Intel Gaudi vs NVIDIA T40

Intel Gaudi
NVIDIA T40
2048 Shaders
32GB HBM2
610MHz
4608 Shaders
24GB GDDR6
1560MHz
Peak AI Performance
159.91 TFLOPS
BF16 Tensor
Peak AI Performance
460.06 TOPS
INT4 Tensor
-
-
FP32
14.38 TFLOPS
-
-
FP16
28.75 TFLOPS
Form Factor
OAM Module
-
Form Factor
PCIe Card
1.0-Slots
TDP
350W
TDP
150W
-
-
-
-
-
Power Connectors
1x 6-Pin
1x 8-Pin
-
-
GB6 OpenCL N/A
0%
GB6 OpenCL N/A
0%
GB6 Metal N/A
0%
GB6 Metal N/A
0%
GB6 Vulkan 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
159.91 TFLOPS
BF16 Tensor
Peak AI Performance
460.06 TOPS
INT4 Tensor
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
FP16
28.75 TFLOPS
115.02 TFLOPS Tensor (FP16 Accumulate)
-
115.02 TFLOPS Tensor (FP32 Accumulate)
-
-
-
-
-
FP32
14.38 TFLOPS
-
-
-
-
-
FP64
450 GFLOPS
-
BF16
-
159.91 TFLOPS Tensor
-
-
-
-
-
-
-
-
-
-
-
-
-
-
INT4
460.06 TOPS Tensor
-
-
-
-
-
INT8
-
230.03 TOPS Tensor
-
-
-
-
-
-
-
Ray Tracing
56.2 TOPS
Pixel Fillrate
-
Pixel Fillrate
149.76 GPixel/s
-
-
-
-
Texture Fillrate
-
Texture Fillrate
449.28 GTexel/s
Manufacturer
Intel
Manufacturer
NVIDIA
Chip Designer
Intel
Chip Designer
NVIDIA
Architecture
Gaudi
Architecture
Turing
Family
Gaudi
Family
Server
Codename
Gaudi
HL-2000
Variant
HL-2000
Codename
NV162
TU102
Variant
TU102-890-KCD-A1
Market Segment
Server
Market Segment
Server
Release Date
6/17/2019
Release Date
9/13/2018
Foundry
TSMC
-
Foundry
TSMC
-
Fabrication Node
16FF
-
Fabrication Node
12FFN
-
-
-
-
Die Size
754 mm²
-
-
-
Transistor Count
18.6 Billion
-
-
-
-
Transistor Density
24.67M/mm²
-
Form
OAM Module
Form
PCIe Card
Shading Units
2048 Shaders
-
Shading Units
4608 Shaders
-
Texture Mapping Units
-
Texture Mapping Units
288 TMUs
Render Output Units
-
Render Output Units
96 ROPs
Tensor Cores
8 T-Cores
Tensor Cores
576 T-Cores
-
-
Ray-Tracing Cores
72 RT-Cores
-
-
Streaming Multiprocessors
72 SMs
Compute Units
1 CU
-
-
-
-
-
-
-
-
-
-
-
-
-
610MHz
-
-
1305MHz Base
1560MHz
-
-
-
-
L1
-
-
-
Unknown
L1
32KB/SM Tex
64KB/SM
-
-
L2
Unknown
L2
6MB Shared
-
-
-
-
-
-
32GB
HBM2
ECC
24GB
GDDR6
-
Bus Width
4096Bit
Bus Width
384Bit
Clock
980MHz
Transfer Rate
2GT/s
Bandwidth
1003.5GB/s
Clock
1625MHz
Transfer Rate
13GT/s
Bandwidth
624GB/s
-
-
-
-
eSRAM
24MB
3200GB/s
-
-
-
-
-
-
-
-
-
-
-
TDP
350W
TDP
150W
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
No Ports
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
No Ports
Max Resolution
Unknown
Max Resolution
Unknown
Max Resolution Refresh Rate
-
Max Resolution Refresh Rate
-
Variable Refresh Rate
-
-
-
Variable Refresh Rate
G-Sync
FreeSync
-
Display Stream Compression (DSC)
Not Supported
Display Stream Compression (DSC)
Not Supported
Multi Monitor Support
Unknown
Multi Monitor Support
Unknown
-
-
-
-
No Encoders
-
Model
NVENC 7
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Codec
-
-
-
-
-
-
-
-
AVC (H.264)
HEVC (H.265)
-
-
-
-
No Decoders
Model
NVDEC 4
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Codec
MPEG-1
MPEG-2
MPEG-4
-
VC-1
VP8
VP9
-
AVC (H.264)
HEVC (H.265)
-
-
-
-
-
-
-
-
Direct X
12
Direct 3D
12_2
-
-
OpenCL
3.0
-
-
OpenGL
4.6
OpenCL
3.0
Vulkan
1.2
-
-
-
-
-
-
-
-
-
-
Shader Model
6.6
CUDA
7.5
-
-
PureVideo HD
VP10
VDPAU
Feature Set J
Not a Card
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Power Connectors
-
-
1x 6-Pin
1x 8-Pin
-
-
-
-
-
PCIe Version
4.0
PCIe Lanes
16
Slots Required
1.0
PCIe Version
3.0
PCIe Lanes
16
Multi GPU Support
Supported
Type
RoCE
Multi GPU Support
Supported
Type
NVLink
-
-
-
-
-
-
Height
111 mm (4.37 in)
Width
267 mm (10.51 in)
Depth
20 mm (0.79 in)
Change Comparison