Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Stadia died because no one trusts Google

techcrunch.com

Post reply on HN
Print view View original

561–570 of 626 posts

Re: Stadia died because no one trusts Google

#561

Post by impossiblefork »

panick21_ wrote: post #548
impossiblefork wrote: post #309

Earlier quoted context omitted.

The Gigapress is an Italian innovation though, and the Italians built it because they understood that it could be incredibly useful, not because there was a custom order. Tesla were early adopters by making use of them, but not innovators.

Tesla specifically came to them and asked them to developed a machine to their specifications, financed and developed the product and they are getting virtually all the machines the company can produce. Also its a part of company from China operating in Italy. These machines are also continuously improved in prosecution at Tesla and that knowlage is flowing back to the company. The machines deployed in Texas are alre…

It's an Italian company that was very recently bought by a Chinese company.

My understanding is that they saw the usefulness of these machines quite early and were developing them in this direction and that they had already produced 4000 tonne presses of this general type. Now it's at 6000 and 8000 tonnes.

Customers are needed to make use of good ideas-- you can't take them otherwise, but the innovation is the work the Italians did.

Re: Stadia died because no one trusts Google

#562

Post by maximus-decimus »

tgsovlerkhgsel wrote: post #56
j-kent wrote: post #19

Imagine if you had a Netflix subscription, and still had to pay for every movie you watched. This is what Stadia felt like to me.

Isn't that basically the model Disney has?

Most titles on Disney+ are included. Just not the very very new stuff, sometimes.

Re: Stadia died because no one trusts Google

#563

Post by chrisseaton »

solardev wrote: post #558
chrisseaton wrote: post #31

Earlier quoted context omitted.

Streaming Netflix is relatively simple - each customer just needs to read a stored stream. I think they even put dedicated servers at edges like ISPs, so it's not even on the public internet! Streaming a game requires an actual real GPU being (I guess) exclusively used by that customer while they're playing. Not feasible to put those in edge locations. Completely different set of costs.

Somehow Nvidia, Shadow, PSNow, Xbox Cloud, Luna, Paper, etc. manage to do this...

Manage to do what? Use edge locations? I'm not sure they do - where did you read that?

Re: Stadia died because no one trusts Google

#564

Post by mrkramer »

choppaface wrote: post #218
stingraycharles wrote: post #150

Earlier quoted context omitted.

At least for Google, I don’t think you can call what they did with search low hanging fruit. Their solution was anything but, and it was truly innovative (at the time) and their success was well deserved. I do agree that they’re victims of their early success, but rather that it proves that it’s rather difficult to make multiple successful products. When you own a whole ecosystem, like Microsoft and Apple do, it make…

Search was full of low-hanging fruit. TF-IDF was well-known. PageRank less so but when the nascent internet has so much signal-to-noise it’s often hard to go wrong. Execution was key—- especially compute efficiency—- but a lot of those are nerdy problems detached from real people. Which is wear Google does well. Stadia was a rich technical problem with an application area that’s accessible to Googlers. Even if Google…

>Search was full of low-hanging fruit. TF-IDF was well-known. PageRank less so but when the nascent internet has so much signal-to-noise it’s often hard to go wrong.

PageRank actually originates from Economics[0] where Input–output model tells you that you need to care about balancing inputs and outputs in order to have an efficient economy. Speaking of internet search engines and ranking websites inputs would be links(backlinks) and outputs would be ranked web documents.

[0] https://en.wikipedia.org/wiki/Input%E2%80%93output_model

Re: Stadia died because no one trusts Google

#565

Post by mrkramer »

jrockway wrote: post #247
h0l0cube wrote: post #131

Earlier quoted context omitted.

> As a gamer, I think Stadia failed because the system they created doesn't make sense. Google, Facebook, etc. are victims of early success. They made their billions on low hanging fruit, by throwing a lot of resources at problems with very high demand for a solution that weren't yet tackled well (e.g. query the internet, keep in touch with friends). So it's no wonder that in this day in age they are incapable of und…

I don't think people at Google have a problem innovating or understanding product/market fit. Rather, they have 100,000 employees (think about that for a second, how long it would take to count that many people), and not everyone's project is a good idea. That doesn't mean people don't try hard, it just means that there isn't necessarily a business just because they want there to be. I think the problem is simply tha…

Larry Page said he is worried because of Google was and is doing too many things. But they simply have to do it because their cash pile is so big that it can't just hang out in bank for decades.

Re: Stadia died because no one trusts Google

#566

Post by solardev »

chrisseaton wrote: post #563
solardev wrote: post #558

Earlier quoted context omitted.

Somehow Nvidia, Shadow, PSNow, Xbox Cloud, Luna, Paper, etc. manage to do this...

Manage to do what? Use edge locations? I'm not sure they do - where did you read that?

Not sure about on the edge, but the problem of gpu allocation isn't a dealbreaker. All the other steaming companies do it and have yet to go bankrupt. Heck, OnLive was doing it ten years ago on much worse hardware and internet.

Re: Stadia died because no one trusts Google

#567

Post by impossiblefork »

geertj wrote: post #410
impossiblefork wrote: post #309

Earlier quoted context omitted.

The Gigapress is an Italian innovation though, and the Italians built it because they understood that it could be incredibly useful, not because there was a custom order. Tesla were early adopters by making use of them, but not innovators.

> Tesla were early adopters by making use of them, but not innovators. There are two versions of the Gigapress history going around. In one version the idea came to the Idra CEO in a dream, while in the other version Elon gave multiple die casting manufacturers a call and got a hard no from everyone, expect for Idra who gave him a ‘maybe’. Whichever it it is not that important in my view. What is undisputed is that T…

As I understand it Idra already had the idea for large presses and understood their potential. The demands Tesla had may still have been difficult to realize. Consequently the stories are in some sense reconcilable.

Structural die-casting has been a trend in the automotive industry, having started before the Gigapress. One article on this, from 2018, is https://www.spotlightmetal.com/opportunities-for-die-casting... and in that you see an analysis of the economic benefits, even though it's an article for public consumption.

The article itself, back in 2018, mentions cars already using structural die-casting and basically forecasts structural die-casting becoming a more typical manufacturing method.

Re: Stadia died because no one trusts Google

#568

Post by shmerl »

solardev wrote: post #559
shmerl wrote: post #553

Earlier quoted context omitted.

You don't need upscaling for a streaming gaming service. DLSS is pointless for it because you can't possibly have framerates that would force you to upscale anyway.

What do you mean?

Upscaling is useful if for whatever reason you need a target framerate that your GPU can't sustain at your desired resolution. I.e. roughly, you reduce target resolution which allows GPU to render more frames in the same time period, then you upscale the result to counteract that resolution reduction. That way you trade image quality for higher framerate.

In the network case (i.e. "streaming"), you are limited by the bandwidth and latency of the network itself. I.e. you will sooner bottleneck by the network than by GPU being unable to render frames faster than that. So upscaling is just pointless for it.

Re: Stadia died because no one trusts Google

#569

Post by solardev »

shmerl wrote: post #568
solardev wrote: post #559

Earlier quoted context omitted.

What do you mean?

Upscaling is useful if for whatever reason you need a target framerate that your GPU can't sustain at your desired resolution. I.e. roughly, you reduce target resolution which allows GPU to render more frames in the same time period, then you upscale the result to counteract that resolution reduction. That way you trade image quality for higher framerate. In the network case (i.e. "streaming"), you are limited by the…

I don't think that's true though. It's not very bandwidth demanding to stream 4k 60fps or 1440p 120fps, GeForce Now does that already out of the box for less than 100 Mbps.

But it's hard to sustain those frame rates at max graphics. Ue5 will raise the ceiling quite a bit too. The gfx cards will continue to be a bottleneck for AAA games.

Re: Stadia died because no one trusts Google

#570

Post by shmerl »

solardev wrote: post #569
shmerl wrote: post #568

Earlier quoted context omitted.

Upscaling is useful if for whatever reason you need a target framerate that your GPU can't sustain at your desired resolution. I.e. roughly, you reduce target resolution which allows GPU to render more frames in the same time period, then you upscale the result to counteract that resolution reduction. That way you trade image quality for higher framerate. In the network case (i.e. "streaming"), you are limited by the…

I don't think that's true though. It's not very bandwidth demanding to stream 4k 60fps or 1440p 120fps, GeForce Now does that already out of the box for less than 100 Mbps. But it's hard to sustain those frame rates at max graphics. Ue5 will raise the ceiling quite a bit too. The gfx cards will continue to be a bottleneck for AAA games.

Upscaling is more useful if you need some crazy framerates (like 240 fps) with high graphics settings for some reason. Even most high end GPUs can't handle that for demanding games on common resolutions. Or more reasonably for mid range and lower end GPUs which are just weaker and can't handle even 144 fps for the above without upscaling.

But low end GPUs use case isn't relevant for streaming since they can have highest end ones. So we are talking about those who want to push those crazy framerates, and I don't think networks allow that even if GPUs can sustain them.

Some middle case can happen I suppose if you get an outlier super demanding game, but I always saw it as questionable to push max settings if you are ready to trade image quality for more framerate with upscaling anyway. I'd rather stick to lower framerate and keep image quality of max settings intact by avoiding upscaling. So I just don't see a value in having it, let alone in dedicated hardware vs something like FSR which can work on regular compute units.

Post reply on HN