GPU Comparison
GEFORCE
NVIDIA B200
CORE STATE
GB100
VRAM
90 GB
CLOCK SPEED
1965 MHz
TDP
1000 W
BUS WIDTH
4096 bit
ARCHITECTURE
Blackwell
PROCESS
5 nm
LAUNCH DATE
—
VS
GEFORCE
GeForce RTX 3070
CORE STATE
GA104
VRAM
8 GB
CLOCK SPEED
1725 MHz
TDP
220 W
BUS WIDTH
256 bit
ARCHITECTURE
Ampere
PROCESS
8 nm
LAUNCH DATE
2020
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
B200
RTX 3070
Core Specs
Shading Units
18,944
5,888
-68.9%
Shaders
18,944
5,888
-68.9%
TMUs
592
184
-68.9%
ROPs
24
96
+300.0%
SM Count
148
46
-68.9%
Clocks
Base Clock
700 MHz
1500 MHz
Boost Clock
1965 MHz
1725 MHz
Memory Clock
2000 MHz
8 Gbps effective
1750 MHz
14 Gbps effective
Memory
Memory Size
90 GB
8 GB
VRAM (MB)
92,160
8,192
-91.1%
Memory Type
HBM3e
GDDR6
Memory Bus
4096 bit
256 bit
Bandwidth
4.10 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
47.16 GPixel/s
165.6 GPixel/s
Texture Rate
1,163.3 GTexel/s
317.4 GTexel/s
FP32 (TFLOPS)
74.45 TFLOPS
20.31 TFLOPS
FP64 (TFLOPS)
37.22 TFLOPS (1:2)
317.4 GFLOPS (1:64)
FP16 (TFLOPS)
1,191.2 TFLOPS (16:1)
20.31 TFLOPS (1:1)
AI/RT
RT Cores
—
46
Tensor Cores
592
184
-68.9%
Power
TDP
1000 W
220 W
TDP (W)
1,000
220
-78.0%
Suggested PSU
1400 W
550 W
Power Connectors
—
1x 12-pin
Architecture
Architecture
Blackwell
Ampere
GPU Name
GB100
GA104
Generation
Server Blackwell
(Bxx)
GeForce 30
Process Size
5 nm
8 nm
Transistors
104,000 million
17,400 million
Die Size
—
392 mm²
Foundry
TSMC
Samsung
Density
—
44.4M / mm²
API Support
DirectX
—
12 Ultimate (12_2)
OpenGL
—
4.6
Vulkan
—
1.4
OpenCL
3.0
3.0
CUDA
10.0
8.6
Shader Model
—
6.8
Physical
Slot Width
SXM Module
Dual-slot
Length
—
242 mm
9.5 inches
Height
—
112 mm
4.4 inches
Outputs
No outputs
1x HDMI 2.13x DisplayPort 1.4a
Bus Interface
PCIe 5.0 x16
PCIe 4.0 x16
Other
Launch Price
—
499 USD
Production
Active
End-of-life
Predecessor
Server Hopper
GeForce 20
Successor
Server Rubin
GeForce 40