NVIDIA RTX 5060 Ti (8 GiB)

NVIDIA · Dedicated memory

Installed memory
8 GiB
Usable per device
8 GiB
Bandwidth
448 GB/s
Largest fitting model
8.83B
Power
180 W

Inference configuration

One sequence per batch, a 512-token prompt and 128 generated tokens. Multi-GPU estimates use pipeline parallelism over PCIe 4.

Open in calculator

Decode speed by model

Compare up to four checkpoints at the settings above. Bars show generated tokens per second.

Llama 3.1 8B Instruct58.4 tok/s
DeepHermes-3-Llama-3-3B-Preview143.3 tok/s
DeepHermes-3-Llama-3-8B-Preview58.4 tok/s

Model performance matrix

Reference rates use the published 32K-cache benchmark when the calculator matches it. Other speeds use its roofline estimator; TTFT covers prompt processing.

#QuantizationRate source
1SmolLM2-135M0.13B3,047.10.01 sFitsQ4Roofline
2SmolLM2-135M-Instruct0.13B3,047.10.01 sFitsQ4Roofline
3SmolLM2-360M0.36B1,188.80.025 sFitsQ4Roofline
4SmolLM2-360M-Instruct0.36B1,188.80.025 sFitsQ4Roofline
5Qwen2-0.5B0.49B941.40.032 sFitsQ4Roofline
6Qwen2-0.5B-Instruct0.49B941.40.032 sFitsQ4Roofline
7Qwen2.5-0.5B0.49B941.40.032 sFitsQ4Roofline
8Qwen2.5-0.5B-Instruct0.49B941.40.032 sFitsQ4Roofline
9Qwen2.5-Coder-0.5B0.49B941.40.032 sFitsQ4Roofline
10Qwen2.5-Coder-0.5B-Instruct0.49B941.40.032 sFitsQ4Roofline
1 / 33

Inference capacity

Llama 3.1 8B Instruct · Fits · 6.52 / 8 GiB

Selection notes

Largest catalog checkpoint inside this memory budget · Yi-1.5-9B Use the remaining memory as headroom before increasing concurrency.

Fine-tuning capacity

Llama 3.1 8B Instruct · 4,096 tokens · batch 1 · calculator defaults

QLoRA

8.4 GiB

Over memory

LoRA

20.8 GiB

Over memory

Full fine-tuning

154.1 GiB

Over memory

Frequently asked questions

How is usable memory determined?

The calculator uses the device profile’s available fraction · 100% × 8 GiB. System use can reduce the amount available at runtime.

Why is a speed missing?

Speed needs a fitting checkpoint and either matching calibration or both bandwidth and a comparable dense BF16 peak.