Next Generation Hardware Speculation with a Technical Spin [2018]

Status
Not open for further replies.
Multithreading is superior on ARM over x86 when using 2x ARM cores, with lower power consumption.
That's a non sequitur argument though, in a situation where you have a chip with twice the amount of half-performance cores compared to another chip.

Couple multithreading cores with high performance single-thread cores for overall higher performance AND higher efficiency solution.
Oh yeah? In our dreams, can we have a set of full-fledged development tools ready-made, and low-level programming experience for ARM and Nvidia GPUs pre-loaded into game developers too perhaps?

Consistency also has value. Maybe you could cook up a piece of ARM/NV hardware which performs better than an AMD APU, but absolute performance isn't everything. Being able to take existing code and run it with predictable performance on the new system is also valuable. Being able to use your existing experience on new projects (instead of having to throw it all away and re-learn something new) also has value.

K12 ST cores = >Zen ST performance with higher power efficiency.
You have a source link for that claim, I assume.
 
It could be the next big thing for manufacturers, it will be interesting to see how "dedicated" hardware affect performances and the amount of times it saves in the greater scheme of things.
I believe manufacturers (at least one) should stick to incremental upgrade. I think foremost Sony as the jump from the PS4 to the PRO was quite reasonable and the system was introduce at the same price as the original system. I think they should continue down that road with something incremental while making clear what the future is about (ish).
It should be about presenting something sleak with strong point and a comfortable price (costumers AND manufacturers). Sony does not have the army of developpers to having crazy things happening for BC for example so I think it would be more complicated for them to move to CPU+GPU /NUMA and have everything working.
They should be "conservative" as they were with the PS4 (till they could upgrade the RAM). Clearly 8GB is fine, at least it is clearly not the thing that old the PRO the most. For me the main issue compared to modern PC is CPU power: a 4 core ZEN processor should do the trick.
Things that could improve significantly though that are not as critical is the overall power efficiency, especially the GPU so the CPU has some room to "breathe".
A more bandwidth efficient, feature rich GPU, slightly more throughput to please the masses (though not the selling point but under shooting the X1X would create to much drama).
More Bandwidth, not necessarily more memory.
If it proves to be necessary for new rendering techniques or others tasks, news type of cores (if AMD can pull the thing out of its GPU ala Nvidia).
A big HDD I mean it sounds like nothing but games and patches and DLC are so big...

The idea is pretty much to do a Xbox 1X "right" (and slim), the later really feel like a new system by it specs but it is pretty big, I guess costly to produce, no matter MSFT best efforts (no money pinching clearly) the system is still held by aging tech. They also set themselves in a situation where even topping themselves won't be that trivial (though they are likely to ship later than Sony).
Sony could use Zen core (or newer revision) which will provide a significant boost in CPU performances, a more power efficient GPU (I hope the next generation of AMD GPU). It could be a sleak box as the PS4 was may be slightly better.

Sony can/should hit the market ahead of MSFT (if the tech is available that is it...), they have a lot of users on the standard PS4, they don't to release a crazy beast with matching price but grow on what they have while continuing pushing content. They just have to deliver good Backward/forward compatibility and have up-to-date "features" running.
IF the tech is here Sony should launch in 2019, not an "impressive" system but it would fit nicely in their incremental upgrade agenda, that is great for lots and lots of std PS4 users, etc. They should be transparent about upcoming more powerful SKU too, to keep the enthusiast crown happy an get some extra dollars.
 
The idea is pretty much to do a Xbox 1X "right" (and slim), the later really feel like a new system by it specs but it is pretty big, I guess costly to produce, no matter MSFT best efforts (no money pinching clearly) the system is still held by aging tech. They also set themselves in a situation where even topping themselves won't be that trivial (though they are likely to ship later than Sony).
.

The One X is too big? :oops:

It's technically smaller by volume than the slim (One S)which had easier cooling due to traditional nide shrink.

The X is close to the limit of the wattage envelope and as much as can be expected of a new console released at that time.
The cooling and die size would have been the same most likely if it was a new gen, the architecture may have differed with no BC but wattage and die size are pretty maxed.

The X could also take slightly less cost conscious decisions as it was a premium device. A new gen will likely have to play a little more conservative to be safer for mass adoption, something the X did not have to factor as much.

I would not expec smaller for the true next gen.
 
