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

Meta-Llama-3.1-8B-Instruct

Performance benchmark · measured on 22.08.2026 08:37

Benchmark-IDrun-20260822-084953-8baabe
Timebench 3 - Kombi (Prefill + Generation)Dense8BRuntime: llama.cppQuantisierung: Q4_K_M
Generation48,95tok/s
Prefill1.278,20tok/s
Time to First Token1.931,50ms
Total duration45,71s
Concurrency1parallel
Ranking in the field
5of 15 systems

Performance benchmark · Primary metric: Generation-Speed (tok/s) · 1× concurrent

This run is better than 71 % of all comparable systems.
Generation 49,0 tok/s
+13 % vs Ø 43,5
Prefill 1.278,2 tok/s
+0 % vs Ø 1.275,6
Time to First Token 1.932 ms
+6 % vs Ø 1.820
Distribution in the field14 – 81 tok/s
Ø 43 Median Dieser Lauf

Wie schlägt sich dieser Benchmark mit anderen Modellen?

How does this benchmark compare on other GPUs?

Same model on different hardware · 1× concurrent · Generation (tok/s)

Hardware

GPU: 2x NVIDIA GeForce RTX 2060 · 6 GB VRAM
CPU: Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz
RAM: 7 GB
Mainboard: ASRock H110 Pro BTC+

Setup

Runtime: llama.cpp
Quantization: Q4_K_M
Model: Meta-Llama-3.1-8B-Instruct

Anmerkung

PCIe-Bandbreiten-Test (identische RTX 2060, Single-GPU): Karte #0 lief an PCIe Gen3 x16 = ~15,75 GB/s effektive Bandbreite. Gegenprobe: Karte #0 an PCIe Gen3 x16 (~15,75 GB/s) vs Karte #1 an PCIe Gen2 x1 (~0,5 GB/s, x1-Mining-Riser) = ~31x Bandbreiten-Unterschied. Frage: bremst langsames PCIe die reine Inferenz-Geschwindigkeit (tok/s) bei Single-GPU?

Configuration

benchmark-konfiguration — run-20260822-084953-8baabe
# LLM-Benchmark Konfiguration # Modell : Meta-Llama-3.1-8B-Instruct # Engine : llama.cpp # Run-ID : run-20260822-084953-8baabe # GPU : 2x NVIDIA GeForce RTX 2060 # CPU : Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz # RAM : 7 GB bench@llm-benchmark:~$ /home/godcore/llama.cpp/build/bin/llama-server \ -m Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf \ -ngl 999 \ -c 4000 \ -np 1 CUDA_VISIBLE_DEVICES=0
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
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.4000
Modellpfad?Pfad zur GGUF-Modelldatei, die geladen und ausgeliefert wird.Meta-Llama-3.1-8B-Instruct-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.999
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.4000
np1

