppFacebookPost on X
Breakout is usually the point at this tier

A 100GbE port is rarely bought to feed one machine at the full rate - it is four 25G lanes in one connector, and a breakout cable is the standard, documented way to split them into four independent 25G links. Computed live: one Intel E810 (QSFP28) as the origin, one QSFP28-to-4x25G breakout DAC, and 4 × Intel E810 (SFP28) - one per host - comes to $1163.84 total, $290.96 per host not counting the shared origin card.

ipolex ipolex QSFP28 card (Intel E810 claimed) $405.99 · 10Gtek breakout DAC $45.89 · H!Fiber H!Fiber SFP28 card (Intel E810 claimed) $177.99 each

100GbE (QSFP28) network cards in stock now

CardPortsInterfaceConditionPrice
Intel E810
ipolexseller claims this chipset
2QSFP28New$405.99Buy
Intel E810
10Gtekseller claims this chipset
2QSFP28New$479.99Buy
ConnectX-5
Mellanox
1QSFP28New$485.00Buy
ConnectX-5
Mellanox
2QSFP28Refurb$489.99Buy
ConnectX-5
Mellanox
1QSFP28New$679.00Buy
ConnectX-5
Mellanox
2QSFP28New$750.00Buy

Straight point-to-point (no breakout): the cheapest Intel E810 card is $405.99chipset attribution is the seller's; two of those plus the cheapest QSFP28 DAC ($25.99, 0.5m) comes to $837.97 for one full-rate link between two machines.

QSFP28 direct-attach cables (including 100G-to-4x25G breakout)

Passive direct-attach copper, by length. A DAC is a fixed-length cable with the transceivers moulded on - you buy the whole run at once rather than a cable plus two modules.

LengthBrandTypePrice
0.5m10GtekQSFP28 passive$25.99Buy
1m10GtekQSFP28 passive$27.99Buy
1m10GtekQSFP28 passive$27.99Buy
1m10GtekQSFP28 passive$45.99Buy
2m10Gtek1 seller compat claimQSFP28 passive$33.99Buy
2m10GtekQSFP28 passive$45.89Buy
2mipolexQSFP28 passive$48.99Buy
2m10GtekQSFP28 passive$50.99Buy
3m10GtekQSFP28 passive$35.09Buy
3m10GtekQSFP28 passive$35.09Buy
3m10GtekQSFP28 passive$38.99Buy
3m10GtekQSFP28 passive$57.99Buy

QSFP28 transceivers

Every "compatible with" phrase below is the seller's own wording, recorded as a claim and badged. We have not tested any module against any switch.

ModuleTypeBrandSeller compatibility claimsPrice
QSFP28 100GBASE-SR4 module100GBASE-SR410Gteknone stated$39.99Buy
QSFP28 100GBASE-SR4 module100GBASE-SR410Gteknone stated$39.99Buy
QSFP28 100GBASE-SR4 module100GBASE-SR410Gteknone stated$39.99Buy
QSFP28 100GBASE-LR4 module100GBASE-LR4QSFPTEKfor QSFP-100G-LR4-Sseller claim$229.99Buy
QSFP28 100GBASE-LR4 module100GBASE-LR4QSFPTEKnone stated$269.99Buy

The PCIe requirement, per card

A 100GbE link needs real PCIe bandwidth to the host, and the vendor's own stated requirement - not a throughput number of ours - is what decides whether a slot works. NVIDIA's manuals state ConnectX-5's 100G SKUs at PCIe 3.0 x16, with the ConnectX-5 Ex variant adding PCIe 4.0 x16 on its 100G SKUs. Intel's PCIe figure for E810's 100G SKU was not sourced in our research pass - ark.intel.com did not render for us this session - so that card's own page states the gap rather than assuming a figure from the family pattern. Either way: check the exact part number's own PCIe line, and check your slot's electrical width rather than its physical size, before buying a 100G card.

Switches at this tier

SwitchPorts (as listed)PriceCondition
QNAP QSW-M7308R-4X-USQSW-M7308R-4X-US, 4x 100GbE QSFP28 + 8x 25GbE SFP28$1299.00NewBuy
Dell S5248F-ONS5248F-ON, 48x 25GbE SFP28 + 4x 100GbE QSFP28$2599.90RefurbBuy
Mellanox Spectrum-based Open Ethernet Switch (exact model not stated)10/25GbE and 100GbE 1U Open Ethernet switch$4995.00RefurbBuy
Mellanox MSN2100-BB2F Spectrum SN2100SN2100, 16x 40 Gigabit QSFP28, L3 managed$5849.00RefurbBuy

None of the switches in this table had a vendor datasheet render with a power or acoustic figure in our sourcing pass - NVIDIA/Mellanox's per-model SN2000-series PDFs downloaded but did not extract to legible text, and Dell's and QNAP's datasheets were not part of this pass. Stated as a sourcing gap rather than papered over with an estimate; what is fair to say without a figure: every unit here is 1U rack-mount datacenter hardware built for a cooled server room, not a desk.

