RTX 3060 12GB最佳性价比12 GB VRAM (+2)
C
Makes the model fit on the accelerator instead of staying completely out of reach.16.2 tok/s 解码
Makes the model fit on the accelerator instead of staying completely out of reach.
~$329 MSRP
Can Phi 3 Medium 14B run on RTX 3080 10GB?
YES — With Q2_K
C54Usable
Estimated from fit model
Phi 3 Medium 14B needs ~10.7 GB VRAM. RTX 3080 10GB has 10.0 GB. With Q2_K quantization, expect ~63 tok/s.
Runtime: OllamaCapacity: OffloadBandwidth: MediumStack: BasicBottleneck: Host offload
Operating mode
Choose the run profile you care about
Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
Phi 3 Medium 14B at Q4_K_M needs 13.8 GB — too much for RTX 3080 10GB (10.0 GB). Runs at Q2_K (10.7 GB) with low quality.
Select quantization to explore
Q4_K_M (Medium quality) — 13.8 GB, exceeds 10.0 GB available
13.8 GB required10.0 GB available
138% VRAM needed
3.8 GB over capacity — needs offload or smaller quantization
Fit status
Too heavy
Decode
27.7 tok/s
TTFT
6985 ms
Safe context
4K
Memory
13.8 GB / 10.0 GB
Offload
30%
Memory breakdown
Weights8.5 GB
KV Cache3.1 GB
Runtime1.2 GB
Headroom1.0 GB
See how fast it feels
With memory offload — actual speed may be lowerSee how fast it feelsPhi 3 Medium 14B on RTX 3080 10GB
1st promptCold start — includes initialization
>Solve: If a train travels at 60 mph for 2.5 hours, how far does it go?
Let me work through this step by step.
Given information:
- Speed = 60 mph (miles per hour)
- Time = 2.5 hours
Formula: Distance = Speed x Time
Calculation:
Distance = 60 mph x 2.5 h
= 150 miles
The train travels 150 miles.
Double-check: at 60 mph the train covers 60 miles each hour. In half an hour it covers 30 miles. So in 2.5 hours: 60 + 60 + 30 = 150 miles. Confirmed.
2nd promptWarm — model ready, no init overhead
>What if the train stops for 20 minutes halfway?
The stop doesn't change the total distance covered while moving, but it affects total trip time.
Breakdown:
- First half: 1.25 hours at 60 mph = 75 miles
- Stop: 20 minutes (no distance)
- Second half: 1.25 hours at 60 mph = 75 miles
Total distance: 150 miles (same)
Total time: 2.5 h + 0.33 h = 2.83 hours
Average speed: 150 / 2.83 = 53 mph
The distance stays the same but average speed drops to 53 mph because of the stop.
Estimated: 27.7 tok/s decode · 7.0s TTFT (warm) · 69 tok/s prefill
What limits this setup
It fits through host-memory offload, and offload is the main reason performance drops.
CPU or host-memory offload is active
About 10% of the working set spills out of accelerator memory, which usually hurts latency and sustained decode throughput.
Very little memory headroom
You can run the model, but there is not much room left for longer context, bigger batches, extra apps, or future model updates.
Best improvement path
Remove offload with more accelerator memory
Prioritize a GPU or unified-memory tier that fits the whole model natively. Removing offload usually helps more than small compute gains.
Buy headroom, not only minimum fit
A slightly larger memory tier gives you safer context growth and makes the recommendation more future-proof.
Increase host RAM if you keep offloading
This setup may need roughly 0.4 GB of extra host RAM just for the offloaded portion, before OS and other tools.
Performance by workload
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Chat | F | Too heavy | 35.5 tok/s | 2977 ms | 4K |
| Coding | F | Too heavy | 27.7 tok/s | 6985 ms | 4K |
| Agentic Coding | F | Too heavy | 18.2 tok/s | 15475 ms | 4K |
| Reasoning | F | Too heavy | 27.7 tok/s | 8255 ms | 4K |
| RAG | F | Too heavy | 18.2 tok/s | 19344 ms | 4K |
Quantization options
How Phi 3 Medium 14B (14B params) fits at each quantization level on RTX 3080 10GB (10.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 5.5 GB | Low | B64 |
Q3_K_SBest for your GPU | 3 | 6.9 GB | Low | B63 |
NVFP4 | 4 | 7.8 GB | Medium | F0 |
Q4_K_M | 4 | 8.5 GB | Medium | F0 |
Q5_K_M | 5 | 10.1 GB | High | F0 |
Q6_K | 6 | 11.5 GB | High | F0 |
Q8_0 | 8 | 15.0 GB | Very High | F0 |
F16 | 16 | 28.7 GB | Maximum | F0 |
Get started
Copy-paste commands to run Phi 3 Medium 14B on your machine.
Run
ollama run phi3:medium升级选项
能流畅运行 Phi 3 Medium 14B 的硬件
RTX 3060 12GB最佳性价比RTX 5060 Ti 16GB经济之选16 GB VRAM (+6)
B
Makes the model fit on the accelerator instead of staying completely out of reach.35 tok/s 解码
Makes the model fit on the accelerator instead of staying completely out of reach.
Removes host-memory offload, which is usually the single biggest latency and throughput win.
~$449 MSRP
RTX 4060 Ti 16GBNVIDIA 升级16 GB VRAM (+6)
B
Makes the model fit on the accelerator instead of staying completely out of reach.26.5 tok/s 解码
Makes the model fit on the accelerator instead of staying completely out of reach.
Removes host-memory offload, which is usually the single biggest latency and throughput win.
~$499 MSRP
Frequently asked questions
Can RTX 3080 10GB run Phi 3 Medium 14B?
Yes, RTX 3080 10GB can run Phi 3 Medium 14B at Q2_K quantization (Runs with offload (needs ~0.4 GB host RAM)). The recommended Q4_K_M requires 13.8 GB which exceeds available memory, but at Q2_K it needs only 10.7 GB. Expected decode speed: 62.7 tok/s.
How much VRAM does Phi 3 Medium 14B need?
Phi 3 Medium 14B (14B parameters) requires approximately 13.8 GB at Q4_K_M quantization. On RTX 3080 10GB, it fits at Q2_K using 10.7 GB.
What is the best quantization for Phi 3 Medium 14B?
The recommended quantization is Q4_K_M, but on RTX 3080 10GB the best fitting quantization is Q2_K, which uses 10.7 GB.
What speed will Phi 3 Medium 14B run at on RTX 3080 10GB?
On RTX 3080 10GB, Phi 3 Medium 14B achieves approximately 62.7 tokens per second decode speed with a time-to-first-token of 3086ms using Q2_K quantization.
Can RTX 3080 10GB run Phi 3 Medium 14B for coding?
For coding workloads, Phi 3 Medium 14B on RTX 3080 10GB receives a F grade with 27.7 tok/s and 4K context.
What context window can Phi 3 Medium 14B use on RTX 3080 10GB?
On RTX 3080 10GB, Phi 3 Medium 14B can safely use up to 12K tokens of context at Q2_K quantization. The model's official context limit is 128K, but available memory constrains the safe maximum.
What should I upgrade first if Phi 3 Medium 14B feels slow on RTX 3080 10GB?
Remove offload with more accelerator memory. Prioritize a GPU or unified-memory tier that fits the whole model natively. Removing offload usually helps more than small compute gains.
Embed this result▼
Paste this snippet into any page to show a live fit card.
<iframe src="https://willitrunai.com/embed/phi-3-medium-14b-on-rtx-3080-10gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>
Preview: