Next Generation Hardware Speculation with a Technical Spin [2018]

[Shifty Geezer]

Yes, but that's besides the point unless we think that means the console can be Itaniums and 64 GBs RAM. What would the ideal hardware for an Azure server be? That's got to be where MS's focus is as that's where their money is and they definitely won't compromise that service to their cloud renters. Also, what does AMD have targeting servers? I thought they were all going low-power ARM these days, so where does Zen fit in?

 

[AlphaWolf]

What do you mean? The One was sold at 500 while the PS4 was sold at 400. After removing the camera to compete in price, the One couldnt compete with performance.

Well they have been the same price for 5 years. 1x and pro are the same price. PsPro still manages to sell.

 

[Jay]

I believe it's supposed to be console first, server second
The main priority of the SoC going into the data centre is to serve streaming.

Disagree, I believe they design it with use cases in mind.
They knew it was going into data centres, and not just for streaming either.Console is just one of the main considerations.
That doesn't mean you end up with a compromised apu for console either.
You're more likely to end up with comprised apu if you design it with just console (first), especially for the other uses they plan to put them to.
Console can be subset of features, what does the console need that would cripple it for uses in other places that couldn't be resolved if was part of the initial design on how to handle it?

The other reason I doubt they are just repurposing console apu's and putting them in the cloud for streaming is, MS has made it a point that gaming and it's tech has a much broader use that it can be put to, hence it's become one of their pillars. It's why gaming is getting so much investment

 

[Jay]

Yes, but that's besides the point unless we think that means the console can be Itaniums and 64 GBs RAM. What would the ideal hardware for an Azure server be? That's got to be where MS's focus is as that's where their money is and they definitely won't compromise that service to their cloud renters. Also, what does AMD have targeting servers? I thought they were all going low-power ARM these days, so where does Zen fit in?

I think this can be a real focus of the discussion in regards to MS console/apu design considerations.
And what it would mean with a monolithic type chip compared to mcm etc

 

[mrcorbo]

Best case, the hardware designed for Azure servers is also ideal for consoles. If true, it needn't be specifically designed for servers - the perfect console design would naturally be a perfect fit for servers. Worst case, in designing a box that works ideally as a server box, it doesn't work well as a console box. Like designing a car base designed to be both an SUV and a sports car given a different shell.

I'm not saying that's a perfect analogy, as computing is different and it's possible that the computing structures used for servers are good for consoles too. I'm also not saying it's bad business for MS. The quality of the end results will show whether it's a comprimise for console owners who left playing second fiddle to the money mainstream of cloud services, or if MS do something clever and manage to bring a synergy across platforms that works to console owners' advantage.

The important point here in this hardware predicition thread is we have MS telling us they are considering Scarlet's HW use in servers as well as consoles. By considering what a server needs vresus a console, we can shift our expectations of Scarlet. Does that point to more CPU power than usual for a console? Is CPU totally out of fashion in servers? What about RAM and BW, typically significant components in servers - will Scarlet support wider busses and more RAM in different configurations to the console? Could that point to multiple hardware configs?

There's really plenty here to discuss of hardware merit instead of talking about business strategies and whether console gamers are being sidelined or not.

There's not how I read it, though. It's don't think it's designed for Azure. It will instead be designed for Project xCloud as well as Scarlet and the Project xCloud servers' primary design criteria will still be to run games well. However, since Project X cloud servers will be co-located with other types of Azure servers, when not being used for game instances, those CPU and GPU compute resources will be available to service Azure clients as well.

 

[mrcorbo]

To expand on my previous post, the one Azure workload that is called out that the Scarlet/xCloud SoC will be tasked with helping out with is machine learning. AMD have products that they market as being suitable for this workload and they are just variants of their GPUs.

It was never stated that this chip was intended to be the only type of chip that will appear in Azure servers going forward, and would therefore have to be able to handle all cloud server workloads well. It was only stated that it would be tasked with handling machine learning which is a workload that could be handled by the CUs in the GPU that is going to be in the SoC anyway.

 

[MrFox]

The nodes from cloud providers have a very large range of specs. It depends on the clients, the demand for compute on AWS and google from companies like mine is a very large number of dual xeons with many nvidia cards, and at least 128GB ram. The other end of the spectrum is those who only need more network nodes to dynamically grow a simple web service based on demand, micro cloud with a very very large number of Atom style chip (or arm) per U is the most efficient there. Another trend is massive nodes which can be easily virtualised into smaller nodes on a need basis to serve more workload types.

