NVIDIA H100 vs AMD Radeon Instinct MI8
NVIDIA H100
AMD Radeon Instinct MI8
|
14592 Shaders
80GB HBM2e
1755MHz
|
4096 Shaders
4GB HBM
1000MHz
|
|
Peak AI Performance
3.03 POPS
INT8 Tensor Sparse
|
Peak AI Performance
8.19 TFLOPS
FP32
|
|
FP32
51.22 TFLOPS
|
FP32
8.19 TFLOPS
|
|
FP16
102.44 TFLOPS
|
-
-
|
|
Form Factor
PCIe Card
2.0-Slots
|
Form Factor
PCIe Card
2.0-Slots
|
|
TDP
350W
|
TDP
175W
|
|
Power Connectors
-
-
-
1x 16-Pin 12VHPWR
|
Power Connectors
-
1x 8-Pin
-
-
|
|
GB6 OpenCL N/A
0%
|
|
|
GB6 Metal N/A
0%
|
GB6 Metal N/A
0%
|
|
GB6 Vulkan N/A
0%
|
|
GB5 OpenCL N/A
0%
|
GB5 OpenCL N/A
0%
|
|
GB5 CUDA N/A
0%
|
GB5 CUDA N/A
0%
|
|
GB5 Metal N/A
0%
|
GB5 Metal N/A
0%
|
|
GB5 Vulkan N/A
0%
|
GB5 Vulkan N/A
0%
|
|
OCT 2020.1 N/A
0%
|
OCT 2020.1 N/A
0%
|
|
OCT Metal N/A
0%
|
OCT Metal N/A
0%
|
|
Peak AI
Performance
3.03 POPS
INT8 Tensor Sparse
|
Peak AI
Performance
8.19 TFLOPS
FP32
|
|
-
-
-
|
-
-
-
|
|
FP8
-
1.51 PFLOPS
Tensor (FP16 Accumulate)
3.03 PFLOPS
Tensor (FP16 Accumulate) Sparse
1.51 PFLOPS
Tensor (FP32 Accumulate)
3.03 PFLOPS
Tensor (FP32 Accumulate) Sparse
|
-
-
-
-
-
-
|
|
FP16
102.44 TFLOPS
756.45 TFLOPS
Tensor (FP16 Accumulate)
1.51 PFLOPS
Tensor (FP16 Accumulate) Sparse
756.45 TFLOPS
Tensor (FP32 Accumulate)
1.51 PFLOPS
Tensor (FP32 Accumulate) Sparse
|
-
-
-
-
-
-
|
|
FP32
51.22 TFLOPS
-
-
|
FP32
8.19 TFLOPS
-
-
|
|
FP64
25.61 TFLOPS
47.28 TFLOPS
Tensor
|
FP64
510 GFLOPS
-
|
|
BF16
102.44 TFLOPS
756.45 TFLOPS
Tensor
1.51 PFLOPS
Tensor Sparse
|
-
-
-
-
|
|
TF32
378.23 TFLOPS
Tensor
756.45 TFLOPS
Tensor Sparse
|
-
-
-
|
|
-
-
-
|
-
-
-
|
|
INT8
-
1.51 POPS
Tensor
3.03 POPS
Tensor Sparse
|
-
-
-
-
|
|
INT32
25.61 TOPS
|
-
-
|
|
-
-
|
-
-
|
|
Pixel
Fillrate
42.12 GPixel/s
|
Pixel
Fillrate
64 GPixel/s
|
|
-
-
|
-
-
|
|
Texture
Fillrate
800.28 GTexel/s
|
Texture
Fillrate
256 GTexel/s
|
|
Foundry
TSMC
-
|
Foundry
TSMC
-
|
|
Fabrication Node
4N
-
|
Fabrication Node
28nm
-
|
|
Die Size
814 mm²
-
|
Die Size
596 mm²
-
|
|
Transistor Count
80 Billion
-
|
Transistor Count
8.9 Billion
-
|
|
Transistor Density
98.28M/mm²
-
|
Transistor Density
14.93M/mm²
-
|
|
Form
PCIe Card
|
Form
PCIe Card
|
|
Shading Units
14592 Shaders
-
|
Shading Units
4096 Shaders
-
|
|
Texture Mapping Units
456 TMUs
|
