NSA vs SA: The Two Ways to Deploy 5G, and Why It Matters

When your phone shows a 5G icon, that tells you almost nothing about which network you're actually on. It might be Non-Standalone 5G — NR radio bolted onto an LTE core — or Standalone 5G with a full 5G Core underneath.

The difference isn't cosmetic. Network slicing, the 5G QoS model, RedCap, and true URLLC do not exist on an NSA network. Not "perform worse" — do not exist, because they live in the 5G Core.

---

Why two options exist

Operators had spent a decade and enormous capital on LTE: radio equipment, fibre backhaul, and the Evolved Packet Core. Replacing all of it simultaneously to launch 5G was neither practical nor financeable.

3GPP defined several deployment options. Two dominate commercially:

Option 3x (NSA) — LTE anchors the control plane, NR adds capacity, both connect to the LTE EPC.

Option 2 (SA) — gNB connects directly to a new 5G Core. No LTE dependency.

NSA got 5G radio into networks fast. SA is what delivers the architecture.

---

How NSA works

NSA runs on EN-DC (E-UTRAN New Radio Dual Connectivity). The UE maintains simultaneous connections to an LTE eNB and a 5G gNB.

Roles are asymmetric and worth being precise about:

The eNB is the master node. It owns the RRC connection, handles all control-plane signalling, manages mobility, and terminates NAS toward the EPC. There is no NR RRC connection in Option 3x — the UE's RRC state is LTE's.

The gNB is the secondary node. It adds NR user-plane capacity and has no independent control-plane relationship with the core.

The "3x" specifically describes where user-plane traffic splits. In Option 3x, the gNB terminates the bearer split — traffic arrives from the EPC at the gNB, which decides how much to serve over NR and how much to route to the eNB. This matters because it avoids making the LTE eNB a throughput bottleneck, which is why 3x won out over plain Option 3.

What operators got from it

Speed to market, above all. NR carriers could be added to existing sites without touching the core, using existing LTE coverage as the mobility anchor.

Coverage robustness came free: LTE handles mobility, so losing NR coverage degrades throughput rather than dropping the connection. For early deployments with sparse NR coverage, that was genuinely valuable.

And the capital efficiency was decisive — EPC investment was preserved and 5G launched on a radio budget rather than a full network rebuild.

---

How SA works

The gNB connects directly to the 5G Core over NG. No LTE anchor, no dual connectivity requirement, NR RRC as the actual control plane.

What the 5GC brings is architectural, not incremental:

Service-Based Architecture. Network functions expose HTTP/2 REST APIs and discover each other through the NRF, rather than communicating over fixed point-to-point reference points. This is what makes cloud-native deployment and independent function scaling possible.

Full control/user plane separation. The SMF controls sessions; the UPF forwards packets. They scale independently, and UPFs can be placed at the edge — which is what makes low-latency edge computing work.

QoS Flows instead of bearers. Finer-grained differentiation within a single PDU Session, with policy able to change mid-session.

Native slicing. S-NSSAI-based slice selection at registration, with slice-aware function selection throughout.

---

EPC vs 5GC

The MME split is the structural change worth understanding. EPC combined mobility and session management in one function. 5GC separates them into AMF and SMF — which is what allows a UE to be served by one AMF while its sessions terminate at different SMFs in different slices. Slicing isn't a feature added to the 5G Core; it's a consequence of how the 5G Core is decomposed.

---

What NSA cannot do

This is the part that gets glossed over in coverage maps.

Network slicing. Requires S-NSSAI-based selection in the 5G Core. The EPC has no equivalent.

5G QoS Flows. NSA uses EPS bearers. No QFI, no 5QI, no reflective QoS.

True URLLC. Requires 5GC session management and edge UPF placement. LTE's control plane adds latency that can't be engineered away.

RedCap. Defined for NR connected to the 5G Core.

Edge computing with UPF selection. No UPF in an EPC.

Voice over NR (VoNR). NSA carries voice over LTE. VoNR needs the 5G Core and IMS integration.

Full NR coverage benefit. The UE must maintain LTE coverage for the anchor, which constrains where NR-only deployment is viable.

An NSA network delivers NR throughput. It does not deliver 5G capabilities. When a vendor demonstrates slicing or URLLC, that demonstration is on SA — always.

---

Migration

Most operators followed the same path: LTE, then NSA overlay, then SA — with a long period of both running in parallel.

The parallel operation phase is where the real complexity sits. Devices must support both. Voice must work in both, meaning VoLTE and VoNR fallback paths. Mobility must handle transitions between SA coverage and NSA or LTE coverage — EPS fallback being the main mechanism, where a UE on SA drops to LTE for a voice call when VoNR isn't available. That's a common source of user-visible issues, and a frequent source of field debugging.

Two migration realities are often understated. SA sometimes shows worse coverage than NSA initially, because there's no LTE anchor to fall back on and NR alone may not cover the same footprint. And the business case for SA isn't the radio — throughput is similar. It's the core network features, which only pay off if you have enterprise customers, slicing demand, or edge deployments to sell into. That's exactly why SA migration has been slower than early roadmaps predicted.

---

Where it stands

SA deployment is now well established in major markets, though NSA remains widespread — and in many networks both operate simultaneously, with SA in dense areas and NSA elsewhere.

The direction is clear. Slicing, RedCap, URLLC, edge computing, and everything 5G-Advanced builds toward all assume the 5G Core. NSA was a bridge, and bridges get crossed.

For engineers, the practical implication: when someone says "5G," ask which one. Feature availability, QoS behaviour, latency characteristics, and device support all differ, and a surprising amount of confused 5G troubleshooting traces back to an unstated assumption about which architecture is actually in play.

---

Key takeaways

NSA (Option 3x) — LTE eNB is the master node, gNB is secondary, EPC is the core. Fast to deploy, NR throughput only.

SA (Option 2) — gNB connects directly to the 5G Core. Full feature set.

EN-DC is the dual connectivity mechanism NSA runs on. The "3x" describes where the bearer splits.

Slicing, QoS Flows, RedCap, VoNR, URLLC and edge UPF all require SA. They're 5G Core features.

The SA business case is the core, not the radio. Throughput is comparable; capabilities are not.

---

Further reading

• 3GPP TS 37.340 — Multi-connectivity, EN-DC procedures
• 3GPP TS 23.501 — 5G System architecture
• 3GPP TR 21.915 — Release 15 description, deployment options
• What is 5G NR? — the radio side of the picture