ℹ

Public note: Dell PowerEdge R820 . 2x AMD Radeon RX 7900 XTX (48 GB VRAM gesamt, ROCm/HIP) . RAM: 32x 16 GB SK Hynix DDR3-1600 ECC RDIMM (HMT42GR7AFR4A-PB) = 512 GB @ 1600 MT/s

Generation2,46tok/s

Prefill96,03tok/s

Time to First Token123.967,00ms

Total duration1.200,00s

Concurrency5parallel

Ranking in the field

This GPU All GPUs

1443of 1469 systems

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

This run is better than 2 % of all comparable systems.

Generation 2,5 tok/s

-99 % vs Ø 246,5

Prefill 96,0 tok/s

-98 % vs Ø 5.365,8

Time to First Token 123.967 ms

+311 % 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

Mistral-Medium-3.5-128B this run2× 2x AMD Radeon RX 7900 XTX · run-20261009-090853-181fcc

2,5 tok/s

How does this benchmark compare on other GPUs?

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

NVIDIA RTX PRO 6000 Blackwell Workstation Editionrun-20260729-105332-853af7

73,6 tok/s

NVIDIA GeForce RTX 5070 Tirun-20260729-150924-1f3042

4,2 tok/s

NVIDIA GeForce RTX 3090 Tirun-20260731-071559-a98068

2,8 tok/s

2× AMD Radeon RX 7900 XTX this runrun-20261009-090853-181fcc

2,5 tok/s

NVIDIA GeForce RTX 5090run-20260806-094440-e0f588

1,9 tok/s

NVIDIA GeForce RTX 5090run-20260806-094437-d31d85

1,9 tok/s

Hardware

GPU: 2x AMD Radeon RX 7900 XTX · 24 GB VRAM
CPU: 4x Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz
RAM: 504 GB
Mainboard: Dell Inc. PowerEdge R820

Setup

Runtime: llama.cpp
Quantization: Q4_K_M
Model: Mistral-Medium-3.5-128B

Configuration

benchmark-konfiguration — run-20261009-090853-181fcc

# LLM-Benchmark Konfiguration # Modell : Mistral-Medium-3.5-128B # Engine : llama.cpp # Run-ID : run-20261009-090853-181fcc # GPU : 2x AMD Radeon RX 7900 XTX # CPU : 4x Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz # RAM : 504 GB bench@llm-benchmark:~$ llama-server \ -m mistralai_Mistral-Medium-3.5-128B-IQ4_XS/mistralai_Mistral-Medium-3.5-128B-IQ4_XS-00001-of-00002.gguf \ --alias Mistral-Medium-3.5-128B \ -ngl 40 \ -sm layer \ -fa on \ -c 40960 \ -np 10 '(2x' RX 7900 XTX ROCm/HIP, 'dense-ngl40)'

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.	Mistral-Medium-3.5-128B
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.	mistralai_Mistral-Medium-3.5-128B-IQ4_XS/mistralai_Mistral-Medium-3.5-128B-IQ4_XS-00001-of-00002.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.	40
Split-Mode?Verteilung ueber mehrere GPUs: none (nur eine GPU), layer (Layer aufteilen) oder row (Tensoren zeilenweise).	layer
fa	on
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.	40960
np	10
execution_type	local

All benchmarks of this model To leaderboard

Model comparison

Mistral-Medium-3.5-128B 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 141,6 tok/sAMD Radeon AI PRO R9700 5,8 tok/sNVIDIA GeForce RTX 5070 Ti 4,2 tok/sNVIDIA GeForce RTX 3090 Ti 2,8 tok/s★ AMD Radeon RX 7900 XTX 2,5 tok/s this runNVIDIA GeForce RTX 5090 1,9 tok/s

CPUby processor

AMD Ryzen 9 9950X 16-Core Processor 141,6 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 5,8 tok/sAMD Ryzen Threadripper PRO 5975WX 32-Cores 4,2 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 2,8 tok/s★ Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz 2,5 tok/s this runAMD Ryzen 7 5800X3D 8-Core Processor 1,9 tok/s

MBby mainboard

ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 141,6 tok/sASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 5,8 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 4,2 tok/sMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 2,8 tok/s★ Dell Inc. PowerEdge R820 2,5 tok/s this runASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 1,9 tok/s

ENGby engine

★ llama.cpp 141,6 tok/s this rununbekannt 5,8 tok/s

DRVby driver

unbekannt 141,6 tok/sAMD 7.0.0-27-generic 5,8 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 URL Reset

⚡ Electricity price EUR/kWh

⚙️ System utilization 100 %

🖥️ Acquisition EUR

🔌 Idle 35 W

⚡ TDP 749 W

☁️ External LLM (API) show

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

Electricity0.30 EUR/kWh

Avg power (incl. idle)749 W estimated (TDP)GPU 710 + CPU 29 + Board 10 W full load

Avg cost / hourEUR 0.22

Electricity / 1M tokensEUR 25.36

Token / kWh11.83K

Acquisition (system)EUR 4,353 partial priceGPU EUR 2,098 · CPU EUR 59 · RAM EUR 2,016 · PSU EUR 180

Electricity (2 years)–

TCO (2 years)EUR 8,288

Output tokens (2 years)155.16M

☁️ 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 (35 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

	Mistral-Medium-3.5-128B2x AMD Radeon RX 7900 XTX	Mistral-Medium-3.5-128BNVIDIA RTX PRO 6000 Blackwell Workstation Edition	Mistral-Medium-3.5-128BNVIDIA GeForce RTX 5070 Ti	Mistral-Medium-3.5-128BNVIDIA GeForce RTX 3090 Ti
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)	–	–	–	–

2× AMD Radeon RX 7900 XTX · this run 2 tok/s NVIDIA RTX PRO 6000 Blackwell Workstation Edition 74 tok/s NVIDIA GeForce RTX 5070 Ti 4 tok/s NVIDIA GeForce RTX 3090 Ti 3 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.