New pricing: More AI power, less cost!
Learn more

GPU Cloud

AMD MI300X

192 GB
3.2 Tbps Network
Deploy
 On-Demand

AMD MI250

128 GB
1.6 Tbps Network
runpod demo
opencv logo
replika logo
datasciencedojo logo
jina logo
defined ai logo
otovo logo
abzu logo
aftershoot logo
krnl logo
opencv logo
replika logo
datasciencedojo logo
jina logo
defined ai logo
otovo logo
abzu logo
aftershoot logo
krnl logo
AMD Instinct™ MI300X
On-Demand MI300X GPUs available.
Scale up to hundreds of MI300X on-demand.
Deploy On-Demand
1-month to 3-year reservations available.
Schedule a time with our team to learn more.
Book a Call
MI300X
H100 SXM
VRAM
192 GB
80 GB
Memory Bandwidth
5.3 TB/s
3.4 TB/s
FP64
81.7 TFLOPs
67 TFLOPs
FP32
163.4 TFLOPs
67 TFLOPs
FP16
2,610 TFLOPs
1,979 TFLOPs
FP8
5,220 TFLOPs
3,958 TFLOPs
Interconnect
GPU to GPU
PCIe 128 GB/s
NVLink 900 GB/s
PCIe 128 GB/s
Network
3.2 Tb/s
-
AMD Instinct MI300X GPU
AMD Instinct™ MI250
1-month to 3-year reservations available.
Schedule a time with our team to learn more.
On-Demand options coming soon.
Arriving as early as August 2024.
Book a Call
MI250
A100 SXM
VRAM
128 GB
80 GB
Memory Bandwidth
3.2 TB/s
2 TB/s
FP64 / FP32
45.3 TFLOPs
19.5 TFLOPS
FP16
362 TFLOPs
312 TFLOPs
Interconnect
GPU to GPU
PCIe 100 GB/s
NVLink 600 GB/s
PCIe 64 GB/s
Network
1.6 Tb/s
-
AMD Instinct MI250 GPU

Scale up to 10,000+ GPUs seamlessly

Scale quickly to thousands of GPUs without sacrificing cost or reliability.