NVIDIA B100 vs NVIDIA L20
We compared two Desktop platform GPUs: 96GB VRAM B100 and 48GB VRAM L20 to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.
Main Differences
NVIDIA B100 's Advantages
Released 1 years late
More VRAM (96GB vs 48GB)
Larger VRAM bandwidth (4.10TB/s vs 864.0GB/s)
5120 additional rendering cores
NVIDIA L20 's Advantages
Boost Clock has increased by 37% (2520MHz vs 1837MHz)
Lower TDP (275W vs 1000W)
Score
Benchmark
FP32 (float)
B100
+4%
62.08 TFLOPS
L20
59.35 TFLOPS
Graphics Card
Nov 2024
Release Date
Nov 2023
Server Blackwell
Generation
Tesla Ada
Desktop
Type
Desktop
PCIe 5.0 x16
Bus Interface
PCIe 4.0 x16
Clock Speeds
1665 MHz
Base Clock
1440 MHz
1837 MHz
Boost Clock
2520 MHz
2000 MHz
Memory Clock
2250 MHz
Memory
96GB
Memory Size
48GB
HBM3e
Memory Type
GDDR6
4096bit
Memory Bus
384bit
4.10TB/s
Bandwidth
864.0GB/s
Render Config
-
-
-
132
SM Count
92
16896
Shading Units
11776
528
TMUs
368
24
ROPs
128
528
Tensor Cores
368
-
RT Cores
92
256 KB (per SM)
L1 Cache
128 KB (per SM)
50 MB
L2 Cache
96 MB
-
-
-
Theoretical Performance
44.09 GPixel/s
Pixel Rate
322.6 GPixel/s
969.9 GTexel/s
Texture Rate
927.4 GTexel/s
248.3 TFLOPS
FP16 (half)
59.35 TFLOPS
62.08 TFLOPS
FP32 (float)
59.35 TFLOPS
31.04 TFLOPS
FP64 (double)
927.4 GFLOPS
Board Design
1000W
TDP
275W
1400 W
Suggested PSU
600 W
No outputs
Outputs
4x DisplayPort 1.4a
-
Power Connectors
1x 16-pin
Graphics Processor
GB102
GPU Name
AD102
-
GPU Variant
AD102
Blackwell
Architecture
Ada Lovelace
TSMC
Foundry
TSMC
5 nm
Process Size
5 nm
104 billion
Transistors
76.3 billion
Unknown
Die Size
609 mm²
Graphics Features
N/A
DirectX
12 Ultimate (12_2)
N/A
OpenGL
4.6
3.0
OpenCL
3.0
N/A
Vulkan
1.3
10.1
CUDA
8.9
N/A
Shader Model
6.7
