Microsoft Teases Next-Gen Xbox: “Largest Technical Leap in a Hardware Generation”

Tsing

The FPS Review
Staff member
Joined
May 6, 2019
Messages
11,391
Points
83
Xbox isn't planning on leaving the console business any time soon, according to Xbox President Sarah Bond, who revealed during yesterday's episode of the Official Xbox Podcast that Microsoft is not only working on a next-generation console, but that it's being designed to deliver what will apparently be the biggest performance leap seen in an Xbox console ever.

See full article...
 
32 gigs of ram. With 16 allocated to video memory! (just a guess)
 
Well if they go Nvidia they may get a big leap in raytracing performance otherwise I don't see what AMD can come up with to make a big enough difference to warrant that statement.

Also if they do another Series S they will ruin another gen like the current one
 
Yea they need to just release a full fat, all in console to start. Though I'm betting on more storage and no optical drive.
 
Anything is possible if you are willing to accept it sounding like a vacuum cleaner, needing a dedicated outlet, and/or heating the room up several degrees.
 
Anything is possible if you are willing to accept it sounding like a vacuum cleaner, needing a dedicated outlet, and/or heating the room up several degrees.
Even a return to Intel... or so the Reddit rumors say is under consideration
 
Even a return to Intel... or so the Reddit rumors say is under consideration
Sure it's possible. I don't know how it helps out - Consoles are such an odd niche. It would be a huge win for Intel and Xess... but I don't see it really making the console any better than AMD (or nVidia, or anyone else) could really do. It's all about the power envelope and what you're willing to deal with on the hardware side - so you pick someone you think will help you drive down costs over time.

No console has ever sold well just because it has great hardware. It's all about the system as a whole. You can have the best hardware possible, but if the ecosystem sucks no one will care. To steal a line from James Carville: It's all about the games, stupid.
 
It is indeed all about the games. And if you want good games you make the platform easy to program on.

Suddenly you have good games coming to it!

My thought is if they can get some BIG and I mean BIG IPC improvements for the console processors. Then they can keep the power envelope around where it is and drive more performance. The issue here is nobody as of yet is claiming even double digit IPC improvements going into the next generation... and honestly to get consoles into 4k greater than 60 FPS gaming... you need some BIG improvements.

OR... you need a feature that runs well that makes games SO MUCH more immersive. Like I don't know... AI's that have a local LLM that has conversation objectives and unlocks that are capable of understanding and responding to human speech.

That would take a lot of juice for a console to run well... but some AI accelerators and a big chunk of memory... (8 gig just for the LLM) Might do it...
 
It would be a huge win for Intel and Xess... but I don't see it really making the console any better than AMD (or nVidia, or anyone else) could really do.
Intel seems to be very serious about RT in their shipping hardware, while AMD hasn't been, and Nvidia lacks the CPU IP to compete (and they'd most likely have to use ARM). One has to wonder if Intel wasn't pushing their Xe graphics for this very purpose, even if just aspirationally.

But the disadvantages are also real, can't leave those out - Intel's cores while likely to be more performant on an IPC basis would also likely be less efficient than a solution from AMD. Intel could use their P+E arrangement while AMD could leverage their Zen4 + Zen4c arrangement, and I'd be hard pressed to predict which would be better for consoles.

XeSS might also be better, but FSR has also shown room for improvement too, and I think either could be made to work as well as needed for the upcoming console generation. Honestly I expect significant development to be focused here as Nvidia has shown that this family of technologies can more or less just be used everywhere if well implemented.
 
My thought is if they can get some BIG and I mean BIG IPC improvements for the console processors. Then they can keep the power envelope around where it is and drive more performance. The issue here is nobody as of yet is claiming even double digit IPC improvements going into the next generation... and honestly to get consoles into 4k greater than 60 FPS gaming... you need some BIG improvements.
This I don't really see the need for.

For example, the 7800X3D is between a 12600K and 12700K in multi-core and below the 12600K in single-core Cinebench (raw FP compute), yet tops the charts when it comes to 1.0% and 0.1% lows while remaining highly competitive in the averages when it comes to game performance.

Point being, it's the balance of resources on the CPU that matters to games more than raw core performance. So, if we assume Zen4 with say 64MB of L3 cache and eight cores that can hit 5.0GHz, we'd have a world-beating gaming CPU that'd still be pretty efficient.

As for the AI stuff, I honestly can't really speculate on that. There are many, many ways for that to shake out, but most are likely to be phone-centric from the perspective of the average consumer. I can understand putting hardware on desktops, laptops, and tablets too, but consoles have been strictly limited to gaming. Do we think Microsoft is willing to disrupt that paradigm?
 