Why breakout is usually the point

Very few home or homelab builds have one machine that genuinely needs a full 100Gb/s link to another single machine - the far more common shape is one fast uplink feeding several 25G hosts, which is exactly what a QSFP28-to-4x25G breakout cable does, per the standard rather than as a workaround.

That reframes the buying decision: a single 100G card and a breakout cable can be cheaper per connected host than four separate 25G point-to-point links would be, because the origin card and the switch port (where one exists) are shared costs. The fanout arithmetic above is computed from exactly that shape, live.

The trade is topology, not cost: a breakout cable's four legs are four independent links from one port, not one device getting four times the bandwidth. Where a build genuinely needs one very fast link rather than several independent ones, a straight QSFP28 DAC or optic is the simpler, more standard answer.

Where to go next

The 25/40/100GbE hub has the full cluster and the form-factor map · 40GbE is the tier below this one, and often the cheapest switched build of the three · ConnectX-5 and Intel E810 are the two cards this page's live builds use · the cabling guide covers breakout cabling in full · rack disk shelves and NAS devices are the usual hardware on the other end - this page prices the link, not what your storage can do with it.

Frequently asked questions

Do I need 100GbE, or is 25G/40G enough?

This page prices the link and states what vendors document; whether your storage or workload can use it is a question we don't answer here - see /enclosures/jbod and the NAS pages for the hardware facts on the other end. What we can say: a 100G port is frequently bought to break out to several 25G hosts rather than to feed one machine at the full rate - see the breakout arithmetic below.

What PCIe slot does a 100GbE card need?

Depends on the exact SKU, and this is the category's most expensive mistake to get wrong. NVIDIA's own manuals state ConnectX-5 100G SKUs at PCIe 3.0 x16 (PCIe 4.0 x16 on the Ex variant); Intel's PCIe requirement for E810 100G SKUs was not sourced in our research pass. Check the specific part number's PCIe line before buying, and check your slot's electrical width, not just its physical size - a card in a physically x16 but electrically x4 slot cannot use the lanes it needs.

What is a 100G-to-4x25G breakout cable?

One QSFP28 connector on one end, four independent SFP28 legs on the other - the standard way a single 100G port electrically is four 25G lanes, split back out into four separate links to four separate hosts. Live listings and the computed cost of a full fan-out are below.

Are the 100GbE switches in your table new or used?

Mixed, and the table states condition per row - several Mellanox and Dell listings are used/renewed enterprise pulls, and one QNAP unit is current-production SMB hardware aimed at exactly this kind of homelab breakout build.

Is Intel E810 or Mellanox ConnectX-5 the better 100G card?

We don't make that call - see each card's own page for its sourced facts (driver, PCIe where available, EOL status, third-party board prevalence) and the live prices above. The two differ meaningfully on how third-party-board-heavy the market is (E810 far more so) and on how completely we could source PCIe/EOL facts (ConnectX-5 far more completely, since ark.intel.com did not render for our Intel research pass).

Sources

Specification claims on this page were checked against the pages below on 4 August 2026. Where a widely-repeated figure did not survive that check, we report what the source says now and say so in the text rather than repeating the familiar number.

