Didn't realize you made this tool. It's super useful.
Some unsolicited feedback, if you're still actively developing:
- You should consider some of the lesser known cloud providers (e.g. Coreweave/Lambda Labs/TensorDock).
- Add information about whether the servers support NVLink/NVswitch. For example, A100s come in 3 flavors: PCIe without NVLink bridges, PCIe with NVLink bridges, and SXM with NVlink/NVSwitch fabric.
GPU pricing of alternative clouds (lowest price to highest):
[A100 PCI]
Lambda Labs: $1.10/hr
TensorDock: $2.06/hr
Coreweave: $2.46/hr
Paperspace: $3.09/hr
[A100 SXM]
Lambda Labs: $1.25/hr
TensorDock: $2.06/hr
Coreweave: N/A (I think PCI only)
Paperspace: N/A (I think PCI only)
[A40]
TensorDock: $1.28/hr
Coreweave: $1.68/hr
Paperspace: N/A
Lambda Labs: N/A
[A6000]
Lambda Labs: $0.80/hr
Tensordock: $1.28/hr
Paperspace: $1.89/hr
Coreweave: $1.68/hr
[V100 SXM4]
Lambda Labs: $0.55/hr
TensorDock: $0.80/hr
Coreweave: $1.00/hr
Paperspace: $2.30/hr
[A5000]
TensorDock: $0.77/hr
Coreweave: $1.01/hr
Paperspace: $1.38/hr
Lambda Labs: N/A
Jonathan thanks for the post. A question: it sounds like TensorDock partners with 3rd-parties who bought these servers and TensorDock doesn't actually own any of the servers you rent out. If that's the case, how do you ensure security? If not, please ignore.
Lambda/Paperspace/Coreweave do own their own servers (Lambda being the cheapest). That alleviates some security concerns.
It all depends on your security requirements though.