Would AMD's K12 core (2016+) be a candidate for next generation console APUs ?

[Megadrive1988]

AMD's next generation CPU core, K12, isn't due for about another two or so years, 2016+

http://techreport.com/review/26418/amd-reveals-k12-new-arm-and-x86-cores-are-coming
http://wccftech.com/amd-announces-20142016-roadmap-20nm-project-skybridge-k12-64bit-arm-cores-2016/

K12 is meant to support both ARM and x86

Assuming a next generation Xbox and PS5 are going to be slated for ~2018, wouldn't a semi-custom APU with at least 8 K12 x86 cores make sense for next generation APUs / SOCs?

Even if they used 8 cores, one would think whatever K12 x86 architecture is, it would be a major improvement in capability over the Jaguar cores used in PS4/XB1, not to mention going to a much higher clockspeed (over 1.6 ~ 1.75 GHz ----> going to 3~4 GHz).

When coupled with at least a 10~12 TFLOP GPU and some stacked DRAM (TSV), would make for a very nice console. Chip manufacturing should be something like 14 nm FinFET or the like.

Games could have more complex scenes, lighting, physics, etc at either 4K 30 FPS or 1080 60FPS (upto the developer on a per game basis, as always) and decent enough performance for VR (compared to how well PS4 will drive Project Morpheus).

 

[Rangers]

Sure

They might well go with ARM though. They really wanted to for some business reason this last time, it just wasn't quite ready yet. My guess is ARM in some form (perhaps from AMD) is the heavy favorite.

Also I hope for 2018, but it sounds aggressive (5 years).

 

[function]

Assuming there's still a perceived market for "high end" consoles in 2018, it'd seem like a good fit.

Who else is going to have genuinely high performance (not phone level) CPU cores that are ripe for integration into a SoC with high end graphics?

I'm really rooting for AMD. It's been a painful few years. Watching Bulldozer wheeze along on 32nm with AMD trying to find *somewhere* that it's competetive(ish) is pretty sad. Kaveri almost managed to make good despite its heritage.

Puma looks the business though. A genuinely competitive 4.8W part that works despite only having a single DDR3 channel. And this time it'll clock up to 2.4 gHz under turbo!

It's time to get on the AMD rollercoaster of hope and despair again, folks, because K12 is a commin'!

 

[3dilettante]

K12 is meant to support both ARM and x86

Most of what has been said by AMD has been more consistent with K12 being AMD's custom ARM core, with a generally equivalent x86 sibling of undisclosed designation.

Assuming a next generation Xbox and PS5 are going to be slated for ~2018, wouldn't a semi-custom APU with at least 8 K12 x86 cores make sense for next generation APUs / SOCs?

There'd be nothing wrong with it, I suppose.

Games could have more complex scenes, lighting, physics, etc at either 4K 30 FPS or 1080 60FPS (upto the developer on a per game basis, as always) and decent enough performance for VR (compared to how well PS4 will drive Project Morpheus).

The more important factors there are probably not the CPU cores.
By 2018 there should be a decent number of architectures out there capable of beating Jaguar.
Right now, Jaguar actually beats Jaguar, since Puma has a finally bug-fixed power management and turbo function that the consoles missed out on.

 

[Megadrive1988]

And when I say 2018 (5 years after PS4/X1 release) I meant to say 2018 or 2019. The 5 or 6 year timeframe.

Now if Microsoft and Sony go with AMD K12, would it not make a whole lot of sense to use an x86 version rather than ARM, for backward compatibility and just programmer familiarity?

I think we could totally see Nintendo going with ARM cores as it makes much more sense, a console and handheld with the same architecture (just different performance levels).


Edit: I understand well enough things like higher scene complexity (geometry, etc) lighting, shaders, effects, etc would be handled by a more powerful GPU. Yet at the same time there is also little doubt that any future Xbox and PS5 would assuredly get a newer, better CPU with more capabilities and a higher clock, just given the timeframe.

