4 ms·

Things that aren't possible on consumer hardware are implausible to be financially viable as an on demand service. Even for Google, it will never affordable to

by fooey 7y ago

Things that aren't possible on consumer hardware are implausible to be financially viable as an on demand service. Even for Google, it will never affordable to spin up giant clusters of servers for one person playing a video game. The basic premise of Stadia being able to match and surpass a high end desktop gaming machine makes no sense. The cloud isn't magic, and it takes people thinking it is magic to buy the pitch. Stadia reminds me of SimCity 2013 promising it had to be online based because of all the magic cloud power they needed to make it work, which was almost immediately proven to be utter bullshit. There's no way to offer the power of a $1,000 GPU without your own $1,000 GPU, and their promise of 4k at 60 FPS takes a very serious amount of specialized gaming hardware.

csunbird 7y ago
but you can rent a $1000 GPU for $100 per month and make a profit of $200 per year
drusepth 7y ago
>There's no way to offer the power of a $1,000 GPU without your own $1,000 GPU I think the whole idea of economics at scale (and much of cloud computing) is predicated on the idea that users want to use that $1,000 GPU without having to pay $1,000 to own it, and that companies can sell access to many users per $1,000 GPU to recoup the costs of that GPU, both over time (as the initial cost gets amortized over recurring subscription/usage revenues) and horizontally (selling access to many users who timeshare that GPU, maximizing its use compared to owning one you'd only use N hours per day). Scale that up to thousands of GPUs making a profit from hundreds of thousands of users each accessing those resources at a cost cheaper than what it'd take for them to purchase those resources for themselves and you've got the magic of the "cloud". tl;dr Users can pay <$1,000 for access to a $1,000 GPU and companies can afford to do so because the number of users multiplied by what they're paying exceeds the cost and upkeep of that GPU. Everyone* wins.

by fooey 7y ago

csunbird 7y ago

but you can rent a $1000 GPU for $100 per month and make a profit of $200 per year

drusepth 7y ago

>There's no way to offer the power of a $1,000 GPU without your own $1,000 GPU I think the whole idea of economics at scale (and much of cloud computing) is predicated on the idea that users want to use that $1,000 GPU without having to pay $1,000 to own it, and that companies can sell access to many users per $1,000 GPU to recoup the costs of that GPU, both over time (as the initial cost gets amortized over recurring subscription/usage revenues) and horizontally (selling access to many users who timeshare that GPU, maximizing its use compared to owning one you'd only use N hours per day). Scale that up to thousands of GPUs making a profit from hundreds of thousands of users each accessing those resources at a cost cheaper than what it'd take for them to purchase those resources for themselves and you've got the magic of the "cloud". tl;dr Users can pay <$1,000 for access to a $1,000 GPU and companies can afford to do so because the number of users multiplied by what they're paying exceeds the cost and upkeep of that GPU. Everyone* wins.