GPU Models
NVIDIA GPU nodes supporting CUDA 12.x
host |
GPU model |
Arch |
GPU RAM |
ECC |
nGPUs |
NVLink |
CC |
MPs |
cores/MP |
GPU driver |
CUDA driver |
|---|---|---|---|---|---|---|---|---|---|---|---|
m3 |
Tesla P4 |
Pascal |
8 GB |
y |
1 |
n |
6.1 |
20 |
128 |
535.154.05 |
12.2 |
m4 |
Tesla P4 |
Pascal |
8 GB |
y |
1 |
n |
6.1 |
20 |
128 |
535.154.05 |
12.2 |
g303 |
Tesla P4 |
Pascal |
8 GB |
y |
3 |
n |
6.1 |
20 |
128 |
535.154.05 |
12.2 |
g304 |
Tesla P4 |
Pascal |
8 GB |
y |
3 |
n |
6.1 |
20 |
128 |
535.154.05 |
12.2 |
g305 |
Tesla P4 |
Pascal |
8 GB |
y |
3 |
n |
6.1 |
20 |
128 |
535.154.05 |
12.2 |
g306 |
Tesla P4 |
Pascal |
8 GB |
y |
3 |
n |
6.1 |
20 |
128 |
535.154.05 |
12.2 |
g5 |
RTX 1080 Ti |
Pascal |
12 GB |
n |
2 |
n |
6.1 |
28 |
128 |
535.154.05 |
12.2 |
g8 |
Tesla V100 |
Volta |
32 GB |
y |
1 |
n |
7.0 |
80 |
64 |
535.154.05 |
12.2 |
g9 |
Tesla V100 |
Volta |
32 GB |
y |
1 |
n |
7.0 |
80 |
64 |
535.154.05 |
12.2 |
g6 |
RTX 2080 Ti |
Turing |
12 GB |
n |
2 |
n |
7.5 |
68 |
64 |
535.154.05 |
12.2 |
g7 |
RTX 2080 Ti |
Turing |
12 GB |
n |
2 |
n |
7.5 |
68 |
64 |
535.154.05 |
12.2 |
g10 |
RTX 2080 Ti |
Turing |
12 GB |
n |
2 |
n |
7.5 |
68 |
64 |
535.154.05 |
12.2 |
g11 |
RTX 2080 Ti |
Turing |
12 GB |
n |
2 |
n |
7.5 |
68 |
64 |
535.154.05 |
12.2 |
g12 |
RTX 2080 Ti |
Turing |
12 GB |
n |
2 |
n |
7.5 |
68 |
64 |
535.154.05 |
12.2 |
g13 |
A100 |
Ampere |
80 GB |
y |
4 |
y |
8.0 |
108 |
64 |
535.154.05 |
12.2 |
g14 |
RTX A6000 |
Ampere |
48 GB |
n |
2 |
y |
8.6 |
84 |
128 |
535.154.05 |
12.2 |
NVIDIA GPU nodes supporting CUDA 11.x
host |
GPU model |
Arch |
GPU RAM |
ECC |
nGPUs |
NVLink |
CC |
MPs |
cores/MP |
GPU driver |
CUDA driver |
|---|---|---|---|---|---|---|---|---|---|---|---|
g301 |
Tesla K40m |
Kepler |
12 GB |
y |
2 |
n |
3.5 |
15 |
192 |
470.57.02 |
11.4 |
g302 |
Tesla K40m |
Kepler |
12 GB |
y |
2 |
n |
3.5 |
15 |
192 |
470.57.02 |
11.4 |
g401 |
Tesla K40m |
Kepler |
12 GB |
y |
2 |
n |
3.5 |
15 |
192 |
470.57.02 |
11.4 |
g402 |
Tesla K40m |
Kepler |
12 GB |
y |
2 |
n |
3.5 |
15 |
192 |
470.57.02 |
11.4 |
The product of (MPs)*(cores/MP) is the total number of CUDA cores
CC = Compute capability
MP = streaming multiprocessors
Arch = NVIDIA GPU generation
AMD GPU nodes
host |
GPU model |
GPU RAM size |
nGPUs |
CUs |
cores/CU |
|---|---|---|---|---|---|
g500 |
VEGA 20 (gfx906) |
32 GB |
8 |
60 |
64 |
g501 |
VEGA 20 (gfx906) |
32 GB |
8 |
60 |
64 |
g502 |
VEGA 20 (gfx906) |
32 GB |
8 |
60 |
64 |
g503 |
VEGA 20 (gfx906) |
32 GB |
8 |
60 |
64 |
g504 |
VEGA 20 (gfx906) |
32 GB |
8 |
60 |
64 |
The product of (CUs)*(WaveFronts/CU) is the total number of “cores”.