Created bymario-alka.dePowered bygodcore.denoob2claw.detricoma.de
Contributed byMario AlkaGoogle

gemma-2-2b-it

Performance benchmark · measured on 08.10.2026 07:04

Benchmark-IDrun-20261008-053613-fa5272
Timebench 3 - Kombi (Prefill + Generation)Dense2BRuntime: llama.cppQuantisierung: Q4_K_M
ℹ
Public note: Arduino Q (ventunoq) . ARM RK3588 8-Core (4x Cortex-A78C + 4x Cortex-A55) . 14 GB RAM . llama.cpp CPU (Edge-Geraet)
Generation7,49tok/s
Prefill35,16tok/s
Time to First Token283.919,00ms
Total duration1.198,42s
Concurrency5parallel
Ranking in the field
1389of 1469 systems

Performance benchmark · Primary metric: Generation-Speed (tok/s) · 5× concurrent

This run is better than 5 % of all comparable systems.
Generation 7,5 tok/s
-97 % vs Ø 246,5
Prefill 35,2 tok/s
-99 % vs Ø 5.365,8
Time to First Token 283.919 ms
+842 % vs Ø 30.155
Distribution in the field0 – 1.349 tok/s
Ø 247 Median Dieser Lauf

Wie schlägt sich dieser Benchmark mit anderen Modellen?

gemma-4-E2B-itNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181606-0adf6d
1.349,3 tok/s
gemma-4-E2B-itNVIDIA GeForce RTX 5090 · run-20260728-184455-5b4937
1.301,9 tok/s
gemma-4-E2B-it3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-184456-4111a9
1.195,7 tok/s
Nemotron-3-Nano-4BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181607-30d864
1.188,7 tok/s
Nemotron-3-Nano-4BNVIDIA GeForce RTX 5090 · run-20260728-194135-5432cd
1.164,8 tok/s
gpt-oss-20bNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260728-140954-9e81f9
1.136,3 tok/s
gpt-oss-20bNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260730-035052-09627f
1.135,0 tok/s
Nemotron-3.5-Lightning-30B-A3BNVIDIA GeForce RTX 3090 Ti · run-20261021-093815-b22926
1.135,0 tok/s
gpt-oss-20bNVIDIA GeForce RTX 5090 · run-20260729-032121-058a31
1.113,5 tok/s
Nemotron-3-Nano-4B3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-836a91
1.090,6 tok/s
Nemotron-3-Nano-4B3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260727-181606-8b381e
1.077,4 tok/s
Nemotron-3-Nano-4B3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-b8a818
1.071,0 tok/s
Nemotron-3-Nano-4B3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-038e92
1.051,1 tok/s
gpt-oss-20b3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260729-032121-ffb18a
1.015,5 tok/s
gemma-2-2b-it this runKeine GPU (CPU-only) · run-20261008-053613-fa5272
7,5 tok/s

How does this benchmark compare on other GPUs?

Same model on different hardware · 5× concurrent · Generation (tok/s)

Hardware

GPU: Keine GPU (CPU-only)
CPU: 4 x Cortex-A78C + 4 x Cortex-A55
RAM: 15 GB

Setup

Runtime: llama.cpp
Quantization: Q4_K_M
Model: gemma-2-2b-it

Configuration

benchmark-konfiguration — run-20261008-053613-fa5272
# LLM-Benchmark Konfiguration # Modell : gemma-2-2b-it # Engine : llama.cpp # Run-ID : run-20261008-053613-fa5272 # GPU : Keine GPU (CPU-only) # CPU : 4 x Cortex-A78C + 4 x Cortex-A55 # RAM : 15 GB bench@llm-benchmark:~$ llama-server \ -m gemma-2-2b-it-Q4_K_M.gguf \ --alias gemma-2-2b-it \ -ngl 0 \ -t 8 \ -c 40960 \ -np 10 '(ARM' RK3588 'CPU)'
Engine?Die Inferenz-Software, die das Modell ausliefert (z.B. vLLM oder llama.cpp). Sie bestimmt Geschwindigkeit, unterstuetzte Modellformate und welche Parameter ueberhaupt verfuegbar sind.llamacpp
Modellalias?Der Name, unter dem das Modell ueber die API angesprochen wird. Genau dieser Wert muss im Request-Feld 'model' stehen.gemma-2-2b-it
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.40960
Modellpfad?Pfad zur GGUF-Modelldatei, die geladen und ausgeliefert wird.gemma-2-2b-it-Q4_K_M.gguf
GPU-Layer?Anzahl der auf die GPU ausgelagerten Modell-Layer. Hoeher = mehr VRAM und schneller; der Rest laeuft auf der CPU. 999 = alles auf GPU.0
Threads?Anzahl CPU-Threads fuer die Token-Generierung (Decode).8
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.40960
np10
execution_typelocal

All benchmarks of this model To leaderboard

Anzeige
Model comparison

gemma-2-2b-it on various hardware

All published performance runs of this model – each bubble a variant: position = prefill (X) × generation (Y), bubble size = number of runs. Closer to the top right = faster. ★ Marked gold = this benchmark.

