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

Gemma 2 27B on AMD MI325X 288GB: Does It Fit, How Fast, What It Costs

Gemma 2 27B at FP8 (E4M3) needs 27.0 GB for its weights and 51.7 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
51.7 GB needed / 288 GB available
Decode speed
216.7
tok/s per stream
Aggregate
2600
tok/s @ 8 concurrent
Cost / 1M tokens
$0.3739
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 GB51.7 GB216.72600711.5 ms$0.3739✓
2× AMD576 GB51.7 GB368.38840371.2 ms$0.2200✓
4× AMD1152 GB51.7 GB736.735360200.6 ms$0.1100✓
8× AMD2304 GB51.7 GB1473141440115.3 ms$0.0550✓
Concurrency vs context length
On 1× AMD MI325X 288GB at FP8 (E4M3). KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K1691.54 GB✓ fits
8K843.09 GB✓ fits
33K2112.35 GB✓ fits
66K1024.70 GB✓ fits
131K549.39 GB✓ fits
Other GPUs that run Gemma 2 27B
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB2×2640$0.1052
AMD MI300X 192GB1×2297$0.3024
L40S 48GB2×1273$0.3055
A100 80GB SXM41×866.7$0.3846
Intel Gaudi 3 128GB1×1430$0.3885
H100 SXM5 80GB1×1452$0.4784
Frequently asked questions

Can Gemma 2 27B run on AMD MI325X 288GB?

Yes. Gemma 2 27B at FP8 (E4M3) needs 51.7 GB including the KV cache, and AMD MI325X 288GB has 288 GB — so it fits on a single GPU.

How fast is Gemma 2 27B on AMD MI325X 288GB?

216.7 tokens/sec per stream and 2600 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 711.5 ms, and inter-token latency is 4.6 ms.

What does Gemma 2 27B on AMD MI325X 288GB cost per million tokens?

$0.3739 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 Gemma 2 27B need on AMD MI325X 288GB?

27.0 GB for the weights at FP8 (E4M3), plus 24.7 GB of KV cache for 8 concurrent requests at 8K context — 51.7 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is AMD MI325X 288GB the right choice for Gemma 2 27B?

It works. Compare $0.3739 per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 8× RTX 3080 10GB at $0.0248, and the fastest is 2× Cerebras CS-3 (WSE-3) at 29467 tok/s.

Adjust the workload yourself

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

Open in calculator