So a console type chip in the racks can hit some good price point for middle ground usage where they don't need much ram and have a compute workload that can be easily distributed. Not having to buy overpriced nvidia cards is a big plus!

Saying they designed the silicon for the cloud is a bit weird. It's a balance of cpu, gpu, and memory, and it has to be designed for the console, the cloud will get whatever they ended up with and fit some sort of use case in the cloud.

But... I don't believe a word spencer says anymore, it's a crap shot whether that's their usual PR crap, in line with: xb1x will support high end wireless VR, each xb1 gamers will get an additional 3x xbox power from the cloud, xb1 silicon was designed for kinect/online and cannot be removed, nobody can see the difference between 900p and 1080p on a 60" TV, etc... And their use of "exclusive" versus "exclusive exclusive" was pretty much the reason we have to deconstruct everything they say to figure out where the misdirection is. It was never a "message" problem, it still looks like the PR side saying anything that sells, and not even talking to engineers. We can't use what spencer says to deduce anything about the hardware, it's a waste of time.

 

[mrcorbo]

The nodes from cloud providers have a very large range of specs. It depends on the clients, the demand for compute on AWS and google from companies like mine is a very large number of dual xeons with many nvidia cards, and at least 128GB ram. The other end of the spectrum is those who only need more network nodes to dynamically grow a simple web service based on demand, micro cloud with a very very large number of Atom style chip (or arm) per U is the most efficient there. Another trend is massive nodes which can be easily virtualised into smaller nodes on a need basis to serve more workload types.

So a console type chip in the racks can hit some good price point for middle ground usage where they don't need much ram and have a compute workload that can be easily distributed. Not having to buy overpriced nvidia cards is a big plus!

That was my understanding. Thanks for the insight.

Saying they designed the silicon for the cloud is a bit weird. It's a balance of cpu, gpu, and memory, and it has to be designed for the console, the cloud will get whatever they ended up with and fit some sort of use case in the cloud.

Bringing the actual quote forward, as opposed to people's interpretation of it, it would seem that's exactly what he said.

The thing that’s interesting for us as we roll forward, is we’re actually designing our next-gen silicon in such a way that it works great for playing games in the cloud, and also works very well for machine learning and other non-entertainment workloads. As a company like Microsoft, we can dual-purpose the silicon that we’re putting in.

We have a consumer use for that silicon, and we have enterprise use for those blades as well. It all in our space around driving down the cost to serve. Your cost to serve is made up by two things, how much was the hardware, and how much time does that hardware monetize.

So if we can monetize that hardware over more cycles in the 24 hours through game streaming and other things that need CPU and GPU in the cloud, we will drive down the cost to serve in our services. So the design as we move forward is done hand-in-hand with the Azure silicon team, and I think that creates a real competitive advantage.”

 

[MrFox]

Hahaha ok, yeah I should read the quote instead. still if they specifically designed the silicon in such a way it's good for both uses, it implies there's something to change, it implies a normal console design isn't good for the cloud, it's not designed "in such a way". What is cloud-specific about the silicon design itself, versus just desiging the best console chip and using it for cloud too?

I was thinking maybe the codec aspect is beefed up, that would help the cloud by reducing latency (like an overkill compresion speed), and the additional silicon cost wouldn't be a big one. Networking isn't a big deal. Not sure what else would make them change the natural design of a console to make it "good for both".

 

[mrcorbo]

Hahaha ok, yeah I should read the quote instead. still if they specifically designed the silicon in such a way it's good for both uses, it implies there's something to change, it implies a normal console design isn't good for the cloud, it's not designed "in such a way". What is cloud-specific about the silicon design itself, versus just desiging the best console chip and using it for cloud too?

I was thinking maybe the codec aspect is beefed up, that would help the cloud by reducing latency (like an overkill compresion speed), and the additional silicon cost wouldn't be a big one. Networking isn't a big deal. Not sure what else would make them change the natural design of a console to make it "good for both".

Even if it wasn't being done as a primary intention, the design could still be being designed "in such a way" that the result is that it is suitable for multiple uses.

Ultimately, the way people should parse this should be based on what makes the most sense. Maybe someone should tweet him and ask for clarification.

Edit: I do think that the Azure Team will have input on the design, but I don't think that they will be equal partners. If a design decision can be made that makes the design work better in the datacenter it could be implemented, but not at the expense of harming its effectiveness in the console.

 

[Nesh]

Well they have been the same price for 5 years. 1x and pro are the same price. PsPro still manages to sell.