K12 x86 makes the most sense, at least at this early point in time (summer 2014) barring any change of chip vendor, or something else on AMD's roadmap they haven't disclosed publicly. Who knows what else is coming CPU wise from AMD, but consoles take several years to R&D. 2016 would seem the time to really begin full-scale development of the CPU/GPU (APU) for the next generation consoles.

I'm hoping it's not the Puma core or any of Jaguar's successors, as these CPUs (Jaguar) are really for the netbook and low-end laptop markets (well just going by things said about Jaguar once it was known XB1 and PS4 would be using them).

I don't know if it would make sense to go beyond 8 cores though, as I would imagine a semi-costom x86 K12 core, its improvements, I mean, being beefier cores all-round, coupled by a large increase in clockspeed (over current Jaguar) would be enough even for the next generation of games, and, most importantly, leaving enough room on the die for the biggest improvement to be a bigger, fatter GPU, of at least the 10+ TFLOP class (if not more) in 2018/2019.

Anyway, the overall purpose of this thread is for early discussion on any of the next consoles that would be coming out in the 2nd half of this decade, especially the ones later in that 2nd half, the next Xbox and PS5.

 

[DieH@rd]

Nintnendo needs to move to modern architecture, especially since there are rumblings that they want the SAME architecture for their next portable and home console sto that they can more easily share games between them.

AMD is their best choice, no matter of they go with X86 or ARM for CPUs. They could easily get great home console APU from them that can also be greatly cut/downclocked for use in 3DS successor. But who knows when they will move away from WiiU/3DS.

 

[Megadrive1988]

Wii U used a half-way modern but still outdated GPU architecture (equivalent of Shader Model 4.1) yet the raw performance of Latte was still below Xenos and RSX. Sure that's partly made up for with more efficient (than 3.0) shaders, but still... And the CPU, it is what it is. Nintendo should aim for "respectable" performance (whatever that might mean to you) for the time their next console is released. Hardware / power isn't everything (how many times have we heard this) but shouldn't be so far on the low end that it hurts your console.

 

[Rangers]

Anyway, the overall purpose of this thread is for early discussion on any of the next consoles that would be coming out in the 2nd half of this decade, especially the ones later in that 2nd half, the next Xbox and PS5.

Tonga is neat...some stuff that we wish was in X1/PS4

http://techreport.com/review/26997/amd-radeon-r9-285-graphics-card-reviewed

The MSI card's higher boost clocks would give it a bit more oomph in some categories than the stock numbers shown for the R9 285 above, but it wouldn't do anything to address the biggest issue. The R9 285's 176 GB/s of memory bandwidth is just a lot less than the R9 280's 240 GB/s—and quite a bit less than what the competing GeForces have to offer, too.

Compared to the Radeon HD 7950, the R9 285 is a big move up and to the left, in the direction of goodness. The R9 285 easily makes the most effective use of its memory bandwidth of any of the cards we tested. Like I said earlier, this is generational change on the GPU front.

By far the most consequential innovation in Tonga is a new form of compression for frame buffer color data. GPUs have long used various forms of compression in order to store color information more efficiently, but evidently, the method Tonga uses for frame buffer data is something novel. AMD says the compression is lossless, so it should have no impact on image quality, and "delta-based." Tonga's graphics core knows how to read and write data in this compressed format, and the compression happens transparently, without any special support from applications.

We don't have many details on exactly how it works, but essentially, "delta-based" means the compression method keys on change. My best bet is that whenever a newly completed frame is written to memory, only the pixels whose color have changed from the frame prior are updated. ARM does something along those lines with its Mali mobile GPUs, and I expect AMD has taken a similar path.

The payoff is astounding: AMD claims 40% higher memory bandwidth efficiency. I'm not quite sure what the basis of comparison is for that claim, nor am I clear on whether 40% is the best-case scenario or just the general case. But whatever; we can measure these things.

3DMark Vantage's color fill test has long been gated primarily by memory bandwidth, rather than the GPU's raw pixel fill rate. Here's how Tonga fares in it.

Between the new color compression method and the geometry performance gains, Tonga could plausibly claim to usher in a new generation of Radeon technology. The use of the GCN or "Graphics Core Next" label has proven incredibly flexible inside the halls of AMD, but what we're seeing here sure feels like a fundamental shift.

