Hugging Face's logo

Datasets:
witcheer
/
rtx-5090-benchmarks

Tasks:
Text Generation
Modalities:
Image
Tabular
Text
Formats:
csv
Size:
< 1K
Tags:
benchmark
inference
llm
nvidia
rtx-5090
llama-cpp
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
witcheer commited on
Commit
5451aec
·
verified ·
1 Parent(s): 8c54ac6

Upload benchmarks.csv with huggingface_hub

Browse files
Files changed (1) hide show
  1. benchmarks.csv +2 -0
benchmarks.csv CHANGED
@@ -299,3 +299,5 @@ Qwen3.8-27B,Dense (hybrid attn),27.32,Q8_0,27.04,llama.cpp,CUDA,RTX 5090,32,pp40
299
  Qwen3.8-27B,Dense (hybrid attn),27.32,Q8_0,27.04,llama.cpp,CUDA,RTX 5090,32,pp8192,3680.71,5.43,2026-08-15
300
  Qwen3.8-27B,Dense (hybrid attn),27.32,Q8_0,27.04,llama.cpp,CUDA,RTX 5090,32,pp16384,3543.55,2.17,2026-08-15
301
  Qwen3.8-27B,Dense (hybrid attn),27.32,Q8_0,27.04,llama.cpp,CUDA,RTX 5090,32,tg128,53.07,0.06,2026-08-15
 
 
 
299
  Qwen3.8-27B,Dense (hybrid attn),27.32,Q8_0,27.04,llama.cpp,CUDA,RTX 5090,32,pp8192,3680.71,5.43,2026-08-15
300
  Qwen3.8-27B,Dense (hybrid attn),27.32,Q8_0,27.04,llama.cpp,CUDA,RTX 5090,32,pp16384,3543.55,2.17,2026-08-15
301
  Qwen3.8-27B,Dense (hybrid attn),27.32,Q8_0,27.04,llama.cpp,CUDA,RTX 5090,32,tg128,53.07,0.06,2026-08-15
302
+ Qwen3.8-27B,Dense (hybrid attn),27.32,NVFP4,21.02,vllm-0.25.1,CUDA,RTX 5090,32,tg128,68.95,,2026-08-17
303
+ Qwen3.8-27B,Dense (hybrid attn),27.32,NVFP4+MTP,21.81,vllm-0.25.1,CUDA,RTX 5090,32,tg128,119.79,,2026-08-17