GPUs

AMD Instinct MI250X vs Intel Gaudi 3 Full Specs

14,080 Shaders
1.7GHz
16,384 Shaders
438MHz
128GB HBM2e3.28TB/s
128GB HBM2e3.7TB/s
··
47.87 TFLOPS
··
-
Form Factor
OAM Module
Form Factor
OAM Module
TDP
560W
TDP
900W

Instinct MI250XInstinct MI250X383 TOPSINT4 Tensor
x1
Gaudi 3Gaudi 31.84 PFLOPSFP8-32 Tensor
x4.80

Clock Speed
···
Clock Speed
··
Peak OPS
383 TOPSINT4 Tensor
Peak OPS
1.84 PFLOPSFP8-32 Tensor
-
Tensor FP8-32
1.84 PFLOPS
Tensor FP16-16
383 TFLOPS
Tensor FP16-32
383 TFLOPS
Tensor FP16-16
-
Tensor FP16-32
459.3 TFLOPS
Tensor BF16
383 TFLOPS
Tensor BF16
1.84 PFLOPS
Tensor TF32
-
Tensor TF32
459.3 TFLOPS
FP32
47.87 TFLOPS
Tensor FP32
95.74 TFLOPS
FP32
-
Tensor FP32
229.6 TFLOPS
FP64
47.87 TFLOPS
Tensor FP64
95.74 TFLOPS
-
Tensor INT4
383 TOPS
Tensor INT4
-
Tensor INT8
383 TOPS
-
Texture Rate
1.5 TTexel/s
Texture Rate
-

Shaders
14,080 Shaders
Shaders
16,384 Shaders
TMUs
880 TMUs
TMUs
-
Tensor Cores
880 T-Cores
Tensor Cores
64 T-Cores
CUs
220 CUs
CUs
8 CUs

Base Clock
1GHz
Base Clock
-
Boost Clock
1.7GHz
Boost Clock
438MHz

L2 Cache
16.4MB shared
L2 Cache
-

128GB HBM2e
128GB HBM2e
Memory Bus
8192-bit
Memory Bus
8192-bit
Memory Speed
3.2GT/s
Memory Speed
3.6GT/s
Memory Bandwidth
3.28TB/s
Memory Bandwidth
3.7TB/s
ECC
Yes
ECC
Yes
ESRAM
-
ESRAM
96MBSize
12800GB/sBandwidth

TDP
560W
TDP
900W

Encoder Model
2x VCN 2.6
-

Decoder Model
2x VCN 2.6
Decoder Model
14x Gaudi

Form Factor
OAM Module
Form Factor
OAM Module
Cooling
Passive
Cooling
Passive

Manufacturer
Manufacturer
Chip Designer
Chip Designer
Architecture
Architecture
Family
Family
Branding
Instinct branding
Branding
Gaudi branding
Codename
Aldebaran
Codename
Gaudi 3
Chip Variant
Aldebaran XTX
Chip Variant
-
Market Segment
Server
Market Segment
Server
Release Date
Nov 8, 2021
Release Date
Apr 9, 2024

Foundry
TSMC
Foundry
TSMC
Fabrication Node
N6
Fabrication Node
N5
Die Size
1448mm²
Die Size
-
Transistor Count
56 Billion
Transistor Count
-
Transistor Density
38.67 MTr/mm²
Transistor Density
-

No images available
No images available