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
Tesla K20m
CORE STATE
GK110
VRAM
5 GB
CLOCK SPEED
—
TDP
225 W
BUS WIDTH
320 bit
ARCHITECTURE
Kepler
PROCESS
28 nm
LAUNCH DATE
2013
PERFORMANCE BENCHMARKS
geekbench_opencl
geekbench_vulkan
DETAILED SPECIFICATIONS
SPECIFICATION
Instinct MI300X
Tesla K20m
Core Specs
Shading Units
19,456
2,496
-87.2%
Shaders
19,456
2,496
-87.2%
TMUs
1,216
208
-82.9%
ROPs
0
40
+∞%
Compute Units
304
—
Clocks
Base Clock
1000 MHz
—
Boost Clock
2100 MHz
—
GPU Clock
—
706 MHz
Memory Clock
1300 MHz 5.2 Gbps effective
1300 MHz
5.2 Gbps effective
Memory
Memory Size
192 GB
5 GB
VRAM (MB)
196,608
5,120
-97.4%
Memory Type
HBM3
GDDR5
Memory Bus
8192 bit
320 bit
Bandwidth
5.32 TB/s
208.0 GB/s
Cache
L1 Cache
16 KB (per CU)
16 KB (per SMX)
L2 Cache
16 MB
1280 KB
L3 Cache
256 MB
—
Performance
Pixel Rate
0 MPixel/s
36.71 GPixel/s
Texture Rate
2,553.6 GTexel/s
146.8 GTexel/s
FP32 (TFLOPS)
81.72 TFLOPS
3.524 TFLOPS
FP64 (TFLOPS)
40.86 TFLOPS (1:2)
1,174.8 GFLOPS (1:3)
FP16 (TFLOPS)
81.72 TFLOPS (1:1)
—
AI/RT
Matrix Cores
1,216
—
Power
TDP
750 W
225 W
TDP (W)
750
225
-70.0%
Suggested PSU
1150 W
550 W
Power Connectors
None
1x 6-pin + 1x 8-pin
Architecture
Architecture
CDNA 3.0
Kepler
GPU Name
Aqua Vanjaram
GK110
Generation
Instinct (MIx)
Tesla Kepler
(Kxx)
Process Size
5 nm
28 nm
Transistors
153,000 million
7,080 million
Die Size
1017 mm²
561 mm²
Foundry
TSMC
TSMC
Density
150.4M / mm²
12.6M / mm²
AMD MCM
MCM
2
—
API Support
DirectX
—
12 (11_0)
OpenGL
—
4.6
Vulkan
—
1.2.175
OpenCL
3.0
3.0
CUDA
—
3.5
Shader Model
—
6.5 (5.1)
Physical
Slot Width
OAM Module
Dual-slot
Length
—
267 mm
10.5 inches
Outputs
No outputs
No outputs
Bus Interface
PCIe 5.0 x16
PCIe 2.0 x16
Other
Launch Price
—
3,199 USD
Production
—
End-of-life
Predecessor
Radeon Instinct
Tesla Fermi
Successor
—
Tesla Maxwell