Texture Mapping Units
256 TMUs
|
|
Render Output Units
24 ROPs
|
Render Output Units
64 ROPs
|
|
Tensor Cores
456 T-Cores
|
-
-
|
|
-
-
|
-
-
|
|
Streaming Multiprocessors
114 SMs
|
-
-
|
|
-
-
|
Compute Units
64 CUs
|
|
-
-
|
-
-
|
|
-
-
|
-
-
|
|
1620MHz Tensor
-
1065MHz Base
1755MHz
|
-
-
1000MHz Base
1000MHz
|
|
-
-
|
-
-
|
|
L1
64KB/SM Tex
256KB/SM
-
-
|
L1
-
-
16KB/CU
-
|
|
L2
50MB Shared
|
L2
2MB Shared
|
|
-
-
-
|
-
-
-
|
|
80GB
HBM2e
-
|
4GB
HBM
ECC
|
|
Bus Width
5120Bit
|
Bus Width
4096Bit
|
|
Clock
1593MHz
Transfer Rate
3.2GT/s
Bandwidth
2039GB/s
|
Clock
500MHz
Transfer Rate
1GT/s
Bandwidth
512GB/s
|
|
-
-
-
-
-
-
-
-
-
|
-
-
-
-
-
-
-
-
-
|
|
TDP
350W
|
TDP
175W
|
|
-
-
|
-
-
|
|
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
No Ports
|
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
No Ports
|
|
Max Resolution
Unknown
|
Max Resolution
4096x2160
|
|
Max Resolution Refresh Rate
-
|
Max Resolution Refresh Rate
60Hz
|
|
Variable Refresh Rate
G-Sync
FreeSync
-
|
Variable Refresh Rate
-
FreeSync
-
|
|
Display Stream Compression (DSC)
Not Supported
|
Display Stream Compression (DSC)
Not Supported
|
|
Multi Monitor Support
Unknown
|
Multi Monitor Support
4
|
|
-
-
|
Content Protection
HDCP 1.4
|
|
No Encoders
-
|
Model
VCE 3.0
|
|
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
|
Codec
-
-
-
-
-
-
-
-
AVC (H.264)
HEVC (H.265)
-
-
-
-
|
|
Model
7x NVDEC 5
|
Model
UVD 6.0
|
|
Codec
MPEG-1
MPEG-2
MPEG-4
-
VC-1
VP8
VP9
-
AVC (H.264)
HEVC (H.265)
-
AV1
-
-
|
Codec
MPEG-1
MPEG-2
MPEG-4
JPEG
VC-1
-
-
-
AVC (H.264)
HEVC (H.265)
-
-
-
-
|
|
-
-
-
-
|
Direct X
12
Direct 3D
12_0
|
|
-
-
OpenCL
3.0
-
-
|
OpenGL
4.6
OpenCL
2.1
Vulkan
1.2
|
|
-
-
CUDA
9.0
-
-
PureVideo HD
VP11
VDPAU
Feature Set K
|
Shader Model
6.5
-
-
GFX
8
-
-
-
-
|
|
-
-
-
-
|
-
-
-
-
|
|
Power Connectors
-
-
-
-
-
-
1x 16-Pin 12VHPWR
|
Power Connectors
-
-
-
1x 8-Pin
-
-
-
|
|
Slots Required
2.0
PCIe Version
5.0
PCIe Lanes
16
|
Slots Required
2.0
PCIe Version
3.0
PCIe Lanes
16
|
|
Multi GPU Support
Supported
Type
NVLink
|
Multi GPU Support
Supported
Type
CrossFire XDMA
|
|
Height
111 mm (4.37 in)
Width
268 mm (10.55 in)
Depth
40 mm (1.57 in)
|
Height
111 mm (4.37 in)
Width
152 mm (5.98 in)
Depth
37 mm (1.46 in)
|
Copy Link