AMD192 GB VRAM5,300 GB/s$2.5/hr

Mixtral 8x7B (MoE) on AMD MI300X 192GB: Does It Fit, How Fast, What It Costs

Mixtral 8x7B (MoE) at FP8 (E4M3) needs 46.7 GB for its weights and 55.3 GB in total at 8K context with 8 concurrent users. That means it fits on a single AMD MI300X 192GB. Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 1 GPU
55.3 GB needed / 192 GB available
Decode speed
400.6
tok/s per stream
Aggregate
4807
tok/s @ 8 concurrent
Cost / 1M tokens
$0.1445
1× AMD MI300X 192GB
What changes as you add AMD MI300X 192GBs
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× AMD192 GB55.3 GB400.64807355.9 ms$0.1445✓
2× AMD384 GB55.3 GB681.016344193.2 ms$0.0850✓
4× AMD768 GB55.3 GB136265375111.6 ms$0.0425✓
8× AMD1536 GB55.3 GB272426150070.8 ms$0.0212✓
Concurrency vs context length
On 1× AMD MI300X 192GB 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?
4K270549.8 MB✓ fits
8K1351.07 GB✓ fits
33K334.29 GB✓ fits
66K168.59 GB✓ fits
131K817.18 GB✓ fits
Other GPUs that run Mixtral 8x7B (MoE)
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB2×5526$0.0503
L40S 48GB2×2664$0.1460
AMD MI325X 288GB1×5442$0.1787
A100 80GB SXM41×1814$0.1838
Intel Gaudi 3 128GB1×2993$0.1856
H100 SXM5 80GB1×3038$0.2286
Frequently asked questions

Can Mixtral 8x7B (MoE) run on AMD MI300X 192GB?

Yes. Mixtral 8x7B (MoE) at FP8 (E4M3) needs 55.3 GB including the KV cache, and AMD MI300X 192GB has 192 GB — so it fits on a single GPU.

How fast is Mixtral 8x7B (MoE) on AMD MI300X 192GB?

400.6 tokens/sec per stream and 4807 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 355.9 ms, and inter-token latency is 2.5 ms.

What does Mixtral 8x7B (MoE) on AMD MI300X 192GB cost per million tokens?

$0.1445 per 1M output tokens at 1× AMD MI300X 192GB ($2.50/hr). That is the marginal hardware cost — add idle capacity, ops, and egress for the real figure.

How much VRAM does Mixtral 8x7B (MoE) need on AMD MI300X 192GB?

46.7 GB for the weights at FP8 (E4M3), plus 8.6 GB of KV cache for 8 concurrent requests at 8K context — 55.3 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is AMD MI300X 192GB the right choice for Mixtral 8x7B (MoE)?

It works. Compare $0.1445 per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 8× RTX 3080 10GB at $0.0119, and the fastest is 2× Cerebras CS-3 (WSE-3) at 61674 tok/s.

Adjust the workload yourself

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

Open in calculator