All benchmarks of this model To leaderboard

Anzeige
Model comparison

Meta-Llama-3.1-8B-Instruct 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

2.0701.5531.0355180,005.87311.74617.619Prefill (tok/s)Generation (tok/s)NVIDIA GeForce RTX 5090 - 1.593,1 tok/s Generation, 11.858 tok/s Prefill, TTFT 3.236 ms (3 Laufe)NVIDIA GeForce RTX 50...NVIDIA RTX PRO 6000 Blackwell Workstation Edition - 1.533,0 tok/s Generation, 13.472 tok/s Prefill, TTFT 2.895 ms (3 Laufe)NVIDIA RTX PRO 6000 B...NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition - 1.433,0 tok/s Generation, 14.214 tok/s Prefill, TTFT 3.334 ms (9 Laufe)NVIDIA RTX PRO 6000 B...NVIDIA GeForce RTX 5070 Ti - 913,4 tok/s Generation, 6.511 tok/s Prefill, TTFT 5.168 ms (3 Laufe)NVIDIA GeForce RTX 50...NVIDIA GeForce RTX 3090 Ti - 837,9 tok/s Generation, 8.047 tok/s Prefill, TTFT 5.721 ms (3 Laufe)NVIDIA GeForce RTX 30...AMD Radeon AI PRO R9700 - 518,3 tok/s Generation, 8.195 tok/s Prefill, TTFT 7.711 ms (10 Laufe)AMD Radeon AI PRO R97...NVIDIA Tesla P100 PCIe 16GB - 67,6 tok/s Generation, 864 tok/s Prefill, TTFT 15.746 ms (3 Laufe)NVIDIA Tesla P100 PCI...Intel Arc Pro B70 - 66,4 tok/s Generation, 860 tok/s Prefill, TTFT 2.539 ms (2 Laufe)Intel Arc Pro B70CPU-only - 3,6 tok/s Generation, 36 tok/s Prefill, TTFT 59.044 ms (1 Lauf)CPU-onlyNVIDIA GeForce RTX 2060 - 130,9 tok/s Generation, 2.353 tok/s Prefill, TTFT 3.410 ms (7 Laufe) | DIESER LAUF★ NVIDIA GeForce RTX 20...
NVIDIA GeForce RTX 5090 1.593,1 tok/sNVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.533,0 tok/sNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 1.433,0 tok/sNVIDIA GeForce RTX 5070 Ti 913,4 tok/sNVIDIA GeForce RTX 3090 Ti 837,9 tok/sAMD Radeon AI PRO R9700 518,3 tok/s★ NVIDIA GeForce RTX 2060 130,9 tok/s this runNVIDIA Tesla P100 PCIe 16GB 67,6 tok/sIntel Arc Pro B70 66,4 tok/sCPU-only 3,6 tok/s

CPUby processor

2.0701.5531.0355180,005.87311.74617.619Prefill (tok/s)Generation (tok/s)AMD Ryzen 7 5800X3D 8-Core Processor - 1.593,1 tok/s Generation, 11.858 tok/s Prefill, TTFT 3.236 ms (3 Laufe)AMD Ryzen 7 5800X3D 8...AMD Ryzen 9 9950X 16-Core Processor - 1.533,0 tok/s Generation, 13.472 tok/s Prefill, TTFT 2.895 ms (3 Laufe)AMD Ryzen 9 9950X 16-...AMD Ryzen Threadripper PRO 9965WX 24-Cores - 1.433,0 tok/s Generation, 14.214 tok/s Prefill, TTFT 3.334 ms (9 Laufe)AMD Ryzen Threadrippe...AMD Ryzen Threadripper PRO 5975WX 32-Cores - 913,4 tok/s Generation, 6.511 tok/s Prefill, TTFT 5.168 ms (3 Laufe)AMD Ryzen Threadrippe...AMD Ryzen 9 8945HX with Radeon Graphics - 837,9 tok/s Generation, 8.047 tok/s Prefill, TTFT 5.721 ms (3 Laufe)AMD Ryzen 9 8945HX wi...AMD Ryzen Threadripper PRO 7955WX 16-Cores - 518,3 tok/s Generation, 8.195 tok/s Prefill, TTFT 7.711 ms (10 Laufe)AMD Ryzen Threadrippe...AMD Ryzen 9 7945HX with Radeon Graphics - 67,6 tok/s Generation, 862 tok/s Prefill, TTFT 10.463 ms (5 Laufe)AMD Ryzen 9 7945HX wi...Intel(R) Xeon(R) CPU E5-4620 v2 @ 2.60GHz - 3,6 tok/s Generation, 36 tok/s Prefill, TTFT 59.044 ms (1 Lauf)Intel(R) Xeon(R) CPU ...Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz - 130,9 tok/s Generation, 2.353 tok/s Prefill, TTFT 3.410 ms (7 Laufe) | DIESER LAUF★ Intel(R) Core(TM) i5-...
AMD Ryzen 7 5800X3D 8-Core Processor 1.593,1 tok/sAMD Ryzen 9 9950X 16-Core Processor 1.533,0 tok/sAMD Ryzen Threadripper PRO 9965WX 24-Cores 1.433,0 tok/sAMD Ryzen Threadripper PRO 5975WX 32-Cores 913,4 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 837,9 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 518,3 tok/s★ Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz 130,9 tok/s this runAMD Ryzen 9 7945HX with Radeon Graphics 67,6 tok/sIntel(R) Xeon(R) CPU E5-4620 v2 @ 2.60GHz 3,6 tok/s

