Next Generation Hardware Speculation with a Technical Spin [2018]

[Syferz]

It makes even less sense for MS to adopt ARM given BC with XB1 and PC cross compatibility. Not saying it won't happen, but it's the least likely one to change.

Well they are software emulating 360 games on XB1 with those netbook cpus, and those were 3.2ghz PowerPC on 360. Moving to a higher clocked ARM CPU shouldn't pose a problem tbh. Also they are emulating windows with ARM as well, so I don't see why they are least likely...

 

[phoenix_chipset]

ARM Like all other Nintendo handhelds?

When we first learned of nx we didn't know it was a hybrid (so, a handheld), We just knew Nintendo wanted to merge their console and handheld ecosystems. Meaning we could've gotten seperate handheld and console that shared a base architecture.

And continuing the power pc 750 line wasn't an option obviously so it had to be arm, anyway.

Don't think Sony or ms go for arm. That's an unnecessary change and it'd axe any chance of bc. DEFINITELY not ms, but I suppose Sony could since they don't seem to give a shit about bc. Easier to sell you the same games again digitally.

 

[function]

Well they are software emulating 360 games on XB1 with those netbook cpus, and those were 3.2ghz PowerPC on 360. Moving to a higher clocked ARM CPU shouldn't pose a problem tbh. Also they are emulating windows with ARM as well, so I don't see why they are least likely...

The PowerPC cores on 360 had incredibly low IPC, while the Jaguar cores are modern out-of-order cores with several times the IPC. Despite this, BC has been a huge engineering effort that's taken years and still occasionally slows on the base X1 CPU.

I wouldn't assume that full speed emulation of Jaguar - with the added complication of using not one but two separate Jaguar modules - will be automatic or easy.

If you want full speed emulation of X1X then you have a much faster core to emulate (2.3 gHz) with several percent worth of IPC improvements.

 

[Syferz]

The PowerPC cores on 360 had incredibly low IPC, while the Jaguar cores are modern out-of-order cores with several times the IPC. Despite this, BC has been a huge engineering effort that's taken years and still occasionally slows on the base X1 CPU.

I wouldn't assume that full speed emulation of Jaguar - with the added complication of using not one but two separate Jaguar modules - will be automatic or easy.

If you want full speed emulation of X1X then you have a much faster core to emulate (2.3 gHz) with several percent worth of IPC improvements.

I don't think they would do it, I'm just saying out of the two companies, they are the ones doing software emulation and showing a desire to be cpu agnostic from a software standpoint. I'd personally guess there is a 10% chance they do it. As for ARM IPC, modern ARM cores are much faster than jaguar cores AFAIK, and the jaguar core clock is only 1.75GHz, in a console with ARM cores, you could double that clock speed.

 

[BRiT]

I don't think they would do it, I'm just saying out of the two companies, they are the ones doing software emulation and showing a desire to be cpu agnostic from a software standpoint. I'd personally guess there is a 10% chance they do it. As for ARM IPC, modern ARM cores are much faster than jaguar cores AFAIK, and the jaguar core clock is only 1.75GHz, in a console with ARM cores, you could double that clock speed.

No. The current Jaguar core clock is 2.3 Ghz.

 

[Syferz]

No. The current Jaguar core clock is 2.3 Ghz.

That isn't what they would need to emulate xb1 games.

 

[BRiT]

That isn't what they would need to emulate xb1 games.

Correct, they would need far more than that. The current Jaguar cores are run at 2.3Ghz in current XBOX console hardware.

 

[MrFox]

Software emulation causes a massive hit in performance so you'd need an ARM much faster than jaguar just to get a similar BC performance from 2013 consoles, let alone at least being on par with mid-gen. Now imagine if the competition uses a modern AMD x86, it would emulate all games at 60fps at launch.

There are no desktop-level ARM implementations so far. Why develop a brand new processor based on ARM, and cause a world of pain from software emulation, when AMD is offering zen(2,3) on a silver platter?

 

[Syferz]

Software emulation causes a massive hit in performance so you'd need an ARM much faster than jaguar just to get a similar BC performance from 2013 consoles, let alone at least being on par with mid-gen. Now imagine if the competition uses a modern AMD x86, it would emulate all games at 60fps at launch.

There are no desktop-level ARM implementations so far. Why develop a brand new processor based on ARM, and cause a world of pain from software emulation, when AMD is offering zen(2,3) on a silver platter?

