Enter for the full results page · Esc to close
A nine-micron core or a fifty-micron one — and why the cheaper fibre is often the more expensive decision.
The old rule was simple: multimode is cheaper for short links because the optics are cheaper, single-mode for long links because the fibre goes further. That rule is now about fifteen years out of date, and following it is how data centres end up re-cabling.
What changed is that single-mode optics got cheap. A 100G CWDM4 module is no longer dramatically more expensive than an SR4, while multimode reach has been falling with every speed generation — 300 metres at 10G, 100 metres at 100G, and less again at 800G. The crossover point has moved a long way.
A core of about nine microns, small enough that only one light path propagates. No modal dispersion at all, so distance is limited by loss and by chromatic dispersion rather than by pulse overlap.
Reach runs from 500 metres for DR4 to 80 kilometres and beyond for coherent variants, on the same glass. The fibre itself costs slightly more than multimode but is essentially future-proof: the plant you install today carries whatever speed you buy in a decade. See OS2 / G.652.D.
A fifty-micron core carrying many light paths at once. Cheap 850 nm VCSELs can be coupled into it easily, which is what makes the optics cheap, but the different paths arrive at different times and that sets a hard distance limit.
That limit shrinks as speed rises: 10GBASE-SR reaches 300 metres on OM3, 100GBASE-SR4 reaches 70, and 800GBASE-VR8 reaches 50 on OM4. No amount of installation care changes it — it is modal bandwidth, not loss. See OM3 vs OM4 vs OM5.
The row that decides most arguments is the last one: what happens at the next speed upgrade.
| Single-mode (OS2 / G.652.D) | Multimode (OM3 / OM4 / OM5) | |
|---|---|---|
| Core diameter | ~9 µm | 50 µm |
| Limiting factor | Loss and chromatic dispersion | Modal bandwidth |
| Reach at 100G | 500 m (PSM4) to 10 km (LR4) | 70 m OM3 / 100 m OM4 |
| Reach at 800G | 500 m (DR8) to 2 km (2×FR4) | 50–100 m OM4 |
| Attenuation | 0.35 dB/km @1310, 0.22 @1550 | ~3.0 dB/km @850 |
| Typical module cost | Higher, but the gap has narrowed a lot | Lowest for SR variants |
| Cable cost | Slightly lower | Slightly higher |
| Connector loss budget | Generous — 3–6.3 dB typical | Tight — 1.8–1.9 dB at 100G+ |
| Survives the next speed upgrade | Yes — same glass | Often not — reach shrinks each generation |
The connector budget row is under-appreciated. Single-mode 400GBASE-DR4 allows 3 dB across the channel; multimode 400GBASE-SR8 allows 1.8 dB. On multimode, four mated pairs at 0.5 dB is already 2 dB — over budget with no fibre counted at all. Multimode is not just shorter, it is less forgiving of the patch panels real installations contain.
Cost, honestly: multimode SR optics are still cheaper per module, and for a rack of sub-50-metre links that saving is real. But the multimode cable and connectors cost slightly more than single-mode, so the advantage is entirely in the optics — the part you replace every few years, not the part you live with for fifteen.
Install single-mode unless you can honestly say every link will stay under 100 metres for the life of the plant. Single-mode fibre carries 10G, 100G, 400G, 800G and coherent 400ZR on the same glass; multimode's reach shrinks with each generation, so a plant sized for today's 100G may not carry tomorrow's 800G over the same distance. The fibre is the long-lived asset and it should be the one chosen conservatively.
If you already have OM4 and your links are under 100 metres, there is no argument for re-cabling. SR optics remain the cheapest per link and the plant is paid for. Watch the loss budget rather than the distance — that is what usually fails first — and keep the mated-pair count down.
The practical arrangement in most data centres is multimode for in-row server links and single-mode for everything crossing a hall or a building. That keeps the cheap optics where they work and puts the durable fibre where distance matters. Label the two plants unmistakably; mixing them produces links that come up degraded rather than failing cleanly, which is much harder to diagnose.
At these rates multimode is a within-the-row technology. DR4 at 500 metres and FR4 at 2 km are the variants that cross a data hall, and both are single-mode. Confirm the budget in the link budget calculator before committing.
The fibre is not — single-mode cable is generally slightly cheaper than OM4. The optics still are, but the gap has narrowed dramatically: CWDM4 and DR4 modules exist precisely to be affordable single-mode options. Over the life of a cabling plant, single-mode is frequently the cheaper total decision.
No, and it does not work well enough to be worth attempting. Mode-conditioning patch cords existed for 1G LX over multimode, but that is a legacy special case. Match the optic to the fibre.
For the links that are short enough. Ask what your longest multimode run is: if it is 80 metres you are fine through 400G-SR8; if it is 150 metres you were already out of spec at 100G. The reach lookup answers this per standard.
Only if you intend to use SWDM or BiDi optics. On a conventional SR link, OM5 reaches exactly as far as OM4 and costs more. It is not a general-purpose upgrade. See OM3 vs OM4 vs OM5.
For parallel optics such as DR4 and DR8, yes. For duplex single-mode, UPC is common and fine. Multimode is conventionally UPC throughout. The rule that never bends is that APC and UPC must not be mated to each other.
We build both plants — single-mode and multimode, patch cords through to 144-fibre trunks, measured fibre by fibre with the report in the box.
Two of these differ by distance. The third differs by something else entirely — and it is widely mis-sold.
Read the comparison →Optical Variants & ReachSame 400 gigabits, three different reaches — and three completely different cabling bills.
Read the comparison →Optical Variants & ReachFour 100G options separated by fibre type, fibre count and how far you actually need to go.
Read the comparison →Fibre Plant & ConnectorsEight degrees of angle, ten decibels of return loss, and one mistake that destroys both ferrules.
Read the comparison →Send the specification and an engineer answers it — no obligation, and no sales sequence. If the answer is that you do not need what we sell, that is the answer you get.
Usually answered within one working day.