MBby mainboard

2.0701.5531.0355180,005.56611.13216.698Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING - 1.593,1 tok/s Generation, 11.858 tok/s Prefill, TTFT 3.236 ms (3 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI - 1.533,0 tok/s Generation, 13.472 tok/s Prefill, TTFT 2.895 ms (3 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE - 1.433,0 tok/s Generation, 11.046 tok/s Prefill, TTFT 5.638 ms (19 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 913,4 tok/s Generation, 6.511 tok/s Prefill, TTFT 5.168 ms (3 Laufe)ASUSTeK COMPUTER INC....Meigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) - 837,9 tok/s Generation, 8.047 tok/s Prefill, TTFT 5.721 ms (3 Laufe)Meigao Innovation Tec...Shenzhen Meigao Electronic Equipment Co.,Ltd F1FXM (DeskMini Series) - 67,6 tok/s Generation, 864 tok/s Prefill, TTFT 15.746 ms (3 Laufe)Shenzhen Meigao Elect...Shenzhen Meigao Electronic Equipment Co.,Ltd DRFXI (MotherBoard Series) - 66,4 tok/s Generation, 860 tok/s Prefill, TTFT 2.539 ms (2 Laufe)Shenzhen Meigao Elect...Dell Inc. PowerEdge R820 - 3,6 tok/s Generation, 36 tok/s Prefill, TTFT 59.044 ms (1 Lauf)Dell Inc. PowerEdge R...ASRock H110 Pro BTC+ - 130,9 tok/s Generation, 2.353 tok/s Prefill, TTFT 3.410 ms (7 Laufe) | DIESER LAUF★ ASRock H110 Pro BTC+
ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 1.593,1 tok/sASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 1.533,0 tok/sASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 1.433,0 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 913,4 tok/sMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 837,9 tok/s★ ASRock H110 Pro BTC+ 130,9 tok/s this runShenzhen Meigao Electronic Equipment Co.,Ltd F1FXM (DeskMini Series) 67,6 tok/sShenzhen Meigao Electronic Equipment Co.,Ltd DRFXI (MotherBoard Series) 66,4 tok/sDell Inc. PowerEdge R820 3,6 tok/s

ENGby engine

2.0631.5471.0315160,003.3116.6219.932Prefill (tok/s)Generation (tok/s)unbekannt - 42,1 tok/s Generation, 821 tok/s Prefill, TTFT 2.298 ms (1 Lauf)unbekanntllama.cpp - 1.593,1 tok/s Generation, 8.129 tok/s Prefill, TTFT 6.770 ms (43 Laufe) | DIESER LAUF★ llama.cpp
★ llama.cpp 1.593,1 tok/s this rununbekannt 42,1 tok/s

DRVby driver

2.0581.5431.0295140,003.3796.75910.138Prefill (tok/s)Generation (tok/s)unbekannt - 1.593,1 tok/s Generation, 8.301 tok/s Prefill, TTFT 6.865 ms (42 Laufe)unbekanntIntel 26.18.38308.4 - 66,4 tok/s Generation, 860 tok/s Prefill, TTFT 2.539 ms (2 Laufe)Intel 26.18.38308.4
unbekannt 1.593,1 tok/sIntel 26.18.38308.4 66,4 tok/s
💰 Economics

Economics of this run

Operating cost, TCO and comparison with the next-best runs of the same model at identical concurrency (1× 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) missing
Electricity (2 years)
TCO (2 years)
Output tokens (2 years)3.09B
☁️ 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

Meta-Llama-3.1-8B-Instruct2x NVIDIA GeForce RTX 2060Meta-Llama-3.1-8B-InstructNVIDIA GeForce RTX 5090Meta-Llama-3.1-8B-InstructNVIDIA RTX PRO 6000 Blackwell Workstation EditionMeta-Llama-3.1-8B-Instruct3x 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)

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.