GPUby graphics card

4193152101050,004.9069.81214.717Prefill (tok/s)Generation (tok/s)AMD Radeon RX 7900 XTX - 323,8 tok/s Generation, 11.875 tok/s Prefill, TTFT 1.279 ms (6 Laufe)AMD Radeon RX 7900 XTXCPU-only - 7,9 tok/s Generation, 40 tok/s Prefill, TTFT 249.190 ms (3 Laufe) | DIESER LAUF★ CPU-only
AMD Radeon RX 7900 XTX 323,8 tok/s★ CPU-only 7,9 tok/s this run

CPUby processor

4193152101050,007.79715.59423.391Prefill (tok/s)Generation (tok/s)Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz - 323,8 tok/s Generation, 4.880 tok/s Prefill, TTFT 1.976 ms (3 Laufe)Intel(R) Xeon(R) CPU ...AMD Ryzen Threadripper PRO 3955WX 16-Cores - 322,9 tok/s Generation, 18.870 tok/s Prefill, TTFT 583 ms (3 Laufe)AMD Ryzen Threadrippe...4 x Cortex-A78C + 4 x Cortex-A55 - 7,9 tok/s Generation, 40 tok/s Prefill, TTFT 249.190 ms (3 Laufe) | DIESER LAUF★ 4 x Cortex-A78C + 4 x...
Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz 323,8 tok/sAMD Ryzen Threadripper PRO 3955WX 16-Cores 322,9 tok/s★ 4 x Cortex-A78C + 4 x Cortex-A55 7,9 tok/s this run

MBby mainboard

4193152101050,007.79715.59423.391Prefill (tok/s)Generation (tok/s)Dell Inc. PowerEdge R820 - 323,8 tok/s Generation, 4.880 tok/s Prefill, TTFT 1.976 ms (3 Laufe)Dell Inc. PowerEdge R...ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 322,9 tok/s Generation, 18.870 tok/s Prefill, TTFT 583 ms (3 Laufe)ASUSTeK COMPUTER INC....unbekannt - 7,9 tok/s Generation, 40 tok/s Prefill, TTFT 249.190 ms (3 Laufe)unbekannt
Dell Inc. PowerEdge R820 323,8 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 322,9 tok/sunbekannt 7,9 tok/s

ENGby engine

35634032330729007.65215.30422.956Prefill (tok/s)Generation (tok/s)vLLM - 322,9 tok/s Generation, 18.870 tok/s Prefill, TTFT 583 ms (3 Laufe)vLLMllama.cpp - 323,8 tok/s Generation, 2.460 tok/s Prefill, TTFT 125.583 ms (6 Laufe) | DIESER LAUF★ llama.cpp
★ llama.cpp 323,8 tok/s this runvLLM 322,9 tok/s

DRVby driver

3563403243082917.4547.7718.0888.405Prefill (tok/s)Generation (tok/s)unbekannt - 323,8 tok/s Generation, 7.930 tok/s Prefill, TTFT 83.916 ms (9 Laufe)unbekannt
unbekannt 323,8 tok/s
💰 Economics

Economics of this run

Operating cost, TCO and comparison with the next-best runs of the same model at identical concurrency (5× concurrent). Methodology →

⚙️ ConfigurationAll metrics and charts below follow these settings – based on a 24-month runtime.Save to URLReset
⚡ Electricity price EUR/kWh
⚙️ System utilization 100 %
🔌 Idle 0 W
⚡ TDP 10 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)0 W missing
Avg cost / hour–
Electricity / 1M tokens–
Token / kWh–
Acquisition (system)– partial price
Electricity (2 years)–
TCO (2 years)–
Output tokens (2 years)472.41M
☁️ External LLM (API) – comparison
External LLM cost (2 years)–
Savings vs. external (2 years)–
No power draw measured – values estimated from GPU TDP + CPU (idle + 15 %) + board.

All values above and the charts below take the configured system utilization into account: at X% the system generates only X% of the time, the rest it idles (0 W). Cost per hour drops (more idle), cost per token rises.

Cost over 2 years – electricity only

Cost over 2 years – incl. acquisition (TCO)

Speed vs. tokens per euro

Euro per 1M tokens

Comparison vs. API – economics per benchmark

gemma-2-2b-itKeine GPU (CPU-only)gemma-2-2b-it2x AMD Radeon RX 7900 XTXgemma-2-2b-it2x AMD Radeon RX 7900 XTX
Electricity cost (24 mo.)–––
Acquisition cost–––
Total cost (TCO)–––
Generated tokens (24 mo.)–––
Token price via API–––
Break-even point (days)–––
Result (savings / extra cost)–––

Comparison with up to 3 next-best runs of this model at the same concurrency (at least one on different hardware). Power = GPU TDP + CPU (idle + 15 %) + board (estimated), acquisition = full system (GPU + CPU + board + RAM + PSU), prices = stored market prices.

Contributed by

Mario Alka Administrator

@marioalka

Ich bin Unternehmer, Softwareentwickler und KI-Enthusiast. Seit vielen Jahren entwickle ich Unternehmenssoftware und beschäftige mich inzwischen fast täglich mit lokalen LLMs, KI-Agenten und leistungsfähiger KI-Hardware.

Mit LLM-Benchmark.de möchte ich eine Plattform schaffen, auf der Modelle, GPUs und Agenten objektiv und reproduzierbar miteinander verglichen werden.