What is a realistic timeline for new materials being ready for the consumer market?
Sony PS6, Microsoft neXt Series - 10th gen console speculation [2020]
- Thread starter Megadrive1988
- Start date
Inuhanyou
Veteran
It seems like for tech like this it's not even in planning phase yet. Closer to researching stage
Intel already chose they're going Ångströms, not picometers.
Silicon atom is 0.2nm, not 2nm. Also silicon atom width isn't that relevant yet, the marketing nanometers aren't actually tied to reality in the sense that any feature would actually match the marketing nm. (Also worth pointing out is that they don't actually use just silicon anymore)
The measurement 'nanometre' long ago stopped being a measure of any actual physical sized element in the microprocessors. Its just a generation label now.
Really though a whole new tech is needed. I propose analogue processors!
D
Deleted member 11852
Guest
Graphene is already used in a bunch of consumer applications, e.g. tyres. It's use is tech is more limited and nothing in the consumer space, only microfabs. How far off it is in being used in consumer semiconductors is anybody's guess.
You're right, my bad.
Inuhanyou
Veteran
One more generation until manufacturers including Nintendo all unilaterally give up on hw? That's crazy to think about. I guess that's when I will be out of the industry so I'll mark my calendar if I am still alive by then lmao.
I still wonder what angstroms is I haven't gotten an answer
Intel’s foundry roadmap lays out the post-nanometer “Angstrom” era
Could the "Intel 20A" process node (that's 20 Angstroms) rival TSMC and Samsung?
arstechnica.com
Angstrom - Wikipedia
What use is there to release a new console that doesn't offer up a tangible performance increase? PS6 could be a 10-20 year console.
Last edited:
Inuhanyou
Veteran
Seems like computer tech manufacturing is scrambling to find an answer to that. After all their business in the hw space(not just gaming consoles) outside of business to extremely huge conglomerates goes down the drain if their products also don't have any kind of way forward. And business with those conglomerates doesn't really change pace. It's not like they aren't using farms already.
Cardassian Bob
Newcomer
I guess the future will end up like the Matrix. Once we hit the process limit, the whole planet will be engulfed in server farms.
Well its gonna happen in one way or another. And its going to be our own fault for buying into it and not seeing alternative ways of spending our time valuably.Sounds crappy and dystopian
I think we will have 16 cores in next generation CPU. Here at 6:14, Mark Cerny said some developer ask for 16 cores CPU for PS5. The Jolt Physic Engine of Decima Engine is scaling well up to 16 cores, same for Id Software game engine. I suppose the limit above in the same probably communication between CCX begin to be a problem above 16 cores. The guy who created the Physic engine of Decima Engine told it was the reason when he tested it with 32 cores it was not going much faster and performance even degrade,
Last edited:
Bit what if the memory bus improves
It would be cool to see 16 CPU cores but I think 8 cores will still be "good enough", I think they'd either save money, put that ~70mm^2 (whatever it'll be) to extra GPU which could be another 24-32 CUs or combine both. If it is 2027 that's at worst Zen 5 but probably Zen 6, AMD will keep increasing clocks so 8x Zen 6 cores, 4.5GHz is possible[1]. Even with reduced caches/other cost tradeoffs the uplift from current 3.5/3.6-3.8GHz (PS5/XSS/X) reduced cache Zen 2 to 4.5GHz Zen 6 will be big, Zen 4's already >1.3x IPC vs Zen 2. With similar IPC increases to Zen 4 and 4.5GHz that's circa 2x 1T performance, add in AVX 512, maybe some dedicated accelerators and other new goodies that's a good improvement.
4.5GHz because ~0.75x of the desktop's clock for efficiency reasons, PS5 is 3.5GHz, or 77.8% of the 3800X's 4.5GHz. For the rest of the system maybe 72-80CUs RDNA 5, 3GHz+ core (assuming AMD hit >4GHz desktop), 32GB GDDR7, 28-32Gbit/s 256 bit bus (896-1024GB/s), larger focus on RT and liberal use of "AI", "Neural" or equivalents in marketing material
4.5GHz because ~0.75x of the desktop's clock for efficiency reasons, PS5 is 3.5GHz, or 77.8% of the 3800X's 4.5GHz. For the rest of the system maybe 72-80CUs RDNA 5, 3GHz+ core (assuming AMD hit >4GHz desktop), 32GB GDDR7, 28-32Gbit/s 256 bit bus (896-1024GB/s), larger focus on RT and liberal use of "AI", "Neural" or equivalents in marketing material
This is the last console generation, I am sure they will do 16 cores CPU and some developers wanted 16 cores for PS5 not PS6. I don't think AVX512 will be there too power hungry. Developer have 6 years to improve CPU game engine to be ready for 16 cores CPU. Some engine are ready Id Software, Naughty Dog, Bungie or Guerrilla Games, Unity ECS.
AVX512 isn't power hungry, 512bit ALUs are power hungry. AVX512 on the latests EPYCs is better than AVX2 in power use. Mostly because the ALUs are 256bits so you need to decode 2 AVX2 instructions or 1 AVX512 instruction to achieve a 512bit computation. Also AVX512 is a from the ground up ISA for vectorizing loops so it might be better than AVX2.5.1.5-256bits-LP-2020 Edition-Ti at, well, vectorizing loops.
Why do devs want more CPU cores? What work is there that's better on 16 cores than a GPU if it's that parallelisable?
As I see it, you have three workloads - massive float power, ML work, and that ever doesn't fit into that. Silicon budget should be distributed accordingly to what yields best returns.
I think it's way too early for 16 cores CPU, engines are not ready, don't need it, and it would be too expensive (APU size) for a console.
What they need to do is to improve current efficiency like for starters they could use a 8 cores CCX and more L3 cache on top of the usual tech improvements. For consoles and the specific requirement of always smaller APU a stacked L3 cache would be ideal!
What they need to do is to improve current efficiency like for starters they could use a 8 cores CCX and more L3 cache on top of the usual tech improvements. For consoles and the specific requirement of always smaller APU a stacked L3 cache would be ideal!
Similar threads
