Speculative Decoding: How to Double Your LLM Inference Speed

Speculative Decoding: How to Double Your LLM Inference Speed

Speculative decoding is one of the most effective techniques for speeding up LLM inference — achieving 2–3x throughput gains with zero quality loss. Here's how it works and how to enable it.

The Bottleneck: Autoregressive Decoding

Standard LLM inference generates one token at a time, sequentially. Each token requires a full forward pass through the model. On a 70B model, this means each token takes ~200ms — limiting throughput to 5 tokens/second even on an A100.

The root cause is memory bandwidth saturation. The GPU spends most of its time loading weights from VRAM into compute units, not doing computation. Memory bandwidth comparisons at LLMHardware.io show this bottleneck clearly across different GPU tiers.

How Speculative Decoding Works

Speculative decoding uses a small draft model (e.g., 1B params) to propose a sequence of K tokens cheaply. The large target model then verifies all K tokens in a single parallel forward pass. If most tokens are accepted, you get K tokens for the cost of ~1.2 forward passes of the large model.

Typical results:
  • Llama 3 70B + Llama 3 8B draft: 2.3x speedup
  • Mixtral 8x7B + Mistral 7B draft: 1.9x speedup
  • Acceptance rate: 75–85% for general text

Enabling in vLLM

python -m vllm.entrypoints.openai.api_server   --model meta-llama/Meta-Llama-3-70B-Instruct   --speculative-model meta-llama/Meta-Llama-3-8B-Instruct   --num-speculative-tokens 5   --tensor-parallel-size 2

Enabling in llama.cpp

./llama-speculative   -m models/llama-3-70b.Q4_K_M.gguf   -md models/llama-3-8b.Q4_K_M.gguf   --draft 8   -p "Write a Python function that..."

When Speculative Decoding Helps Most

Speculative decoding works best when the draft and target models agree frequently. This happens with repetitive or predictable text: code generation, structured output, translations. It helps less with creative writing or tasks requiring high diversity.

The speedup also depends heavily on your hardware's memory bandwidth. GPUs with higher bandwidth see more benefit because the bottleneck shifts from memory loading to compute during the verification step.

Comments

Post a Comment

Popular posts from this blog

AMD RX 9070 XT for Local LLMs: ROCm Performance Guide (2026)

The AMD RX 9070 XT launched in early 2026 at $599 with 16GB GDDR6 and 648 GB/s memory bandwidth. For local LLM users who want a mid-range GPU without NVIDIA prices, this changes the calculus significantly. Here is what it actually delivers. RX 9070 XT Specs vs Competition GPU VRAM Bandwidth Price LLM Score RX 9070 XT 16 GB 648 GB/s $599 ~60 tok/s RTX 4070 Ti Super 16 GB 672 GB/s $799 ~62 tok/s RTX 4080 Super 16 GB 736 GB/s $999 ~68 tok/s RTX 4090 24 GB 1,008 GB/s ~$1,100 ~90 tok/s At $599 with 16GB VRAM, the RX 9070 XT is the best value GPU for local LLMs in the mid-range tier in 2026. See the full comparison at LLMHardware.io's AMD GPU guide . ROCm Setup on Linux (2026) # Ubuntu 22.04/24.04 sudo apt install rocm-hip-libraries hip-base # or use the AMD installer: wget https://repo.radeon.com/amdgpu-install/6.3/ubuntu/jammy/amdgpu-install_6.3.60300-1_all.deb sudo dpkg -i amdgpu-install*.deb sudo amdgpu-install --usecase=rocm,lrt,openclsdk # Verify ROCm sees your GPU r...
Read more

llama.cpp flash attention: how to enable and benchmark

Image
Flash attention computes attention with far less memory traffic. The math is identical; the memory movement is not. In llama.cpp it matters most in two situations: long context windows and tight video memory (VRAM). With flash attention enabled, the key-value cache shrinks substantially. A 32k-token context window becomes realistic on a 16GB card, and the prompt-processing phase speeds up, so long documents get ingested faster. The catch is compatibility: not every model and quantization combination supports it, and on some setups the speedup is modest. If you run short chats on a strong GPU, you may barely notice. Where it shines is retrieval-augmented generation (RAG) and long-document work on mid-range cards. Enable it, compare your tokens per second, and keep it if your workload benefits. For the common case of pasting a long PDF and asking questions, it is close to free performance. See how context length changes VRAM needs .
Read more

SillyTavern with Local LLMs: Complete Setup for AI Companions

Image
SillyTavern is the most feature-rich frontend for character-based AI conversations. What began as an AI roleplay tool has evolved into a powerful platform with persona management, memory systems, extensive prompt engineering tools, and connections to every major local inference backend. Here's how to set it up with local models. Quick connect: SillyTavern connects to Ollama, KoboldCpp, llama.cpp server, and TabbyAPI (ExLlamaV2) out of the box. Any of the hardware setups at LLMHardware.io provides a solid backend for SillyTavern. Installation Install via git: git clone https://github.com/SillyTavern/SillyTavern && cd SillyTavern && npm install then node server.js . Access at localhost:8000. For persistent installation, use the included start.sh script. SillyTavern uses Node.js and runs on any platform. The companion app (STMP) enables mobile access to your SillyTavern instance via Tailscale or local network. Character Cards and the TavernAI Format Chara...
Read more