AMD Instinct MI250X Specs

AMD Instinct MI250X
14080 Shaders
128GB HBM2e
1700MHz
Peak AI Performance
382.98 TOPS
INT4 Tensor
FP32
47.87 TFLOPS
FP16
95.74 TFLOPS
Form Factor
OAM Module
-
TDP
560W
-
-
-
-
-
GB6 OpenCL N/A
0%
GB6 Metal N/A
0%
GB6 Vulkan N/A
0%
GB5 OpenCL N/A
0%
GB5 CUDA N/A
0%
GB5 Metal N/A
0%
GB5 Vulkan N/A
0%
OCT 2020.1 N/A
0%
OCT Metal N/A
0%
Peak AI Performance
382.98 TOPS
INT4 Tensor
-
-
-
-
-
-
-
-
-
FP16
95.74 TFLOPS
382.98 TFLOPS Tensor (FP16 Accumulate)
-
382.98 TFLOPS Tensor (FP32 Accumulate)
-
FP32
47.87 TFLOPS
95.74 TFLOPS Tensor
-
FP64
47.87 TFLOPS
95.74 TFLOPS Tensor
BF16
-
382.98 TFLOPS Tensor
-
-
-
-
INT4
382.98 TOPS Tensor
-
INT8
-
382.98 TOPS Tensor
-
-
-
-
-
Pixel Fillrate
-
-
-
Texture Fillrate
1496 GTexel/s
Manufacturer
AMD
Chip Designer
AMD
Architecture
CDNA 2
Family
Instinct
Codename
Aldebaran
Aldebaran XTX
Variant
Aldebaran XTX
Market Segment
Server
Release Date
11/8/2021
Foundry
TSMC
-
Fabrication Node
N6
-
Die Size
2x 724 mm²
-
Transistor Count
2x 28 Billion
-
Transistor Density
38.67M/mm²
-
Form
OAM Module
Shading Units
14080 Shaders
-
Texture Mapping Units
880 TMUs
Render Output Units
-
Tensor Cores
880 T-Cores
-
-
-
-
Compute Units
220 CUs
-
-
-
-
-
-
1000MHz Base
1700MHz
-
-
L1
-
-
16KB/CU
-
L2
16MB Shared
-
-
-
128GB
HBM2e
ECC
Bus Width
8192Bit
Clock
1600MHz
Transfer Rate
3.2GT/s
Bandwidth
3276.8GB/s
-
-
-
-
-
-
-
-
-
TDP
560W
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
No Ports
Max Resolution
Unknown
Max Resolution Refresh Rate
-
Variable Refresh Rate
-
-
-
Display Stream Compression (DSC)
Not Supported
Multi Monitor Support
Unknown
-
-
Model
2x VCN 2.6
Codec
-
-
-
-
-
-
-
-
AVC (H.264)
HEVC (H.265)
-
-
-
-
No Decoders
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
OpenCL
3.0
-
-
-
-
-
-
GFX
9.4
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
PCIe Version
4.0
PCIe Lanes
16
Multi GPU Support
Supported
Type
Infinity Fabric
-
-
-
-
-
-
AMD Instinct MI250X AMD Instinct MI250
AMD Instinct MI250X vs AMD Instinct MI250
Change Comparison