2 × W7900 LLM benchmarks

4 results · 4 measurements · 1 source

Explore all results Image benchmarks Submit a result

Device specifications

Variant	Memory	Memory type	Bus	Bandwidth	FP16 matrix	INT8 matrix
W7900 Dual Slot ↗	48 GB	GDDR6	384-bit	864 GB/s	123 TFLOPS	123 TOPS
W7900 ↗	48 GB	GDDR6	384-bit	864 GB/s	123 TFLOPS	123 TOPS

Manufacturer specifications. Shared system RAM is not dedicated VRAM. Unreported values remain —; measured LLM results retain their reported configuration.

Selected reported tests

4 of 4 results

Source-reported results, selected across reports by documentation quality. Match configurations before comparing values.


View Reportllama-bench	Llama 3 70B (checkpoint unspecified)Q8_0Model ref. ↗	Token speed	—	263.18 tok/sPrefill · per request	—	2 × W7900Layer split · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Linux Engine llama.cpp / llama-bench Backend Not reported Harness llama-bench Prompt / output 512 / ? tokens Context Not reported Concurrency Not reported Date 2024-07-11 Full record →
View Reportllama-bench	Llama 3 70B (checkpoint unspecified)Q8_0Model ref. ↗	Token speed	—	—	9.25 tok/sGeneration · per request	2 × W7900Layer split · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Linux Engine llama.cpp / llama-bench Backend Not reported Harness llama-bench Prompt / output ? / 1024 tokens Context Not reported Concurrency Not reported Date 2024-07-11 Full record →
View Reportllama-bench	Llama 3 70B (checkpoint unspecified)Q8_0Model ref. ↗	Token speed	—	108.18 tok/sPrefill · per request	—	2 × W7900Row split · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Linux Engine llama.cpp / llama-bench Backend Not reported Harness llama-bench Prompt / output 512 / ? tokens Context Not reported Concurrency Not reported Date 2024-07-11 Full record →
View Reportllama-bench	Llama 3 70B (checkpoint unspecified)Q8_0Model ref. ↗	Token speed	—	—	12.85 tok/sGeneration · per request	2 × W7900Row split · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Linux Engine llama.cpp / llama-bench Backend Not reported Harness llama-bench Prompt / output ? / 1024 tokens Context Not reported Concurrency Not reported Date 2024-07-11 Full record →

Explore all 4 results →

Models and software in this collection

Model families: Llama 3

Engines: llama.cpp / llama-bench

Methodology and evidence standards · Download the research dataset

2 × W7900 LLM benchmarks

Selected reported tests

LLMs benchmarked on the W7900