Why Comparing 5G and 4G Requires More Than Peak Numbers

Marketing materials quote 20 Gbps for 5G and 1 Gbps for LTE, but those are theoretical peaks defined in ITU IMT-2020 and IMT-Advanced respectively. Real-world performance depends on spectrum holdings, site density, device penetration, and traffic load. This article presents both the 3GPP-specified theoretical limits and independently measured field data so you can make informed comparisons across three generations.

The specification baselines come from 3GPP TS 36.306 (E-UTRA UE radio access capabilities) for LTE, 3GPP TS 38.306 (NR UE radio access capabilities) for 5G NR, and the older 3GPP TS 25.306 for WCDMA/HSPA+. We use these to anchor the peak numbers, then overlay real measurements from Ookla Speedtest Intelligence and OpenSignal reports published through Q4 2025.

Theoretical Peak Comparison: 3G, 4G, 5G

The peak DL throughput for NR FR1 is derived per 3GPP TS 38.306 clause 4.1.2: the data rate formula is:

Throughput = v_layers x Q_m x f x R_max x (N_PRB x 12 / T_s) x (1 - OH)

Where v_layers = 4, Q_m = 8 (256QAM), f = 1 (scaling), R_max = 948/1024, N_PRB = 273 (for 100 MHz at SCS 30 kHz), T_s = 1 ms / 14 symbols, and OH = 0.14 for DL. This yields approximately 4.6 Gbps.

Real-World Speed Data: Global Operator Measurements

Laboratory peaks tell only part of the story. The table below compiles median speeds reported by Ookla Speedtest Intelligence and OpenSignal for selected countries as of Q4 2025.

Key takeaway: real-world 5G delivers a 5x to 10x improvement over 4G, not the 20x that peak specs suggest. The gap narrows where operators deploy 5G on mid-band (n78 at 100 MHz) and widens where mmWave (n258/n261) is available, as in South Korea and parts of the USA. SK Telecom's n78 deployment with 100 MHz contiguous bandwidth and Massive MIMO 64T64R consistently hits top-tier median speeds.

Worked Example 1: Capacity Calculation per Cell

Suppose an operator deploys a 5G NR cell on band n78 (3.5 GHz) with 100 MHz bandwidth, SCS = 30 kHz, 4x4 MIMO, and 256QAM, versus an LTE cell on band 3 with 20 MHz and 4x4 MIMO.

Step 1 -- Resource Block count.

• NR: 100 MHz / (12 x 30 kHz) = 273 PRBs per slot (per 3GPP TS 38.101-1 Table 5.3.2-1).
• LTE: 20 MHz / (12 x 15 kHz) = 100 PRBs per subframe (per 3GPP TS 36.101 Table 5.6-1).

Step 2 -- Symbols per second.

• NR (SCS 30 kHz): 14 symbols per 0.5 ms slot = 28,000 symbols/sec.
• LTE (SCS 15 kHz): 14 symbols per 1 ms subframe = 14,000 symbols/sec.

Step 3 -- Peak bits per symbol.

• NR: 8 bits (256QAM) x 4 layers x 948/1024 coding rate = 29.6 bits/RE.
• LTE: 8 bits x 4 layers x 948/1024 = 29.6 bits/RE (same modulation order if using 256QAM per Cat-12+).

Step 4 -- Peak throughput (single carrier).

• NR: 273 PRBs x 12 subcarriers x 28,000 symbols/s x 29.6 bits x (1 - 0.14 OH) = ~4.59 Gbps.
• LTE: 100 PRBs x 12 subcarriers x 14,000 symbols/s x 29.6 bits x (1 - 0.14 OH) = ~633 Mbps.

Result: The NR cell delivers 7.25x the peak capacity of the LTE cell. The dominant factors are 5x more bandwidth (100 MHz vs 20 MHz) and 2x more symbols per second (SCS 30 kHz doubles the slot rate).

Spectral Efficiency: A Fairer Comparison

When normalized per Hz of bandwidth, 5G NR offers a more modest improvement over LTE-Advanced Pro. The gains come primarily from better reference-signal design, LDPC/Polar coding, and more flexible scheduling.