  1. NVIDIA (Mellanox) - ConnectX-3 EN product line and lifecycle status (nvidia.com, 2026)
    Used for: that ConnectX-3 has reached end-of-life status with NVIDIA, and the MCX311A / MCX312A / MCX314A part-number family. EOL means no new firmware or driver work from the maker; it does not mean the card stops working, and the page says so in those terms.
  2. The Linux kernel project - mlx4, mlx5, ixgbe and i40e in-tree Ethernet drivers (kernel.org, 2026)
    Used for: that mlx4 (ConnectX-3), mlx5 (ConnectX-4 and later), ixgbe (Intel 82599/X520, X540, X550) and i40e (X710) ship in the mainline Linux kernel. This is why an end-of-life card can still be a sane purchase, and it is a documented fact rather than a forum claim.
  3. DatacenterDisk - Live catalogue - NIC, DAC and transceiver listings (datacenterdisk.com, 2026-08-11)
    Used for: every price on this page, read from our catalogue at request time. Seeded 11 August 2026: 24 network cards across 6 models, 24 DAC cables, 8 transceivers. Every card is split tier 1 (the listing is that part) or tier 2 (a third-party board asserting that silicon).
  4. DatacenterDisk - Sourcing note - power figures and compatibility claims (datacenterdisk.com, 2026-08-11)
    Used for: We publish no per-port wattage for either cabling path. The relationship - 10GBASE-T draws more power and runs hotter than SFP+ direct attach - is documented at product level and is stated here. The magnitude is not, because we have not fetched and logged a source we can cite for it. Separately: every 'compatible with' phrase shown on this site is stored as a seller claim at tier 2 and badged wherever it renders. We have not tested any module against any switch.
  5. NVIDIA - ConnectX-5, ConnectX-4 Lx and ConnectX-4 EN user manuals - end-of-life status (networking-docs.nvidia.com, 2026-09-13)
    Used for: that NVIDIA's own manuals state "The ConnectX-5 product line has moved to end-of-life" and "The ConnectX-4 Lx product line has moved to end-of-life" in those exact words (connectx5enhw and connectx4lxhw respectively, both dated 2026). The standard ConnectX-4 EN manual (connectx4enhw) carries an "EOL'd Ordering Part Numbers" table for every MCX413A/414A/415A/416A SKU but not the same plain-English banner sentence - stated as that asymmetry rather than smoothed into one claim. Also the per-SKU PCIe generation and lane count (PCIe 3.0 x8 or x16 depending on SKU; PCIe 3.0/4.0 on ConnectX-5 Ex variants) from the same manuals.
  6. NVIDIA - Linux Kernel Upstream Release Notes - driver and firmware support table (networking-docs.nvidia.com, 2026-09-13)
    Used for: that ConnectX-4, ConnectX-4 Lx and ConnectX-5 all use the mlx5 driver family per NVIDIA's own "Supported Uplinks to Servers" table, dated June 2026 - current documentation, confirming the in-kernel driver is still maintained for hardware NVIDIA itself calls end-of-life for ordering. Also that NVIDIA's current DOCA-Host device list (docs dated September 2026) includes ConnectX-5 and ConnectX-4 Lx but not the standard ConnectX-4 - a real difference in ongoing software-package support between the two, stated as found rather than smoothed over.
  7. The Linux kernel project - i40e and ice in-tree Ethernet driver documentation (kernel.org, 2026-09-13)
    Used for: that the i40e driver explicitly lists Intel Ethernet Controller XXV710 as supported hardware alongside X710/XL710/X722 (same page also notes XXV710's speed cannot be forced via ethtool - auto-negotiation only). The companion ice driver page (device_drivers/ethernet/intel/ice.html) is titled for the "800 Series" - the family E810 belongs to - without naming E810 verbatim in the rendered portion, which is stated as the limit of what this specific page confirms rather than stretched into a direct quote. Intel's own ARK product pages did not render in this sourcing pass (see the 100GbE and NIC pages' own note); PCIe generation and EOL status for XXV710 and E810 parts are marked not sourced for that reason.
  8. Arista Networks - 7050X and 7050X3 Series switch specifications, and End-of-Sale advisories (arista.com, 2026-09-13)
    Used for: the 7050X family's own published figures: "10GbE Typical Power: 1.2W per port", and for the 7050QX-32S specifically "Typical Power Draw: 150W (4.5W per port)" - the only per-port power figures this cluster publishes, because they are the only ones a vendor datasheet actually rendered and stated. Also Arista's own End-of-Sale advisory archive (arista.com/en/support/advisories-notices/endofsale), which carries a dated 2016 end-of-sale notice for "Select Models of the Arista 7050SX, 7050TX and 7050QX Series" and, separately, 2026-dated end-of-sale notices for the newer 7050SX3-48YC12 and 7050SX3-48YC8/48C8 models - a used 7050-family switch may be running silicon its own maker has since end-of-saled, stated with the notice's own date rather than ours.
  9. DatacenterDisk - Live catalogue and sourcing note - 25/40/100GbE cluster (datacenterdisk.com, 2026-09-13)
    Used for: every price in this cluster's tables, read from our catalogue at request time. Seeded 13 September 2026: 12 NICs across 3 models (ConnectX-5, ConnectX-4, Intel E810), 40 DAC cables (25G/40G/100G including breakout), 18 transceivers, 11 used-enterprise and current-SMB switches. Two vendor documentation gaps stated plainly rather than filled: NVIDIA/Mellanox's per-model switch datasheets (SN2010/SN2100/SN2410/SN2700) did not render in this sourcing pass beyond a series-level marketing page, so those four switches carry no sourced power, acoustic or EOL figures; Intel's ark.intel.com did not render at all in this pass, so XXV710 and E810 PCIe generation and formal EOL status are marked not sourced rather than assumed from the 700/800-series family pattern.