# LLM-Benchmark Konfiguration # Modell : Llama-3_3-Nemotron-Super-49B-v1_5 # Engine : llama.cpp # Run-ID : run-20260729-105333-24b92d # GPU : NVIDIA GeForce RTX 5070 Ti # CPU : AMD Ryzen Threadripper PRO 5975WX 32-Cores # RAM : 247 GB bench@llm-benchmark:~$ /home/godcore/llama.cpp/build/bin/llama-server \ -m /home/godcore/.cache/huggingface/hub/models--gabriellarson--Llama-3_3-Nemotron-Super-49B-v1_5-GGUF/snapshots/c324e134daf0a0d4f9e3e136ae525df9fdebea3b/Llama-3_3-Nemotron-Super-49B-v1_5-Q4_K_M.gguf \ --alias Llama-3_3-Nemotron-Super-49B-v1_5 \ --host 0.0.0.0 \ --port 8000 \ -ngl 30 \ -c 16384 \ -np 4 \ --jinja

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.	Llama-3_3-Nemotron-Super-49B-v1_5
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.	16384
Modellpfad?Pfad zur GGUF-Modelldatei, die geladen und ausgeliefert wird.	/home/godcore/.cache/huggingface/hub/models--gabriellarson--Llama-3_3-Nemotron-Super-49B-v1_5-GGUF/snapshots/c324e134daf0a0d4f9e3e136ae525df9fdebea3b/Llama-3_3-Nemotron-Super-49B-v1_5-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.	30
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.	16384
np	4
execution_type	local

All benchmarks of this model To leaderboard

Model comparison

Llama-3_3-Nemotron-Super-49B-v1_5 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

NVIDIA RTX PRO 6000 Blackwell Workstation Edition 364,1 tok/sNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 303,0 tok/sNVIDIA RTX A6000 197,3 tok/sAMD Radeon RX 7900 XTX 108,3 tok/sAMD Radeon PRO W7900 Dual Slot 71,0 tok/s★ NVIDIA GeForce RTX 5070 Ti 25,2 tok/s this runNVIDIA GeForce RTX 5090 20,2 tok/sNVIDIA GeForce RTX 3090 Ti 19,6 tok/sAMD Radeon AI PRO R9700 11,6 tok/sNVIDIA GeForce RTX 2060 1,1 tok/s

CPUby processor

AMD Ryzen 9 9950X 16-Core Processor 364,1 tok/sAMD Ryzen Threadripper PRO 9965WX 24-Cores 303,0 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 197,3 tok/sIntel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz 108,3 tok/s★ AMD Ryzen Threadripper PRO 5975WX 32-Cores 71,0 tok/s this runAMD Ryzen 7 5800X3D 8-Core Processor 20,2 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 19,6 tok/s

MBby mainboard

ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 364,1 tok/sASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 303,0 tok/sDell Inc. PowerEdge R820 108,3 tok/s★ ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 71,0 tok/s this runASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 20,2 tok/sMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 19,6 tok/s

ENGby engine

★ llama.cpp 364,1 tok/s this runvLLM 197,3 tok/s

DRVby driver

unbekannt 364,1 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 URL Reset

⚡ Electricity price EUR/kWh

⚙️ System utilization 100 %

🖥️ Acquisition EUR

🔌 Idle 10 W

⚡ TDP 310 W

☁️ External LLM (API) show

Input €/1M Output €/1M Input per output token

Electricity0.30 EUR/kWh

Avg power (incl. idle)310 W estimated (TDP)GPU 300 + Board 10 W full load

Avg cost / hourEUR 0.093

Electricity / 1M tokensEUR 1.83

Token / kWh163.97K

Acquisition (system)EUR 5,000 partial priceGPU EUR 994 · CPU EUR 1,880 · RAM EUR 1,976 · PSU EUR 150

Electricity (2 years)–

TCO (2 years)EUR 6,629

Output tokens (2 years)890.58M

☁️ External LLM (API) – comparison

External LLM cost (2 years)–

Savings vs. external (2 years)–

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 (10 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

	Llama-3_3-Nemotron-Super-49B-v1_5NVIDIA GeForce RTX 5070 Ti	Llama-3_3-Nemotron-Super-49B-v1_5NVIDIA RTX PRO 6000 Blackwell Workstation Edition	Llama-3_3-Nemotron-Super-49B-v1_53x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition	Llama-3_3-Nemotron-Super-49B-v1_53x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition
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)	–	–	–	–

NVIDIA GeForce RTX 5070 Ti · this run 14 tok/s NVIDIA RTX PRO 6000 Blackwell Workstation Edition 190 tok/s 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 160 tok/s 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 158 tok/s

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.