The cell-edge and average gains are where 5G truly excels -- 70% or more improvement -- driven by grant-free UL, flexible slot formats, and Massive MIMO beamforming. These system-level gains translate directly into the real-world speed ratios observed in the operator data above.

Worked Example 2: Latency Budget Comparison

Consider the one-way user-plane latency budget for LTE FDD (SCS 15 kHz, TTI = 1 ms) versus NR TDD (SCS 30 kHz, slot = 0.5 ms) per 3GPP TR 38.913 clause 7.2.

With mini-slot (2-symbol scheduling), NR at SCS 120 kHz can achieve sub-0.5 ms one-way latency, meeting the URLLC 1 ms round-trip target. LTE cannot approach this even with shortened TTI (sTTI) in Release 15, which only reduces TTI to 0.5 ms but retains the longer processing times.

T-Mobile US reported median 5G latency of 14.2 ms in their Ookla Q4 2025 data, which reflects the full round-trip including core network, internet peering, and server response. The 1--4 ms RAN latency above represents only the air-interface contribution. Verizon's mmWave sites in dense urban areas measured 9.8 ms median round-trip latency in the same report, confirming the lower-latency benefit of wider SCS and shorter slot durations.

Connection Density and IoT Scale

5G NR's mMTC capability targets 1 million devices per km-squared per ITU-R M.2083 (IMT-2020 requirements). LTE-M and NB-IoT (both Release 13) were the first to approach this density class. 5G RedCap (Release 17, per 3GPP TS 38.300 clause 4.4a) extends this with reduced-capability NR devices that support 20 MHz bandwidth, 1--2 Rx antennas, and half-duplex FDD -- bridging the gap between full NR and NB-IoT.

Deutsche Telekom's NB-IoT network in Germany supports over 700,000 connected meters and sensors as of 2025, while SK Telecom has deployed RedCap for wearable and industrial sensor applications with a target of 2 million RedCap devices by 2027.

Capacity Multiplier: How 5G Delivers More Bits per km-squared

The area traffic capacity metric (Mbps/m-squared) is the product of spectral efficiency, bandwidth, and site density. 5G achieves 10 Mbps/m-squared versus LTE's 0.1 Mbps/m-squared through three multipliers:

1. Bandwidth: 5x (100 MHz vs 20 MHz on FR1; up to 400 MHz on FR2).
2. Spectral efficiency: 1.7x average cell throughput improvement (Massive MIMO + LDPC coding).
3. Site densification: 3--10x in dense urban (small cells on FR2 at 200 m ISD vs macro at 500 m ISD).

Combined: 5 x 1.7 x 6 = 51x improvement in area traffic capacity -- aligning with the ITU 100x target when FR2 mmWave layers are included.

When Does 4G Still Win?

Coverage and uplink performance in rural areas remain LTE strengths. Low-band LTE (700 MHz, band 28) propagates farther than mid-band NR (3.5 GHz, n78), requiring fewer sites per km-squared. Jio's rural India network delivers 19.4 Mbps median DL on LTE at 850 MHz with cell radii exceeding 10 km, whereas their 5G coverage on n78 is limited to urban and semi-urban areas with cell radii of 1--3 km. For indoor penetration, LTE on 700 MHz suffers 6--8 dB less wall loss than NR on 3.5 GHz.

Operators pursuing coverage-first strategies -- such as Dish Network (now EchoStar) in the USA deploying on n71 (600 MHz) -- can match LTE coverage footprints at the cost of reduced NR bandwidth (only 10--15 MHz is typically available below 1 GHz).

Summary

5G NR delivers 5--10x real-world speed improvements, 3--8x latency reduction, and up to 50x area capacity gains over LTE. The spectral efficiency improvement per Hz is modest (5--70% depending on scenario), but the combination of wider bandwidth, flexible numerology, and Massive MIMO creates a compounding effect. Operators with 100 MHz mid-band holdings consistently achieve median downloads above 200 Mbps, while mmWave sites push beyond 1 Gbps in dense venues. Understanding these distinctions -- and knowing when LTE still holds an advantage -- is essential for network planning, device strategy, and certification preparation.