Hey guys, wanna rent a GPU for a month?

Hey guys, wanna rent a GPU for a month?

Attached: linode-gpu-hosting.png (767x549, 43K)

Other urls found in this thread:

vast.ai/console/create/
twitter.com/SFWRedditVideos

Trying to think of something to say but I have no words. Why the FUCK wouldn’t you just buy a 1000 dollar gpu?

Even considering the hosting isn't that a bit too expensive?

Because an RTX6000 is much more powerful and is 5-6k to buy.

Because you're after that sweet, sweet, 40/10Gbit network?

Yes, the list price of RTX 6000 is just $4,000.

>nearly 1/4 the cost to rent

that's a horrible deal unless you really only need them for 1-2 months.

This is sad.

vast.ai/console/create/

Landlords should list their rent in dollars/hour too

A legal DDos attack to see how much traffic can handle facebook server....

Attached: snowden.jpg (825x464, 291K)

but right now i don't have money to do that.....

>Ubuntu 16.04.4 (required)
>Fast, reliable internet: at least 10Mbps per machine.
uh

I doubt anyone rents them for constant long time use, not even weeks, only when you need to render/calculate something 'one and done' - so even 6$/hr is not much when you might render your project in ~10 hrs and thats it

What's the cheapest/best place to rent GPU time right now? I know Scaleway offers €100 of free credits when you sign up, I'm hearing stuff about Google collab, there's Amazon's free tier which may offer something... what's the best place / best deal in all of this?

Is this the future? People prefer to rent property, prefer to rent their own phone, lease a car, and now renting a GPU or game streaming will become the norm. For what fucking reason? If you're doing any sort of work that actually requires this much GPU based data crunching, why not just build your own fucking machine as a one time deal?

>For what fucking reason?
Because you don't need that kind of power all the time, only when training models, which for a lot of people are just a few hours here and there. You can do the rest of the stuff on a smaller machine, yet still have access to essentially up to date supercomputers (in Google's and Amazon's cases at least) for not that much money, all things considered, or at least a lot less upfront. The fact that there's virtually no maintenance or support worries is also a big plus for some.

I suppose. I just feel that if you're working with the kind of data that requires 10+ hours of crunching at a time, you'd benefit from just getting even a mediocre quadro. I.e. instead of paying $60. Sure your data processing may take 20 hours now, but you own the hardware as a one time buy. Seems to be pros and cons of both, I just am not a fan of renting or paying for someone else's stuff as a service.

$4000 GPU but i'm going to pay you $1k a month to use one? LMFAO
other components are literally like $500/$600

The absolute state of cloud computing faggots

Serious question, what happens to all of these faggots when someone comes up with a far more efficient algorithm that doesn't require you literally brute-force a state space and come up w/ a meme learning bot? Lets say something that can run on an average desktop computer in someone's home that doesn't require a week straight of meme learning/inference

they need to charge that much in order to make a profit
it's likely that those servers are idle much of the time, because anyone who would need that much power that often would just buy their own card

>meme learning/inference

Redpill me on this broski

For a company, at 1.5 an hour over the course of a month that's nothing.

This is for virtual servers which require parallel processing (whether graphical or otherwise). Probably priced competitively with similar AWS services.