Looks like GCN (as seen in current gen GPU) is finally a bit of old news...

 

[pjbliverpool]

Looks like GCN (as seen in current gen GPU) is finally a bit of old news...

About time too. Can't wait to see bigger variations of this GPU.

 

[function]

Hopefully this "next-gen GCN" has made its way into Carrizo. Such BW savings would be a big boost for an APU that can't call on edram.

Something that might interest Nintendo somewhere down the line?

 

[Megadrive1988]

edit: meant for another thread.

 

[eastmen]

Sure

They might well go with ARM though. They really wanted to for some business reason this last time, it just wasn't quite ready yet. My guess is ARM in some form (perhaps from AMD) is the heavy favorite.

Also I hope for 2018, but it sounds aggressive (5 years).

MS could go with intel . I know there were problems with the xbox original but that was a different time. Both companys are watching market share erode and MS already uses intel in their surface line and the next windows phone is supposed to run on intel soc . So having a top down intel solution could be very good for them .

Yes Intel would have to take a cut on the pricing of their chip's but one of intel's problems is excess fab capacity. But that could in turn allow intel to charge more for desktop and laptop prices esp if MS releases Xbox games across all their platforms with a single Digital store.

 

[Osamar]

Looking at Broadwell-Y first benches, a Pro (with EDRAM) Skywell Soc DX12 will make a very good gaming machine.

http://s14.directupload.net/images/140910/zxffkjxr.png

 

[upnorthsox]

What from that slide indicates anything as to its relevance to a gaming machine?

 

[Globalisateur]

It's interesting to note that, somehow, Sony did go with AMD x86 + ARM for their PS4 design.

It's like they guessed that the future would be Hybrids x86+ARM APUs designs like the 2015 AMD APUs shown here.

Very similar to the elements of HSA they put in their console too just before AMD included HSA in their APUs.

It can't be a coincidence IMO, it's like Cerny did look closely to the future AMD designs...or/and the contrary?

 

[shredenvain]

It's interesting to note that, somehow, Sony did go with AMD x86 + ARM for their PS4 design.

It's like they guessed that the future would be Hybrids x86+ARM APUs designs like the 2015 AMD APUs shown here.

Very similar to the elements of HSA they put in their console too just before AMD included HSA in their APUs.

It can't be a coincidence IMO, it's like Cerny did look closely to the future AMD designs...or/and the contrary?

I was under the impression the Ps4 only had ARM in the offloading chip that is on a completely separate dye.

 

[Deleted member 11852]

I was under the impression the Ps4 only had ARM in the offloading chip that is on a completely separate dye.

Yup, it's incorporated into the South Bridge (ideal location for ARM TrustZone). I've not read anything to suggest it's accessible to developers and I imagine it'd be fairly useless even if it were. It's likely an ultra low power core.

 

[eastmen]

what would have been nice Dsoup is if one of the companies had taken one of the quad core arm chips with 2 or 3 gigs of ram and shoved it in the consoles and used that for the os / apps and given complete acess to the x86 soc



I have to wonder though. Will the next micron drop (22/20nm for amd or 13nm for intel) enable performance on par with these new gen systems in a 30 or 40 watt apu ?

 

[Osamar]

What from that slide indicates anything as to its relevance to a gaming machine?

It just indicates Skywell will be "full" DX12. And I suspect game companys would be more happy with full DX12 tan half DX12.

 

[Osamar]

I have to wonder though. Will the next micron drop (22/20nm for amd or 13nm for intel) enable performance on par with these new gen systems in a 30 or 40 watt apu ?

From http://www.pcper.com/reviews/Processors/IDF-2014-Intel-Shows-Core-M-5Y70-Performance-Numbers 3DMark Ice Storm Graphics = 51.376. From a friend, Phenom II 965 BE + 7850 = 163,782.
A SoC 4xCore M 5Y70, CPU 8 cores + Hiperthreading. GPU 4 slices (96 UEs), 4,5x4 plus not perfect scale, around 20-22 watts. You could add 128mb EDRAM like Iris Pro, and be under 40 watts for sure.