AMD288 GB VRAM6,000 GB/s$3.5/hr

Phi-4 14B on AMD MI325X 288GB: Does It Fit, How Fast, What It Costs

Phi-4 14B at FP16 needs 28.0 GB for its weights and 38.7 GB in total at 8K context with 8 concurrent users. That means it fits on a single AMD MI325X 288GB. Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 1 GPU
38.7 GB needed / 288 GB available
Decode speed
139.3
tok/s per stream
Aggregate
1671
tok/s @ 8 concurrent
Cost / 1M tokens
$0.5817
1× AMD MI325X 288GB
What changes as you add AMD MI325X 288GBs
Tensor-parallel degree vs VRAM, throughput, and cost. Cost scales linearly with GPU count; decode throughput scales near-linearly too, so cost per token stays roughly flat.
GPUsTotal VRAMVRAM neededDecode tok/sAggregate tok/sTTFTCost / 1M tokFits?
1× AMD288 GB38.7 GB139.31671388.2 ms$0.5817✓
2× AMD576 GB38.7 GB236.85683209.7 ms$0.3422✓
4× AMD1152 GB38.7 GB473.622731119.9 ms$0.1711✓
8× AMD2304 GB38.7 GB947.19092674.9 ms$0.0855✓
Concurrency vs context length
On 1× AMD MI325X 288GB at FP16. KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K387687.2 MB✓ fits
8K1931.34 GB✓ fits
33K485.37 GB✓ fits
66K2410.74 GB✓ fits
131K1221.47 GB✓ fits
Other GPUs that run Phi-4 14B
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB2×1697$0.1637
AMD MI300X 192GB1×1476$0.4704
A100 40GB SXM41×431.8$0.5790
A100 80GB SXM41×557.1$0.5983
Intel Gaudi 3 128GB1×919.3$0.6043
H100 SXM5 80GB1×933.2$0.7441
Frequently asked questions

Can Phi-4 14B run on AMD MI325X 288GB?

Yes. Phi-4 14B at FP16 needs 38.7 GB including the KV cache, and AMD MI325X 288GB has 288 GB — so it fits on a single GPU.

How fast is Phi-4 14B on AMD MI325X 288GB?

139.3 tokens/sec per stream and 1671 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 388.2 ms, and inter-token latency is 7.2 ms.

What does Phi-4 14B on AMD MI325X 288GB cost per million tokens?

$0.5817 per 1M output tokens at 1× AMD MI325X 288GB ($3.50/hr). That is the marginal hardware cost — add idle capacity, ops, and egress for the real figure.

How much VRAM does Phi-4 14B need on AMD MI325X 288GB?

28.0 GB for the weights at FP16, plus 10.7 GB of KV cache for 8 concurrent requests at 8K context — 38.7 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is AMD MI325X 288GB the right choice for Phi-4 14B?

It works. Compare $0.5817 per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 4× RTX 3080 10GB at $0.0772, and the fastest is 1× Cerebras CS-3 (WSE-3) at 5571 tok/s.

Adjust the workload yourself

Same formulas, your context length, concurrency, and quantisation.

Open in calculator