> Nvidia contractually prohibits the use of GeForce and Titan cards in datacenters. So Amazon and other providers have to use the $8,500 datacenter version of the GPUs, and they have to charge a lot for renting it.
How can Nvidia anymore enforce that than movie producers can stop rentals of physical media.
Building your own deep learning computer is 10x cheaper than AWS
11–20 of 269 posts
Re: Building your own deep learning computer is 10x cheaper than AWS
#12> Nvidia contractually prohibits the use of GeForce and Titan cards in datacenters. So Amazon and other providers have to use the $8,500 datacenter version of the GPUs, and they have to charge a lot for renting it.
How can Nvidia anymore enforce that than movie producers can stop rentals of physical media.
Re: Building your own deep learning computer is 10x cheaper than AWS
#13> Nvidia contractually prohibits the use of GeForce and Titan cards in datacenters. So Amazon and other providers have to use the $8,500 datacenter version of the GPUs, and they have to charge a lot for renting it.
Re: Building your own deep learning computer is 10x cheaper than AWS
#14Of course building is cheaper than AWS... It also means you need to build the computer, maintain it, upgrade it, store it somewhere, connect it to the internet, and so on. And you can't just scale it if you need to. I mean, the cloud's here for a reason...
Re: Building your own deep learning computer is 10x cheaper than AWS
#15> Nvidia contractually prohibits the use of GeForce and Titan cards in datacenters. So Amazon and other providers have to use the $8,500 datacenter version of the GPUs, and they have to charge a lot for renting it.
So I guess in theory it would be possible for AWS to develop their own (or enhance open source) drivers. On the other hand they would spoil the business relationship with Nvidia and have to do without any discounts.
Re: Building your own deep learning computer is 10x cheaper than AWS
#16Of course building is cheaper than AWS... It also means you need to build the computer, maintain it, upgrade it, store it somewhere, connect it to the internet, and so on. And you can't just scale it if you need to. I mean, the cloud's here for a reason...
Why "of course"? Last time I checked baking my own cookies is more expensive than just buying a bag at Walmart.
Re: Building your own deep learning computer is 10x cheaper than AWS
#17Re: Building your own deep learning computer is 10x cheaper than AWS
#18Of course building is cheaper than AWS... It also means you need to build the computer, maintain it, upgrade it, store it somewhere, connect it to the internet, and so on. And you can't just scale it if you need to. I mean, the cloud's here for a reason...
> Of course building is cheaper than AWS... Why "of course"? Last time I checked baking my own cookies is more expensive than just buying a bag at Walmart.
If you need 5 cookies today (and maybe the next batch next week), go to Walmart. Same for an one-off DL job.
Re: Building your own deep learning computer is 10x cheaper than AWS
#19> Nvidia contractually prohibits the use of GeForce and Titan cards in datacenters. So Amazon and other providers have to use the $8,500 datacenter version of the GPUs, and they have to charge a lot for renting it.
Hasn't stopped Hetzner from offering 1080s in their servers.
Re: Building your own deep learning computer is 10x cheaper than AWS
#20Earlier quoted context omitted.
> Of course building is cheaper than AWS... Why "of course"? Last time I checked baking my own cookies is more expensive than just buying a bag at Walmart.
If you need 10k cookies every day for multiple years, you're better off building your own bakery. If you need multiple years of continuous DL work, you're better off building your own DL system. If you need 5 cookies today (and maybe the next batch next week), go to Walmart. Same for an one-off DL job.