As for the AI stuff, I honestly can't really speculate on that. There are many, many ways for that to shake out, but most are likely to be phone-centric from the perspective of the average consumer. I can understand putting hardware on desktops, laptops, and tablets too, but consoles have been strictly limited to gaming. Do we think Microsoft is willing to disrupt that paradigm?
Do we really think a modern CPU going forward won't have AI accelerators or processors built into them? The question is if MS or Sony wants a SOC without that or says... you know what give it to us as a way to push gaming and our consoles forward?

Let me find a link for you that's kind of interesting...

I don't know about you... but having a reasonably quick AI running on a local LLM that's focused on the game... maybe with tiered unlocks and some preprogramed sequences where needed could be VERY awesome. At least to me. I think the level of immersion would be... very interesting.
 
I absolutely agree with you that they'll put the hardware in. It's going into everything, and it would be a mistake not to; not because there's a guaranteed use, but rather because it's cheap to get it out there now and it's not something that anyone wants to fall behind in.

That said, my pessimism is rooted in the 'what for?' question. I'll admit that I'm among those that doesn't like to talk to my electronics / computing devices, so I'm jaded going in - but really, how does AI improve the user experience without also making it horrifically inconsistent, and wouldn't that run counter to the experience that game developers typically shoot for?

I get that LLMs can be useful at datacenter-scale implementations. What can a local implementation on a console do? My assumption is that any real utility will be using local hardware to handle user interaction and having server-side resources handling anything complex / deep.



On 'AI', what I'm seeing is the effort to make 'use' of information pools. I think Microsoft's 'Copilot' branding covers what I'm expecting, that is, an assistant that can both find things as well as help with complicated work. I also expect that such implementations will be highly domain-specific should the technology become actually useful in a commercial / professional sense.
 
@LazyGamer I'm working on professionally becoming more expert at what AI can do for a company or in this case for a game. The larger the datasets the more information an AI or lets call it a LLM can pull from. It all depends on what/how the AI data was populated.

So some examples. You have a data pool of ALL of the information on the game world. Who everyone is and everything else. Maybe alignments and such as part of it. All of the fauna and animals and such as well as every building and everything else in the game. Lets say this is... 4 gig of data. That is a small data pool. On top of that you have an LLM that understands angry, sad and other such states as well as good, bad, neutral and gender identities and even what different types of people may sound like on a sliding scale. All information for the character definitions.

Then instead of having a character with a set of generic responses to pre programmed requests for data. (pick one of 4 ways to ask about such and such how all games do it today IF they even give you an option mind you.) NOW you can have an AI overlay where you describe the type of person the NPC is, what they sound like, and what areas about the world they known and how well they know it. Maybe even assign generic values to the data set for how known specific data is.

So now instead of preprogrammed responses you have an NPC that has a breadth of knowledge about many things to a shallow level and a lot of knowledge about other more specific areas of focus within the game world.

Now when you meet someone in the game based on those values they can respond in a 'unique' voice about things they may know about depending on their attitude toward you.

AND you can run all of that locally without requiring an internet connection.

THAT is where AI gets really cool and flexible. Sure it'll take more hardware than your xbox has today to do all of that. AND run the game well. But it's not far off from being something that can be done. The less data the LLM has to respond to, and the MORE accurate the responses have to be the better the experience. You can even allow AI to 'guess' at results. (today what we call AI Hallucinations) or make logical jumps with you. AND it can be wrong!

For gaming that will be alright and just add to the depth of NPC interactions even if frustrating.

for professional uses... they need to get rid of the hallucinations. ;)
 
So now instead of preprogrammed responses you have an NPC that has a breadth of knowledge about many things to a shallow level and a lot of knowledge about other more specific areas of focus within the game world.

Now when you meet someone in the game based on those values they can respond in a 'unique' voice about things they may know about depending on their attitude toward you.
I realy do not see a point in that, I'm not going to talk to the hundreds or thousands of NPC's roaming around in a game to find out eveything they might know about a certain subject and it would make finding NPC's you need to talk to to advance the game a pain.

To me it sounds like bloatware and would require a ton of work for just some background info.

Maybe if you could build a game to create something lto make random episodes for a CSI like game where each playtrough would be different but still solvable.
 
I realy do not see a point in that, I'm not going to talk to the hundreds or thousands of NPC's roaming around in a game to find out eveything they might know about a certain subject and it would make finding NPC's you need to talk to to advance the game a pain.
I think we're missing each others thought process here. I see it as you still get pointers to talk to specific NPC's about information you want to know. And they would have access to that information to share. BUT they might also know other stuff about the world. And anyone you talk to that is 'named' would have areas of expertise perhaps more unlocked as you progress in the game world. Even OTHER NPC's that are known to know about specific information.

Yes it would make single player games less linear but I don't see that as a bad thing.
 
Become a Patron!
Back
Top