GPU Specifications / AMD
AMD Instinct MI355X
CDNA 4 datacenter GPU with 288 GB of HBM3e memory, 8000 GB/s of bandwidth and up to 5,000 TFLOPS of FP16 tensor compute.
Key Specifications
Memory
288 GB HBM3e
Memory Bandwidth
8,000 GB/s
TDP
1400 W
Architecture
CDNA 4
Interconnect
Infinity Fabric 4 · 1075 GB/s
Est. On-demand Price
~$13.00/h
Hourly rates are indicative on-demand estimates; actual pricing varies by provider and commitment.
Compute Performance
| Precision | Peak throughput |
|---|---|
| FP64 (double precision) | 78.6 TFLOPS |
| FP32 (single precision) | 157.3 TFLOPS |
| FP24 | 314.6 TFLOPS |
| FP16 (tensor) | 5,000 TFLOPS |
| INT8 (tensor) | 10,100 TFLOPS |
| INT4 (tensor) | 20,100 TFLOPS |
Tensor figures use the vendor's peak numbers (with structured sparsity where supported).
System Requirements
Recommended CPU
AMD EPYC 9755 or Intel Xeon 6
Max VRAM per node (8 GPUs)
2,304 GB
System RAM (min / recommended)
1024 / 2048 GB
Minimum PSU
2600 W
LLMs on the Instinct MI355X
Number of Instinct MI355X GPUs needed to serve popular models at 8-bit quantization (including 20% overhead for activations and KV cache).
| Model | Params | VRAM (8-bit) | GPUs needed |
|---|---|---|---|
| GPT-5.6 Sol | 2400B | 2682 GB | 10x Instinct MI355X |
| GPT-5 Flagship | 2100B | 2347 GB | 9x Instinct MI355X |
| GPT-5.6 Luna | 1400B | 1565 GB | 6x Instinct MI355X |
| Kimi K3 (1.2T) | 1200B | 1341 GB | 5x Instinct MI355X |
| Kimi K2.6 (1T) | 1000B | 1118 GB | 4x Instinct MI355X |
| GPT-5.6 Terra | 800B | 894 GB | 4x Instinct MI355X |
| DeepSeek V4 Pro (671B) | 671B | 750 GB | 3x Instinct MI355X |
| Llama 4 Behemoth (500B) | 500B | 559 GB | 2x Instinct MI355X |
| Claude 5 Fable (480B) | 480B | 536 GB | 2x Instinct MI355X |
| GLM 5.2 (400B) | 400B | 447 GB | 2x Instinct MI355X |
| Grok 4.5 | 350B | 391 GB | 2x Instinct MI355X |
| Claude 4.8 Opus (300B) | 300B | 335 GB | 2x Instinct MI355X |
| Muse Spark 1.1 | 300B | 335 GB | 2x Instinct MI355X |
| Grok 4 | 270B | 302 GB | 2x Instinct MI355X |
| Gemini 3.1 Pro | 250B | 279 GB | 1x Instinct MI355X |
| Qwen 3.7 Max (235B) | 235B | 263 GB | 1x Instinct MI355X |
| Mistral Large 3 (200B) | 200B | 224 GB | 1x Instinct MI355X |
| Grok 3 Mini | 190B | 212 GB | 1x Instinct MI355X |
| Claude 5 Sonnet (175B) | 175B | 196 GB | 1x Instinct MI355X |
| Gemini 3.5 Flash | 150B | 168 GB | 1x Instinct MI355X |
| Gemini 2.5 Flash | 140B | 156 GB | 1x Instinct MI355X |
| Llama 4 Maverick (128B) | 128B | 143 GB | 1x Instinct MI355X |
| DeepSeek V4 Flash (120B) | 120B | 134 GB | 1x Instinct MI355X |
| Qwen 3.6 Plus (110B) | 110B | 123 GB | 1x Instinct MI355X |
| Nova Premier (80B) | 80B | 89 GB | 1x Instinct MI355X |
| Qwen 3 Coder-Next (80B) | 80B | 89 GB | 1x Instinct MI355X |
| Claude 4.5 Haiku (70B) | 70B | 78 GB | 1x Instinct MI355X |
| Llama 3.3 Instruct (70B) | 70B | 78 GB | 1x Instinct MI355X |
| Mistral Medium 3.5 (70B) | 70B | 78 GB | 1x Instinct MI355X |
| Yi 1.5 (40B) | 40B | 45 GB | 1x Instinct MI355X |
| Nova Core (34B) | 34B | 38 GB | 1x Instinct MI355X |
| DeepSeek V3.1 (32B) | 32B | 36 GB | 1x Instinct MI355X |
| Gemma 3 (27B) | 27B | 30 GB | 1x Instinct MI355X |
| Mistral Small 4 (24B) | 24B | 27 GB | 1x Instinct MI355X |
| Yi 1.5 (15B) | 15B | 17 GB | 1x Instinct MI355X |
| Phi 4 (14B) | 14B | 16 GB | 1x Instinct MI355X |
| Nova Lite (12B) | 12B | 13 GB | 1x Instinct MI355X |
| Llama 3.2 Instruct (11B) | 11B | 12 GB | 1x Instinct MI355X |
| Gemma 3 (9B) | 9B | 10 GB | 1x Instinct MI355X |
| Yi 1.5 Lite (9B) | 9B | 10 GB | 1x Instinct MI355X |
| Phi 4 Mini (7B) | 7B | 8 GB | 1x Instinct MI355X |
| Phi 3.5 (3.8B) | 3.8B | 4 GB | 1x Instinct MI355X |
Frequently Asked Questions
How much VRAM does the AMD Instinct MI355X have?
The AMD Instinct MI355X has 288 GB of HBM3e memory with 8000 GB/s of memory bandwidth.
Which LLMs can run on a single Instinct MI355X?
At 8-bit quantization, a single Instinct MI355X (288 GB) can serve models up to roughly 250B parameters, such as Gemini 3.1 Pro. Larger models require multiple GPUs or more aggressive quantization.
How much does it cost to rent a AMD Instinct MI355X?
On-demand cloud pricing for the Instinct MI355X is around $13.00/hour, i.e. about $9,490/month running 24/7. Actual prices vary by provider, region, and commitment.
What are the power and system requirements of the AMD Instinct MI355X?
The Instinct MI355X has a TDP of 1400W. A power supply of at least 2600W per GPU is recommended. Recommended host CPUs: AMD EPYC 9755 or Intel Xeon 6.
Deploy on a GPU cloud
Rent the AMD Instinct MI355X by the hour instead of buying hardware.