I don't think the console should be designed around performance profiles for games that are poorly threaded. Looking at what Mike Acton and others are doing at Unity, I think the general performance and quality of Unity games is going to increase significantly, because it'll be highly optimized for any platform it runs on. I haven't watched the latest Unreal Engine GDC stuff, but I wouldn't be shocked if they're following the same path. I don't think core count is something to really worry about at this point. There may be a number of other things about ARM that makes it unsuitable for a console, but to me, that should not be one of them.

It could also be looked at as putting unnecessary load on the developers with simpler games. GTA has a lot more to gain from a multithreaded workload than devil may cry, so for capcom it would be more work without much benefit. On the other hand, with an 8 core part vs. a 16, you get the much better per core performance for the devs that only need a few threads, but other devs still have a lot to gain from more parallelism.

If there's one thing current consoles aren't lacking in, it's cores. We need as good as possible per core performance.
 
No way is the X1X too big. Probably a bit too costly to manufacture atm, but that'll change as memory prices fall and node shrinks come along.

As for another iteration of PS4, I don't think power is what they should chase until the PS5. At this point, I think form factor is the best thing to pursue.

Vega is being manufactured at 7nm later this year to "clean the pipes" and the 7nm Zen 2 is coming next year. So we know that 2019 will bring us 7nm.

That's what put me in the 2019 camp, but the promise of AMD's post-GCN architecture, coming in 2020, has changed my opinion. Because of that, I think 2019 would be the right time to refresh the PS4 and PS4Pro.

-- PS4Pro Slim --

They should focus on getting this as cheap as possible, with a sturdy design that they can manufacture for a few years to come.

- Smaller and quieter than the current Pro. It doesn't need to be as small as the X1X, but people seem to care about this sort of thing, so I guess it couldn't hurt. It needs to be quieter though - OG PS4 noise levels are unacceptable and disconcerting (seriously, my PS4 makes so much noise now, I'm genuinely concerned to play more CPU/GPU intensive games.)

- UHD Blu-ray drive. Should have been in there from day 1.

- One more USB port at the rear: one for an external HDD, one for wired PSVR.

- Built in hardware for wireless streaming to PSVR2. Or to PSVR with an adaptor attached to the headset's wire.


-- PS4 Micro --

They should take some risks with this iteration, and test the waters with a Switch competitor.

- Base unit consists of the brain and a simple dock with power, HDMI out, analogue video out, and 3 or 4 USB 3.1 type-C ports. The brain contains an SD card large enough to fit system files. The simple dock is quietly forward compatible with the PS5.

- Sell an advanced dock separately, which contains an HDD and UHD Blu-ray drive. Ideally, that advanced dock could still operate as a Blu-ray player when the brain isn't docked. Quietly forward compatible with the PS5.

- Sell a portable shell separately. In my opinion, a laptop form factor would be best, as it could accommodate a large battery, a large screen, and a large amount of storage. Quietly forward compatible with the PS5.

- Each dock/shell contains built in hardware for wireless streaming to PSVR2. Or to PSVR with an adaptor attached to the headset's wire.

- Each dock/shell takes the same power cable. Proprietary if needs be, but consistency is the main thing.

- Each dock/shell is responsible for capturing and streaming video.

- It would be great if each dock/shell could still run media and streaming apps like Youtube, Netflix, and PSNow without the brain, but I don't know the minimum required hardware for running 4K versions of those. Certainly, I've seen dirt cheap Android boxes that profess to do exactly that, but I've no idea if they're any good.

- Power draw of GDDR5 hampers portability. So, use a single 8GB stack of low cost HBM, underclocked enough to still slightly exceed 176GB/s. This could be a high volume enough item to help bring down the cost of HBM. If not, at least they're aware in readiness for the PS5.
 
Last edited:
Power draw of GDDR5 hampers portability. So, use a single 8GB stack of low cost HBM, underclocked enough to still slightly exceed 176GB/s. This could be a high volume enough item to help bring down the cost of HBM. If not, at least they're aware in readiness for the PS5.
It looks like the problem of HBM price is not volume, but the actual production, yield, stacking, integration on the interposer, make it expensive regardless of the cost per individual die. They still have many issues to solve before it's cost competitive. There's been progress recently though, speed issues are solved, and samsung said they have better thermal management.

Gddr5 is certainly too hot, but gddr6 is much better and a PS4 and Pro shrink could use half the chips at 128bit wide. According to some samsung slides, it is even more power efficient than LPDDR4 (in bandwidth/watt at least). Gddr6 also added many power management features which would make it much more usable for portable platforms in the future.
 
