AMD Instinct MI355X
AMD: specifications, comparisons and alternatives.
The AMD Instinct MI355X is a CDNA 4 accelerator for AI training, inference and HPC. According to AMD, it has 288 GB of HBM3E, 8 TB/s of memory bandwidth and a board power of 1,400 W. The OAM module was launched on 12 June 2025 and supports lower compute precision such as MXFP4. All figures apply per accelerator.
AMD
- 288 GB
- Memory
- HBM3E
- Memory type
- 8 TB/s
- Memory bandwidth
- 1.400 W
- Power (board power or TDP)
You can see the current price and availability with the partner as soon as a link has been set up.
How are the points calculated?
Clouwino points are an arithmetic summary of the specifications the manufacturer states, on a scale of 0 to 100. It is not a benchmark and not a judgement on quality or suitability: a product with fewer points may suit your situation better. Products are compared with other products of the same type. For each property we count how far a product sits between the lowest and highest value in the group:
- FP32 compute (for: consumer), carries heavy weight
- Memory (for: consumer), carries heavy weight
- Memory bandwidth (for: consumer), carries heavy weight
- Compute units (for: consumer), carries light weight
- Boost clock (for: consumer), carries light weight
- Memory (for: professional and data centre), carries medium weight
- Memory bandwidth (for: professional and data centre), carries medium weight
- FP16/BF16 (for: professional and data centre), carries heavy weight
- FP8 (for: professional and data centre), carries medium weight
- FP64 (for: professional and data centre), carries medium weight
If too much data is missing, we do not show points. Points are relative: when a new product is added, they can shift. Read more at how we compare.
Specifications of AMD Instinct MI355X
All specifications the manufacturer lists, per item. Want to see another product next to it? Pick one below.
| Specificatie | AMD Instinct MI355X |
|---|---|
| Overview | |
| Clouwino points | 68 punten |
| General | |
| Architecture | CDNA 4 |
| Manufacturing process | TSMC 3nm | 6nm FinFET |
| Form | OAM |
| Launch | 2025-06 |
| Compute | |
| FP32 compute | 157,3 TFLOPS |
| FP64 | 78,6 TFLOPS |
| FP16/BF16 | 2.500 TFLOPS |
| FP8 | 5.000 TFLOPS |
| FP4 | 10.100 TFLOPS |
| Performance notes | FP16/BF16 matrix: 2.5 PFLOPS dense (5 PFLOPS sparse). OCP-FP8 matrix: 5 PFLOPS dense (10.1 PFLOPS sparse). MXFP4: 10.1 PFLOPS, in AMD's dense column; no sparsity figure stated. MXFP6 10.1 PFLOPS. FP64 78.6 TFLOPS (vector and matrix), FP32 157.3 TFLOPS. PFLOPS converted to TFLOPS (unit only). Note: the page extraction did not explicitly label MXFP4/FP16 vector; check the dense/sparse columns manually if necessary. |
| Memory | |
| Memory | 288 GB |
| Memory type | HBM3E |
| Memory bandwidth | 8 TB/s |
| Power and connections | |
| Power (board power or TDP) | 1.400 W |
| GPU interconnect | 7x Infinity Fabric, 153 GB/s per link; PCIe 5.0 x16; scale-out 128 GB/s (1 link) |
| Multiple instances | Not stated by the manufacturer |
Note: Figures per GPU; TBP 1400 W (passive & active cooling). Launch date 12 June 2025. Sister product MI350X (also launched 12 June 2025): 288 GB HBM3E, 8 TB/s, TBP 1000 W, FP64 72.1 TFLOPS, FP16 matrix 2.3 PFLOPS dense (4.6 sparse), FP8 4.6 PFLOPS dense (9.2 sparse), MXFP4 9.2 PFLOPS. Partitioning not described (SR-IOV only). TF32 via software emulation.
Alternatives to AMD Instinct MI355X
Products with similar specifications from the same overview. This is not a ranking and not buying advice.
- AMD Instinct MI325X256 GB · HBM3E · 6 TB/sCompare with AMD Instinct MI355X →
- AMD Instinct MI300X192 GB · HBM3 · 5,3 TB/sCompare with AMD Instinct MI355X →
- NVIDIA Rubin (per GPU)288 GB · HBM4 · 22 TB/sCompare with AMD Instinct MI355X →
- NVIDIA H200 SXM141 GB · HBM3e · 4,8 TB/sCompare with AMD Instinct MI355X →
- NVIDIA H100 SXM80 GB · HBM3 · 3,35 TB/sCompare with AMD Instinct MI355X →
- AMD Instinct MI250X128 GB · HBM2e · 3,2 TB/sCompare with AMD Instinct MI355X →
What the MI355X is
The MI355X forms the CDNA 4 generation of AMD Instinct and is made with TSMC 3nm and 6nm FinFET technology. It is an OAM module, suitable for both passive and active cooling. The sister product MI350X was launched on the same day and has a lower board power of 1,000 W.
The connections are 7 Infinity Fabric links at 153 GB/s per link, a PCIe 5.0 x16 connection and a scale-out link of 128 GB/s. Partitioning is described only as SR-IOV. You can find more background in GPU interconnects explained.
The MI355X and MI350X therefore share many characteristics, but differ in power and compute figures.
The newer CDNA 4 architecture adds lower precisions such as MXFP4, which are intended for AI inference and training where less precise numbers are sufficient.
Memory, power and precision of the MI355X
With 288 GB of HBM3E and 8 TB/s, a large model can fit on a single accelerator. The board power of 1,400 W is one of the higher values in this group, which matters a great deal for cooling. Also read what TDP means.
Per accelerator, AMD states for FP16/BF16 matrix 2.5 PFLOPS dense (5 PFLOPS with sparsity) and for OCP-FP8 5 PFLOPS dense (10.1 PFLOPS with sparsity). The MXFP4 value of 10.1 PFLOPS is in the dense column, without a sparsity figure. FP64 is 78.6 TFLOPS and FP32 157.3 TFLOPS. The labelling of some values is not fully explicit in the source; check this with AMD.
According to the source, the TF32 performance is achieved through software emulation, and we therefore do not state a separate figure.
Choosing the MI355X
A 1,400 W module is a serious demand on power and cooling. Check with your supplier whether the system uses air or liquid cooling and how many modules it contains.
- Only compare MXFP4 with other FP4 formats if the label is the same.
- Read compute precision explained for the differences between formats.
- Consider whether availability via cloud or server suits your plan.
A board with this much power requires coordination with the data centre’s operator.
You can see the current price and availability at the partner via the button on the page.
Frequently asked questions
How much power does the MI355X use?
AMD states a total board power of 1,400 W for the MI355X, suitable for passive and active cooling. That is a peak value for the board, so cooling and power supply are important points to consider.
What is MXFP4 on the MI355X?
MXFP4 is a low compute precision for AI tasks. AMD states 10.1 PFLOPS, listed in the dense column, without a separate sparsity figure. Check the exact labels with AMD itself.
What is the difference from the MI350X?
The MI350X is a sister product that was also launched on 12 June 2025. According to the data, it likewise has 288 GB of HBM3E and 8 TB/s, but a board power of 1,000 W and slightly lower compute figures.
Last updated on
