GPU Comparison
GEFORCE
NVIDIA H200 NVL
CORE STATE
GH100
VRAM
141 GB
CLOCK SPEED
1785 MHz
TDP
600 W
BUS WIDTH
6144 bit
ARCHITECTURE
Hopper
PROCESS
5 nm
LAUNCH DATE
2024
VS
GEFORCE
RTX A4000
CORE STATE
GA104
VRAM
16 GB
CLOCK SPEED
1560 MHz
TDP
140 W
BUS WIDTH
256 bit
ARCHITECTURE
Ampere
PROCESS
8 nm
LAUNCH DATE
2021
PERFORMANCE BENCHMARKS
geekbench_opencl
3dmark_3dmark_steel_nomad_dx12
geekbench_vulkan
passmark_directx_10
passmark_directx_11
passmark_directx_12
passmark_directx_9
passmark_g2d
passmark_g3d
passmark_gpu_compute
DETAILED SPECIFICATIONS
SPECIFICATION
H200 NVL
RTX A4000
Core Specs
Shading Units
16,896
6,144
-63.6%
Shaders
16,896
6,144
-63.6%
TMUs
528
192
-63.6%
ROPs
24
96
+300.0%
SM Count
132
48
-63.6%
Clocks
Base Clock
1365 MHz
735 MHz
Boost Clock
1785 MHz
1560 MHz
Memory Clock
1593 MHz
6.4 Gbps effective
1750 MHz
14 Gbps effective
Memory
Memory Size
141 GB
16 GB
VRAM (MB)
144,384
16,384
-88.7%
Memory Type
HBM3e
GDDR6
Memory Bus
6144 bit
256 bit
Bandwidth
4.89 TB/s
448.0 GB/s
Cache
L1 Cache
256 KB (per SM)
128 KB (per SM)
L2 Cache
50 MB
4 MB
Performance
Pixel Rate
42.84 GPixel/s
149.8 GPixel/s
Texture Rate
942.5 GTexel/s
299.5 GTexel/s
FP32 (TFLOPS)
60.32 TFLOPS
19.17 TFLOPS
FP64 (TFLOPS)
30.16 TFLOPS (1:2)
299.5 GFLOPS (1:64)
FP16 (TFLOPS)
241.3 TFLOPS (4:1)
19.17 TFLOPS (1:1)
AI/RT
RT Cores
—
48
Tensor Cores
528
192
-63.6%
Power
TDP
600 W
140 W
TDP (W)
600
140
-76.7%
Suggested PSU
1000 W
300 W
Power Connectors
8-pin EPS
1x 6-pin
Architecture
Architecture
Hopper
Ampere
GPU Name
GH100
GA104
Generation
Server Hopper
(Hxx)
Workstation Ampere
(Ax000)
Process Size
5 nm
8 nm
Transistors
80,000 million
17,400 million
Die Size
814 mm²
392 mm²
Foundry
TSMC
Samsung
Density
98.3M / mm²
44.4M / mm²
API Support
DirectX
—
12 Ultimate (12_2)
OpenGL
—
4.6
Vulkan
—
1.4
OpenCL
3.0
3.0
CUDA
9.0
8.6
Shader Model
—
6.8
Physical
Slot Width
Dual-slot
Single-slot
Length
267 mm
10.5 inches
241 mm
9.5 inches
Height
111 mm
4.4 inches
112 mm
4.4 inches
Outputs
No outputs
4x DisplayPort 1.4a
Bus Interface
PCIe 5.0 x16
PCIe 4.0 x16
Other
Production
Active
End-of-life
Predecessor
Server Ada
Quadro Turing
Successor
Server Blackwell
Workstation Ada