ai.canshu.com
中文
VRAM Calculator Advisor & TCO API Pricing Compare Models Leaderboard GPUs

Search models, GPUs, and tools

↑↓ Navigate · Enter Open · ★ Watchlist prioritized ai.canshu.com
Q4_K_M: Smooth Fit 12.82 GB / 31.4 GB Usable

DeepSeek-R1-Distill-Qwen-14B on NVIDIA GeForce RTX 5090 32GB: VRAM Requirement & Speed Estimate

Yes. At Q4_K_M 4-bit quantization with a 16K context window, DeepSeek-R1-Distill-Qwen-14B requires 12.82 GB of VRAM (8.97 GB weights + 3.22 GB KV cache + 0.63 GB overhead), fitting within RTX 5090's 31.4 GB usable memory at ~121.9 tokens/s.

Concurrency (Batch):
Dense 14.8B·48L (d=5120)·40Q/8KV·Ctx 128K
Runnable Models →
GPUs
32 GB VRAM·1792 GB/s·Safe Usable 31.4 GB
Smooth Fit12.82 / 31.4 GB~121.9 tok/s
3. Weight Quantization
4. Per-Request Context16K (16,384 tok)
Verified: 2026-11-20·Roofline bandwidth/compute estimate (Batch=1)
Smooth Fit
12.82/ 31.4 GB usable

Physical 32 GB · Load 40.9%

Theoretical Decode Speed
~121.9tokens/s

Actual speed varies ~±25% by runtime

Max Context at Batch=1
108K/ 128K

Enable Q8/Q4 KV quantization to scale batch concurrency

Weights: 8.97 GBKV Cache (1×16K): 3.22 GBOverhead: 0.63 GB
🚀 Blazing Fast (Instant CoT)~4.1s for a 500-token response
Reference Launch CLI:
OLLAMA_CONTEXT_LENGTH=16384 ollama run deepseek-r1:14b

RTX 5090 Fits DeepSeek-R1-Distill-Qwen-14B · Cost & TCO

Approx. $2,199 (~$69/GB VRAM) · 575W TDP

📊 All Quantization Levels for DeepSeek-R1-Distill-Qwen-14B on RTX 5090

Quantization Total VRAM (16K) Verdict Est. Speed Max Context
FP16 / BF16 33.86 GB Partial RAM Offload (Slow) ~17.8 tok/s 3K
FP8 / Q8_0 19.71 GB Smooth Fit ~74.4 tok/s 73K
GGUF Q6_K 16.05 GB Smooth Fit ~93.9 tok/s 92K
GGUF Q5_K_M 14.41 GB Smooth Fit ~106.3 tok/s 100K
Q4_K_M / AWQ 12.82 GB Smooth Fit ~121.9 tok/s 108K
GGUF IQ4_XS 11.69 GB Smooth Fit ~136.2 tok/s 113K
GGUF Q3_K_M 11.05 GB Smooth Fit ~145.9 tok/s 116K
GGUF Q2_K 9.62 GB Smooth Fit ~173.5 tok/s 124K

💡 Recommended Alternative GPUs & Macs for DeepSeek-R1-Distill-Qwen-14B

Frequently Asked Questions

Can NVIDIA GeForce RTX 5090 32GB run DeepSeek-R1-Distill-Qwen-14B locally?

Yes. At Q4_K_M 4-bit quantization with a 16K context window, DeepSeek-R1-Distill-Qwen-14B requires 12.82 GB of VRAM (8.97 GB weights + 3.22 GB KV cache + 0.63 GB overhead), fitting within RTX 5090's 31.4 GB usable memory at ~121.9 tokens/s.

What is the best quantization for DeepSeek-R1-Distill-Qwen-14B on RTX 5090?

The highest-quality quantization that fits cleanly on RTX 5090 is FP8 / Q8_0 (99.9% Near-Lossless), using 19.71 GB VRAM at ~74.4 tokens/s with up to 73K context.

How much VRAM does DeepSeek-R1-Distill-Qwen-14B need in full FP16 precision?

In unquantized FP16/BF16 precision (16K context), DeepSeek-R1-Distill-Qwen-14B requires 33.86 GB VRAM (29.6 GB for weights alone).