They wouldn't, I don't believe they would anyways.

 

[iroboto]

Richard puts his hat into this thread with a 2019 release lol.
http://www.eurogamer.net/articles/digitalfoundry-2018-ps4-pro-and-xbox-one-x-processors-compared

n the here and now, we can safely assume that both Sony and Microsoft are moving on to the core design of their next-gen successors, which become viable from a manufacturing perspective in 2019, when next-gen 7nm processor fabrication is expected to be mature enough to accommodate the demands of a new console launch. And with the arrival of AMD's Ryzen/Vega-based APUs, launching today (with Digital Foundry review coverage to follow), we may well be getting our first look at the building blocks of tomorrow's PlayStation and Xbox. Based on a revised 14nmFF process, AMD has integrated a quad-core Ryzen block and 11 Vega compute units in an area approximately 210mm2. Scaling down those components to what we expect from the new 7nm process on a similar area used by PS4 Pro and Xbox One X offers up some fascinating options for the platform holders - something we'll be returning to soon.

 

[MrFox]

Richard puts his hat into this thread with a 2019 release lol.
http://www.eurogamer.net/articles/digitalfoundry-2018-ps4-pro-and-xbox-one-x-processors-compared

Not exactly predicting it, but he's saying 2019 is where it becomes a possible target because of 7nm, based on the jump we got mid-gen (28 to 14) which is what this article is all about. Still a bit misleading because 7nm is only 1.5x scaling until they do euv in high volume. I suppose he's already planning an article about this once we start getting more data from 7nm gpus. ""something we'll be returning to soon.""

 

[iroboto]

Not exactly predicting it, but he's saying 2019 is where it becomes a possible target because of 7nm, based on the jump we got mid-gen (28 to 14) which is what this article is all about. Still a bit misleading because 7nm is only 1.5x scaling until they do euv in high volume. I suppose he's already planning an article about this once we start getting more data from 7nm gpus. ""something we'll be returning to soon.""

fair enough!

 

[Grall]

Software emulation causes a massive hit in performance

Meh. You'd recompile the binary for your new target platform, and/or rely on JIT transcoding, and then saving the resulting transcoded binary for further runs. Software emulating a modern ~2GHz CPU in a traditional manner would be massively demanding, practically impossible for all intents and purposes.

 

[borntosoul]

Performance per watt and performance per $ will be the two biggest things on Sony’s mind. all this vs their performance target/vision. It’s not unreasonable to expect a bit over 2x ps4 pro speeds, release in 2020 or by Christmas 2019 would make sence.
I’m sure Sony have the cpu and gpu designs that they’ll use already sorted.

 

[OCASM]

I wish they targeted 1080p instead of wasting so much power on 4k. Oh well...

 

[ultragpu]

1440p or 1800p then CBR or frame inject to 4k from that point on would be pretty ideal for a 10 tf console. Again they might have to anyway since a 9-10 tf machine would not have nearly enough power to pull next gen graphics on top of 4k, 15 tf is the bare minimum to do that I reckon.

 

[borntosoul]

15 tf would great if they could pull it off. I doubt it though, not before 2022. Hope I’m wrong.


Edit, changed 2012 to 2022.

 

[Tkumpathenurpahl]

What kind of performance is required to achieve certain effects and techniques?

If we take it that 4*1.84=7.36TF would be enough to brute force 4K on every 1080p PS4 game, what additional power would be required to achieve the likes of Global Illumination?

And would the same kind of >4x increase in bandwidth be ideal?

 

[OCASM]

What kind of performance is required to achieve certain effects and techniques?

If we take it that 4*1.84=7.36TF would be enough to brute force 4K on every 1080p PS4 game, what additional power would be required to achieve the likes of Global Illumination?

And would the same kind of >4x increase in bandwidth be ideal?

Such a simple calculation is not possible. Most effects are done per-pixel and so the power needed depends on the shader complexity. All we can know is that more pixels = more power required hence why I prefer sacrificing resolution over pixel quality.

 

[Globalisateur]

About the CPU I think the only question now is to choose between: 4c4t, 4c8t, 8c8t or 8c16t. I think it's going to be 8c8t, 2 of those (the updated 2019 or 2020 model at 7nm):

AMD Ryzen™ 3 2300U
# of CPU Cores 4
# of Threads 4
Base Clock 2GHz
Default TDP / TDP 15W
14nm


https://www.amd.com/fr/products/apu/amd-ryzen-3-2300u