Hey all! I've been hosting a small fleet of GPU servers for 3 years now on a public marketplace.

A couple of months ago I started building my own GPU Cloud as I believe this way users can get a better service in terms of support, reliability and functionality. Now it's ready for the first users.

So why another GPU Cloud? Well, firstly, high quality and reasonably priced cloud GPU providers are constantly out of capacity.

Another reason is the gripe I have with GPU marketplaces. It's difficult to know whether the server you're renting is in a proper data center or in someone's hot and humid basement. This leads to reliability and performance issues which drags down the quality (and therefore the price) of the whole platform. This incentivises building cheap unreliable servers to save on cost.

Additionally, the current GPU marketplaces are all focused on hosting Docker containers instead of virtual machines. This is fine for most use cases but it is a fundamental limitation. I've seen plenty of user complaints about not being able to run their image because of a driver/kernel version mismatch. There are many more reasons which I could go into.

The goal of Backprop is to address these problems. We're hosted in a tier III data center. All instances are virtual machines with root access and fully dedicated IPv4 addresses. We're reasonably priced (hopefully you agree) with no additional charges on storage/bandwidth.

We currently only have RTX 3090 servers but we're looking to expand our capacity soon.

Please share your feedback!

PS! We're looking to build a high quality marketplace so if you know a host that meets these criteria, please get in touch!