GPU Comparison
RADEON
AMD Instinct MI300X
CORE STATE
Aqua Vanjaram
VRAM
192 GB
CLOCK SPEED
2100 MHz
TDP
750 W
BUS WIDTH
8192 bit
ARCHITECTURE
CDNA 3.0
PROCESS
5 nm
LAUNCH DATE
2023
VS
GEFORCE
A10M
CORE STATE
GA102
VRAM
20 GB
CLOCK SPEED
1635 MHz
TDP
150 W
BUS WIDTH
320 bit
ARCHITECTURE
Ampere
PROCESS
8 nm
LAUNCH DATE
—
PERFORMANCE BENCHMARKS
geekbench_opencl
DETAILED SPECIFICATIONS
SPECIFICATION
Instinct MI300X
A10M
Core Specs
Shading Units
19,456
7,168
-63.2%
Shaders
19,456
7,168
-63.2%
TMUs
1,216
224
-81.6%
ROPs
0
80
+∞%
Compute Units
304
—
SM Count
—
56
Clocks
Base Clock
1000 MHz
975 MHz
Boost Clock
2100 MHz
1635 MHz
Memory Clock
1300 MHz 5.2 Gbps effective
1563 MHz
12.5 Gbps effective
Memory
Memory Size
192 GB
20 GB
VRAM (MB)
196,608
20,480
-89.6%
Memory Type
HBM3
GDDR6
Memory Bus
8192 bit
320 bit
Bandwidth
5.32 TB/s
500.2 GB/s
Cache
L1 Cache
16 KB (per CU)
128 KB (per SM)
L2 Cache
16 MB
6 MB
L3 Cache
256 MB
—
Performance
Pixel Rate
0 MPixel/s
130.8 GPixel/s
Texture Rate
2,553.6 GTexel/s
366.2 GTexel/s
FP32 (TFLOPS)
81.72 TFLOPS
23.44 TFLOPS
FP64 (TFLOPS)
40.86 TFLOPS (1:2)
732.5 GFLOPS (1:32)
FP16 (TFLOPS)
81.72 TFLOPS (1:1)
23.44 TFLOPS (1:1)
AI/RT
RT Cores
—
56
Tensor Cores
—
224
Matrix Cores
1,216
—
Power
TDP
750 W
150 W
TDP (W)
750
150
-80.0%
Suggested PSU
1150 W
450 W
Power Connectors
None
8-pin EPS
Architecture
Architecture
CDNA 3.0
Ampere
GPU Name
Aqua Vanjaram
GA102
Generation
Instinct (MIx)
Server Ampere
(Axx)
Process Size
5 nm
8 nm
Transistors
153,000 million
28,300 million
Die Size
1017 mm²
628 mm²
Foundry
TSMC
Samsung
Density
150.4M / mm²
45.1M / mm²
AMD MCM
MCM
2
—
API Support
DirectX
—
12 Ultimate (12_2)
OpenGL
—
4.6
Vulkan
—
1.4
OpenCL
3.0
3.0
CUDA
—
8.6
Shader Model
—
6.8
Physical
Slot Width
OAM Module
Single-slot
Length
—
267 mm
10.5 inches
Height
—
112 mm
4.4 inches
Outputs
No outputs
No outputs
Bus Interface
PCIe 5.0 x16
PCIe 4.0 x16
Other
Production
—
End-of-life
Predecessor
Radeon Instinct
Tesla Turing
Successor
—
Server Ada