Amarisoft is a commercial LTE/5G software stack. It has three parts: the lteenb eNB/gNB, the ltemme unified 4G/5G core, and the lteue soft UE, which can simulate many UEs on one radio. It is licensed through a floating network license (tags gNB, 5gC and ue). This page is the map: a short introduction, the standardized Amarisoft tutorials, and what each one measures.
Every Amarisoft tutorial on COSMOS shares the same building blocks:
lteenb + ltemme on the RAN server, from the amarisoft-<date>.ndz image;lteue on a UE node, from the lte5gue-<date>.ndz image;.cfg files for configuration.The amarisoft.ndz and lte5gue.ndz symlinks always point at the newest dated images. Amarisoft is COSMOS's commercial reference stack. Its soft UE is also the cross-vendor test UE for the OCUDU and Duranta stacks.
The 5G SA tutorials are standardized on test PLMN 00101 and UE pool 192.168.100.0/22 (gateway 192.168.100.1, with internet access), the same as OCUDU and Duranta. For how Amarisoft compares to them, see the 5G SA platforms overview.
| Component | Role | Where it lives |
|---|---|---|
lteenb |
eNB (4G) and gNB (5G) | RAN server, from amarisoft-<date>.ndz |
ltemme |
the unified 4G EPC / 5G core | RAN server, same image |
lteue |
soft UE — simulates many UEs on one radio, and is COSMOS's cross-vendor test UE for the OCUDU and Duranta stacks | UE node, from lte5gue-<date>.ndz |
| licence client | floating network licence, tags gNB, 5gC, ue |
contacts the licence server on mgmt1 |
.cfg files |
JSON-style configuration for each component | on each host |
amarisoft-<date>.ndz · lte5gue-<date>.ndz |
the node images; amarisoft.ndz and lte5gue.ndz point at the newest |
COSMOS image store |
🔴 The licence is one seat per tag, testbed-wide. A leftover
lteenb,ltemmeorlteue
anywhere on COSMOS holds the seat, and the next user's binary segfaults with no banner and no
licence error — it simply dies. Before blaming a config, look for the current seat holder in
the licence log onmgmt1. Never restart the licence daemon to clear it.
Amarisoft is COSMOS's commercial reference stack: the point of comparison for the two open stacks, OCUDU (srsRAN) and Duranta (OAI).
| # | Tutorial | Modality | What it needs | Start here if… |
|---|---|---|---|---|
| 1 | Over-the-air — gNB + core + soft UE on a USRP | USRP, direct | sb1/grid + two USRPs (or one USRP and a Quectel modem), ~60 min |
…you want a commercial 5G SA cell on air with a UE attached and passing traffic. |
| 2 | Controlled cell: UE capacity sweep and performance — three ways | USRP, direct (4G and 5G) | a RAN server + the sdr1-in3.sb1 UE bank, ~30–90 min |
…you want to measure how a cell scales from 1 to 64 UEs, in LTE or 5G SA, run as CLI, Ansible or Jupyter. |
| — | Soft-only (no radio) | — | — | No soft-only Amarisoft tutorial on COSMOS. Use the OCUDU or Duranta soft-only tutorial. |
| — | O-RAN 7.2 split (vendor or soft O-RU) | — | — | No standardized Amarisoft 7.2 tutorial yet. The older O-RAN Split 7.2 End-to-End page covers an earlier setup. |
This table gives the best measured result for each modality on the three COSMOS server platforms, with the date and image of the measurement. The column to compare is the RAN server's CPU platform. Host names are deliberately left out. Each linked tutorial's own Performance section has every run and the method used.
| CPU platform | Cores / threads |
|---|---|
| A — 2× Intel Xeon Gold 6226 @ 2.70 GHz (Cascade Lake) | 24 cores / 24 threads (HT off) |
| B — 2× Intel Xeon Gold 6126 @ 2.60 GHz (Skylake-SP) | 24 cores / 48 threads (HT on) |
| C — 1× AMD EPYC 9355P (Zen 5) | 32 cores / 32 threads (SMT off), one NUMA node |
| Modality | Radio / UE | Platform | DL, Mbit/s | UL, Mbit/s | Date · image |
|---|---|---|---|---|---|
| Soft-only | — | A, B, C | not applicable — no soft-only tutorial | — | — |
| Over-the-air, 5G SA (n41, 40 MHz TDD) | lteue, 1 UE |
A | 10–66 (median 16) | 0.6–17 (median 4.1) | 2026-09-13 · amarisoft-20260904 + lte5gue-20260905 (8 runs) |
| B | 14–76 (median 57) | 3.9–21 (median 4.4) | 2026-09-13 · same images (5 runs) | ||
| C | not measured — rrail has no USRP of its own | — | — | ||
| Over-the-air, 5G SA (n41, 100 MHz TDD) | lteue on an N310 |
not recorded | ≈ 132 | ≈ 24 | 2026-07-15 · amarisoft.ndz of that date |
| Over-the-air, LTE (10 MHz FDD) | lteue, 1 UE |
A | 37–47 | 17–20 | 2026-09-13 · amarisoft-20260904 + lte5gue-20260905 (6 runs) |
| B | 37–47 | 16.5–20 | 2026-09-13 · same images (3 runs) | ||
| O-RAN 7.2, vendor O-RU | — | A, B, C | not measured — no standardized tutorial | — | — |
| O-RAN 7.2, soft O-RU | — | A, B, C | not applicable | — | — |
UDP capacity is the receiver-side rate at the highest offered load with less than 5 % loss. LTE shows no platform effect. In 5G, the run-to-run spread on each platform is larger than the difference between the platforms; the capacity-sweep tutorial explains why. The 100 MHz figure predates the rule of recording the CPU platform and is due for a re-measurement.
omf-amarisoft. RAN configuration: skill amarisoft-ran. UE: skill quectel-5g-modem.Author(s): COSMOS team · Last verified: 2026-09-14 (landing page; see each linked tutorial for its own verification date; the performance table is compiled from those tutorials' measurements) · Tested image/release: amarisoft-20260904.ndz + lte5gue-20260905.ndz, Amarisoft 2026-06-12 · Tags: amarisoft, lte, 5g, sa, overview