Qwen3.8-27B W4A16-AWQ on RTX 4090
safetensors INT4_G128_ZP+BF16
Engines
vLLM
0.31.0 ↗measured 2026-11-09 19:41 UTCnewer release at the time: 0.33.0
54.4 tok/s
88% of the roofline61.7 tok/s possible
- now, a token
- 18.37 ms
- floor
- 16.20 ms
- read, a token
- 15.59 GB
More
- memory, measured
- 962 GB/s
What was run
python vllm_decode_drive.py --model Qwen3.8-27B-W4A16-AWQ --contexts 512 --batches 1 --tokens 64 --reps 1 --once 64 --max-num-batched-tokens 584 --model-context-limit 584 --capture-pass pairedSpeed at context
128 tokens
512 tokens
Where the time goes
- Now, a token
- 18.37 ms
- Roofline floor, a token
- 16.20 ms
- Gap to the floor
- 2.17 ms
- Kernels a token
- 790
Per role ms a token. floor (bytes ÷ measured memory speed) above the floor
Per kernel (34): ms now / ms at the roofline, launches a token
No floor = the kernel reads no weights (norms, conversions, copies): all of its time is gap.
Run rtx4090-qwen3.8-27b-awq-vllm
Tested on
- GPU
- RTX 4090 · 24 GB
- System RAM
- 189 GB
- CPU threads
- 96
- Memory, measured
- 962 GB/s
- Temp, max
- 64 °C
- Power
- 92 W mean of 450 W
- SM clock
- 2562 MHz
- Driver
- 595.58.03
- Measured
- 2026-11-09 19:41 UTC · context 512 tokens · measured, in-model (nsys): split by role from the engine's own labels, 9 weight roles
Qwen3.8-27B on other GPUs
RTX 5090vLLM + MTP, 4-bit draft layer · W4A16-AWQ · 4-bit head202.1 tok/s H200vLLM · W4A16-AWQ100.8 tok/s · 34%2× Radeon R9700vLLM + MTP k=2 · Quark W4A1692.3 tok/s H100 80GBvLLM · W4A16-AWQ87.8 tok/s · 41%H200 NVLllama.cpp · UD-Q4_K_M78.6 tok/s · 27%2× RTX 3090vLLM · W4A16-AWQ68.6 tok/s · 76%RTX 4090vLLM · W4A16-AWQ54.4 tok/s · 88%RTX 3090vLLM · W4A16-AWQ46.2 tok/s · 80%L40SvLLM · W4A16-AWQ43.6 tok/s · 89%
The fastest setup measured on each GPU, one user. The percentage is that GPU's own roofline.