I was excited when Storj purchased Valdi, in that I may be able to setup a GPU rig to earn Storj. Is this still in the works? What is the timeline?
I assumed they’d stick to their own enterprise cards. Do you think they’re going to compete with salad/vast/clore for GPU rentals?
Not sure, but thought it would be a good fit?
It will be interesting to see what market they chase! I’ve been watching another company getting into bulk-gpu-rentals (www.silicon.net) and in one of their videos they mention that there’s a good market for rentals of multiple-enterprise/high-memory GPUs in well-configured environments (like 4-8 H100/A100/A6000, in systems with 256GB+ RAM, lots of NVMe space, and fast internet). And weak demand for someone renting low-VRAM (like a single 8GB 3060) GPU in a system with 16GB of RAM, maybe no flash, and a 250/50 internet connection.
Basicallly… if you aren’t bringing at least a few 24GB 3090/4090s to the party in a beefy computer… it’s not really worth their time to try to resell it in their marketplace.
But… obviously other projects are reselling some weak-sauce configs… so Storj may try too.
2 Likes
That’s exactly what I would like to get into, large investment upfront, but if the market is there, then this could be much easier than running storj nodes. I think a datacenter level infrastructure would be the only way to go when hosting one of these machines.
I think they are using surplus in exisiting data centers and not individual cards from enthusiasts. That could always change, but the model (as I understand it) is connecting GPU farm with GPU customer.
You could look at the founder’s (Shawn) other venture Prodia.com I believe they have a model for individual card holders to resell their GPU bandwidth but I dont think that aspect is currently available for new suppliers. It would require further investigation. I have only cursory knowledge of it.
Right now we’re just using commercial/enterprise facilities. It’s an exciting concept and there are a number of web3 projects doing decentralized compute in various ways (containers, GPUs, CPU, FaaS, etc.). We may explore it in the future; right now our focus is on making high-quality GPUs available to customers.
4 Likes