I don't think you are following.
The base models are still not like for like. MS original focus did have an effect on the base model's performance and that does not sell as well as the PS4. The X and Pro are limited stock consoles and obviously the Pro is enjoying sales due to the success of the base model.
The argument that sometimes focus or non-gaming features can impact negatively the gaming aspect of the console still stands

 

[vipa899]

Perhaps MS learnt from their misstakes for next gen regarding that?

 

[bgroovy]

Up-front price and performance can be preserved, since the goal is to have service subscriptions, transactions like in-game purchases and lootboxes, elimination of the relevance of the secondary market (can't sell the service), monitoring of user habits, preferences, and consumption can compensate. Gamers can be pushed more quickly to games publishers want, and game experiences can be tailored to get them to spend more money.

You are making a lot of business arguments, with very little to improve the consumer perspective. Like I've said, MS is always focused of business models and focusing on leveraging synergies, often while ignoring the impact on the consumer.

They were/are the same price.

And the Xbox One hardware is demonstrably worse for the same money. That's my whole point.

I never said the ps4 doesn't have it. I said its the only one that does. My switch doesn't do that.

You literally just got finished saying the Switch has a button for quickly taking screenshots or saving video clips. That's what a Share button does. PS4 invented it, Nintendo adopted it. Xbox had to put together a voice command and quick button command to respond to it. The next Xbox will have one.

 

[AlphaWolf]

I don't think you are following.
The base models are still not like for like. MS original focus did have an effect on the base model's performance and that does not sell as well as the PS4. The X and Pro are limited stock consoles and obviously the Pro is enjoying sales due to the success of the base model.
The argument that sometimes focus or non-gaming features can impact negatively the gaming aspect of the console still stands

And sometimes those features are what sell the product, Nintendo would have folded up shop years ago if it wasn't. MS was trying to go there but missed it and they marketed themselves into the ground before launch, a better (more powerful) product still probably would have sold worse.

And I don't think I would use the term limited for the ps4pro or the 1x. They aren't commanding a premium over msrp. Without sales numbers for the sku we can't really tell where they are at or how much the additional power might be helping the 1x.

Edit: Just want to add that it's going to be difficult to make inroads on marketshare if they're only differentiating themselves from the competition by a small amount of performance. Everyone is working in the same reality, similar thermal/power envelope, similar price point means there is not going to be a huge difference in performance.
No one is going to ship a loss leader anymore, the tech just isn't advancing fast enough to recover anymore.
So being innovative is a chance to gain without giving up too much elsewhere. I realize that might not be what the core gamer crowd wants, but the goal extends well beyond the core gaming audience, if it didn't they would probably just pass, because they aren't worth the effort on their own

 

[eastmen]

You are making a lot of business arguments, with very little to improve the consumer perspective. Like I've said, MS is always focused of business models and focusing on leveraging synergies, often while ignoring the impact on the consumer.



And the Xbox One hardware is demonstrably worse for the same money. That's my whole point.



You literally just got finished saying the Switch has a button for quickly taking screenshots or saving video clips. That's what a Share button does. PS4 invented it, Nintendo adopted it. Xbox had to put together a voice command and quick button command to respond to it. The next Xbox will have one.

So it doesn't actually share anything then ? Its a picture button ? I've had that on my pc forever now.

Whats more is even if you want to give sony credit for that its on one other device. Far from industry standard

 

[Karamazov]

when you press the button you can directly share the video saved to youtube or facebook or other sharing sites.

 

[vipa899]

Xfire had it since 2006, it shutdown some years ago. The idea could come from anywhere.

 

[Xbat]

So it doesn't actually share anything then ? Its a picture button ? I've had that on my pc forever now.

When it comes to console with only so many buttons on a controller it's quite the innovation. Considering how big streaming has become Sony was forward thinking with PS4 when it came to the social aspect.

 

[Nesh]

And sometimes those features are what sell the product, Nintendo would have folded up shop years ago if it wasn't. MS was trying to go there but missed it and they marketed themselves into the ground before launch, a better (more powerful) product still probably would have sold worse.

What non gaming features are setting Nintendo's console apart? What makes the XBOX that competes the PS4 on the same experience and same type of games, so similar to a Switch, Wii or WiiU?

 

[Tkumpathenurple]

So it doesn't actually share anything then ? Its a picture button ? I've had that on my pc forever now.

Whats more is even if you want to give sony credit for that its on one other device. Far from industry standard

Would you stop moving the bloody goalposts?

You were wrong about something completely trivial and unimportant, accept it. It happens to us all.