That's a non sequitur argument though, in a situation where you have a chip with twice the amount of half-performance cores compared to another chip.


Oh yeah? In our dreams, can we have a set of full-fledged development tools ready-made, and low-level programming experience for ARM and Nvidia GPUs pre-loaded into game developers too perhaps?

Consistency also has value. Maybe you could cook up a piece of ARM/NV hardware which performs better than an AMD APU, but absolute performance isn't everything. Being able to take existing code and run it with predictable performance on the new system is also valuable. Being able to use your existing experience on new projects (instead of having to throw it all away and re-learn something new) also has value.


You have a source link for that claim, I assume.

Here you are--https://forums.anandtech.com/thread...r-than-their-x86-one-ditch-bulldozer.2395754/

> The ISA advantage will be greatly reduced in the top-end side of the performance spectrum, but will not vanish.
> Keller mentioned during Core Day conference that his K12 core will have a "bigger engine" than its x86 sister
> thanks to the advantages of ARMv8 over x86-64, which allows to spend more transistors on compute.

I happen to know the differences between those two designs. I'm not really sure it's going to translate into a significant performance delta. My guess is maybe 10%.(David Kanter)

 
Last edited:
The ISA advantage will be greatly reduced in the top-end side of the performance spectrum
As we go forward, performance will increase more and more. So from that logic the ISA impact will become less and less.
 
As we go forward, performance will increase more and more. So from that logic the ISA impact will become less and less.
Err, that’s not logic at all.
And what if performance can move forward by methods that are disfavoured by the x86 ISA? Wouldn’t the ISA impact grow?

Then there are higher order effects such as effort of ensuring that your design is bug free, corner cases and so on, does it impact the rate at which you can introduce new designs? That you have to support legacy code at all, what does that cost?
(Apple is currently completely leaving 32-bit ARM code behind in hardware and software - can AMD and Intel? What are the overall implications of that?)

Claiming that ”the ISA doesn’t matter” is an x86 supporter retoric.

Why couldn’t intel make a competitive mobile chip in spite of sinking ten billion dollars into the project and having the best lithographic technique in the industry? How can Apple produce CPUs that perform so well vs. Intels offerings, while still being at a process disadvantage and not having the luxury of binning (and binning covers a rather large range of performance)?

The only x86 advantage are in markets where binary code inertia is strong. There is quite a few such markets, so x86 will hang around for a very long time. If consoles is one such market is debatable. It wasn’t. Is it now?
 
Last edited:
It could be considering the effort Microsoft is putting into BC and FC.
 
K12 is dead. I don't see an Arm core fast enough to allow full BC being available in the next couple of years. And in terms of pure grunt, I'm not aware of any ARM offerings that can match the highly scalable 8 core, 16 thread Zen setup AMD would likely be able to offer console vendors in the 2019/2020 period.
 
The One X is too big? :oops:

It's technically smaller by volume than the slim (One S)which had easier cooling due to traditional nide shrink

I would not expec smaller for the true next gen.
Indeed never seen it in person neither search for comparison. Bad to make assumption sometime :oops:
 
Last edited by a moderator:
Here you are
I meant an actual benchmark link, not just a spurious claim made off the cuff.

FWIW, transistors spent on x86/x64 ISA overhead is not 10% of a core's worth, but much less. I seem to recall claims of 1-2% or so of Intel's cell phone x86 CPUs from a number of years back, and those cores were tiny. So it would likely be much less for a modern desktop x64 core.
 
That's a pre-Zen 4 year old interview..
AMD dropped their ARM development years ago and Zen is pretty successful.

AMD isn't going to start working on ARM cores again just because Softbank is rich.


Masayoshi Son, the 'Warren Buffet' of Japan, 4th largest shareholder of NVIDIA, and CEO of Softbank (owner of ARM), if he wants to partner with Sony and broker a sweetheart deal for PS5, it can happen. PS5 will be a large, high volume contract.

And Sony knows that getting devs working fast and efficiently on the new architecture is orders of magnitude more important than the whims of a japanese magnate.

According to the latest rumors, earlier devkits are already in the hands of some devs.
What do you think it's more likely to be inside those devkits? A special order SoC with an ARM K12 or a regular X370 motherboard with a Ryzen 1700 and a Vega 64?
 
Maybe the Sony PS5 devkits are just Xbox One X's with Sony OS onboard... :runaway:

That's far more believable than Sony or MS switching entirely to ARM for PS5/XBoxNext.
 
Status
Not open for further replies.
Back
Top