GeForce GTX 680 vs A40 PCIe
We compared two Desktop platform GPUs: 2GB VRAM GeForce GTX 680 and 48GB VRAM A40 PCIe to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.
Main Differences
GeForce GTX 680 's Advantages
Lower TDP (195W vs 300W)
A40 PCIe 's Advantages
Released 8 years and 7 months late
Boost Clock has increased by 64% (1740MHz vs 1058MHz)
More VRAM (48GB vs 2GB)
Larger VRAM bandwidth (695.8GB/s vs 192.3GB/s)
9216 additional rendering cores
Score
Benchmark
FP32 (float)
GeForce GTX 680
3250
A40 PCIe
+1051%
37420
Graphics Card
Mar 2012
Release Date
Oct 2020
GeForce 600
Generation
Tesla
Desktop
Type
Desktop
PCIe 3.0 x16
Bus Interface
PCIe 4.0 x16
Clock Speeds
1006 MHz
Base Clock
1305 MHz
1058 MHz
Boost Clock
1740 MHz
1502 MHz
Memory Clock
1812 MHz
Memory
2GB
Memory Size
48GB
GDDR5
Memory Type
GDDR6
256bit
Memory Bus
384bit
192.3GB/s
Bandwidth
695.8GB/s
Render Config
-
SM Count
84
1536
Shading Units
10752
128
TMUs
336
32
ROPs
112
-
Tensor Cores
336
-
RT Cores
84
16 KB (per SMX)
L1 Cache
128 KB (per SM)
512 KB
L2 Cache
6 MB
Theoretical Performance
33.86 GPixel/s
Pixel Rate
194.9 GPixel/s
135.4 GTexel/s
Texture Rate
584.6 GTexel/s
-
FP16 (half)
37.42 TFLOPS
3.250 TFLOPS
FP32 (float)
37.42 TFLOPS
135.4 GFLOPS
FP64 (double)
584.6 GFLOPS
Board Design
195W
TDP
300W
450 W
Suggested PSU
700 W
2x DVI
1x HDMI 1.4a
1x DisplayPort 1.2
Outputs
3x DisplayPort 1.4a
2x 6-pin
Power Connectors
8-pin EPS
Graphics Processor
GK104
GPU Name
GA102
GK104-400-A2
GPU Variant
-
Kepler
Architecture
Ampere
TSMC
Foundry
Samsung
28 nm
Process Size
8 nm
3.54 billion
Transistors
28.3 billion
294 mm²
Die Size
628 mm²
Graphics Features
12 (11_0)
DirectX
12 Ultimate (12_2)
4.6
OpenGL
4.6
3.0
OpenCL
3.0
1.1
Vulkan
1.3
3.0
CUDA
8.6
5.1
Shader Model
6.6
