Tesla P4 8GB

NVIDIA · 8 GB GDDR5 · 192 GB/s · 75 W · MSRP $80

The Tesla P4 8GB is a NVIDIA accelerator with 8 GB GDDR5 and about 192 GB/s peak memory bandwidth. AI Hashrate estimates local LLM decode (batch 1) at 8K context: fit counts weights + KV cache + 1 GB runtime overhead — not weights alone. At Q4, about 16 curated models fit fully on this card; at FP16, about 3 fit. Top Q4 speeds: MiniCPM5-1B ≈ 113.1 tok/s (estimated, fits); Qwen3.5-2B ≈ 61.1 tok/s (estimated, fits); DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx ≈ 53.1 tok/s (estimated, fits); Llama-3.2-3B-Instruct ≈ 38.2 tok/s (estimated, fits); Qwen3-4B-Instruct-2507 ≈ 30.5 tok/s (estimated, fits). Relative ranking is more reliable than absolute tok/s. See methodology for the bandwidth formula and measured-anchor policy. Methodology.

Table: decode speed estimates (Q4 / FP16). Measured rows override estimates. MSRP is list price, not live retail. Context default 8K.

Fit checks

• Qwen3-8B — fits · ≈14.9 tok/s Q4 est.
• gemma-4-26B-A4B-it — does not fit · ≈23.9 tok/s Q4 est.
• Qwen2.5-7B-Instruct — fits · ≈16.1 tok/s Q4 est.
• Qwen3.5-9B — fits · ≈13.7 tok/s Q4 est.
• Llama-3.1-8B-Instruct — fits · ≈15.3 tok/s Q4 est.
• gpt-oss-20b — does not fit · ≈33.6 tok/s Q4 est.
• Qwen3.6-35B-A3B — does not fit · ≈17.0 tok/s Q4 est.
• Qwen3.5-35B-A3B — does not fit · ≈18.2 tok/s Q4 est.
• gemma-4-E2B-it — fits · ≈24.0 tok/s Q4 est.
• Qwen3.6-27B — does not fit · ≈2.8 tok/s Q4 est.
• Mistral-7B-Instruct-v0.3 — fits · ≈17.0 tok/s Q4 est.

All 69 model fit checks — see the linked Fits? column in the table above