https://www.eurogamer.net/articles/...yQotcLukZi9W0M6OAs0DLoUBJ2VkYqZ7sBh_lf8lnhelw
How many TF will the PS10 have, if it ever comes
How many TF will the PS10 have, if it ever comes
What year will be the last for consoles?
- Thread starter PSman1700
- Start date
As I commented there, there's good reason to think lithography will hit a wall long before then. A PS10 would need to be based on some tech that presently doesn't exist. It's very likely we'll be streaming by that point in a best-case scenario.
Easy one;
Year 2147, when Kuoley V get's discontinued indefinetly. You haven't heard of the brand yet because it hasn't beem created yet. You'll have to wait until 2098 for that, but then they wont even be on the games market. They'll start out selling e-staplers, will be the new hot gadget of the late 90's.
Year 2147, when Kuoley V get's discontinued indefinetly. You haven't heard of the brand yet because it hasn't beem created yet. You'll have to wait until 2098 for that, but then they wont even be on the games market. They'll start out selling e-staplers, will be the new hot gadget of the late 90's.
Changed the name of the thread to invite more meaningful discussion.
The end of consoles as we know it, ignoring market changes, will likely be when lithographic advances stop and there's no way to produce faster processors. At that point, whatever is the PlayStation generation will be the last. Ergo, can we speculate when that is, and are there upcoming technologies to keep processors advancements progression, even if at a much reduced pace?
PS4 started at 28 nm. PS5 is entering at 7nm which is three nodes? From 7nm, we have 5nm then 3nm and then maybe something, maybe nothing. So potentially, PS5 will take us to the end of conventional lithography, suggesting it's the last console or PS6 starts the last with no die shrinks and no cost reduction. If we imagine lithography can be pushed to the Angstroms level, there's still only clearly room for one more generation after that.
20 years seems a best-case possibility to me for home consoles. Over that 20 years, communications infrastructure will improve massively making game streaming all the more usable. So, 20 years. Are there good arguments against this future (ignoring all apocalyptical scenarios and assuming humankind is still alive and kicking and able to generate electricity etc)?
The end of consoles as we know it, ignoring market changes, will likely be when lithographic advances stop and there's no way to produce faster processors. At that point, whatever is the PlayStation generation will be the last. Ergo, can we speculate when that is, and are there upcoming technologies to keep processors advancements progression, even if at a much reduced pace?
PS4 started at 28 nm. PS5 is entering at 7nm which is three nodes? From 7nm, we have 5nm then 3nm and then maybe something, maybe nothing. So potentially, PS5 will take us to the end of conventional lithography, suggesting it's the last console or PS6 starts the last with no die shrinks and no cost reduction. If we imagine lithography can be pushed to the Angstroms level, there's still only clearly room for one more generation after that.
20 years seems a best-case possibility to me for home consoles. Over that 20 years, communications infrastructure will improve massively making game streaming all the more usable. So, 20 years. Are there good arguments against this future (ignoring all apocalyptical scenarios and assuming humankind is still alive and kicking and able to generate electricity etc)?
The only thing that comes to mind is the on-going experimentation with 3D stacked topography, but that seems kinda iffy when it comes to non-memory-related things.
Eventually, chiplets might be the only easy solution when balancing yields and die size, but that needs high bandwidth interconnects. hrm....
I wonder if shifting the focus to ray tracing will allow them to rebalance the GPU architecture for somewhat fixed-function...things.
Last edited:
That will never happen. First 2D litography will continue to improve even if more and more slowly with time. But then there is the untapped potential of 3D litography applied to GPU (and CPUs). What we are seeing with HBM is just the beginning.
The problem with chiplets and stacking is you're just using more silicon, not shrinking it, which means performance will not get cheaper. Presently consoles use something like 300 mm² of silicon. Now imagine lithography stopped at 28 nm. To get a console 8x the power of PS4, you'd need 8x as much silicon to fit 8x as many transistors. The PS5 APU now costs 8x what you'd previously had to pay.
We need a solution that gets more performance per mm² to be able to continue console generations. Short of a paradigm shift, I can't see any clear solution. Maybe we'll replace the binary digit with something else? Analogue computing?
We need a solution that gets more performance per mm² to be able to continue console generations. Short of a paradigm shift, I can't see any clear solution. Maybe we'll replace the binary digit with something else? Analogue computing?
Last edited:
3D stacking will allow for smarter and smarter solutions. The miniaturization will be lower and lower, that is in 2D because the miniaturization in 3D (per mm^3) will have bigger benefits than on 2D only. Then speed light to transmit stuff and emit less heat (so they'll be able to overclock it like crazy).The problem with chiplets and stacking is you're just using more silicon, not shrinking it, which means power will not get cheaper. Presently consoles use something like 300 mm² of silicon. Now imagine lithography stopped at 26 nm. To get a console 8x the power of PS4, you'd need 8x as much silicon to fit 8x as many transistors. The PS5 APU now costs 8x what you'd previously had to pay.
We need a solution that gets more performance per mm² to be able to continue console generations. Short of a paradigm shift, I can't see any clear solution. Maybe we'll replace the binary digit with something else? Analogue computing?
Then faster than light transmission of data.
I have faith in the researchers around the world. The market is being milked with silicon chips as long as the advancements continue. 5-10 years maybe?The advancements are rapidly decreasing. The pressure for finding new technologies in the never ending need for more performance is increasing and in turn increases the investments in the area. Naming, compound semiconductors, quantum computing, cold computing, optical transistors and nano magnets. Also this maybe as a step forward
https://www.sciencealert.com/new-ch...e-gap-between-classical-and-quantum-computing
https://www.sciencealert.com/new-ch...e-gap-between-classical-and-quantum-computing
If any of that takes off, it will probably be a long time away. Consoles may go on hiatus for 20 years until new processing technologies and development paradigms become usable in the home.
Somewhere around 2027 or 2029 depending on timelines.
I agree with this. Actually today's internet infrastructure is already pretty good for stream gaming. Bandwidth wise it's never a big problem (20Mbps is pretty good for streaming 4K video, maybe a little higher for low latency video required by games, but it's still kind of a low end bandwidth these days). The bigger problem is latency, but it can be solved by putting more data centers near population centers (for example, the latency of pinging www.google.com from my computer is 10ms).
Some companies are already gaming people with gambling or rewarding player engagement/time.
Playstation will probably just be Playstation, a service. Sony already experiments with this PSnow, maybe people won't have the hardware at home but a 'thin client' like system. But that's 20 years or so to go, a very long time.
They trademark it not per-se for the hardware every 7 years.
They trademark it not per-se for the hardware every 7 years.
DmitryKo
Veteran
The current end of MOSFET scaling is projected at either 3 nm, which is about 5-7 years away, or 2 to 1.4 nm which would be 7-10 years. After that, the move to 450 mm wafers will allow some further cost reduction.
Meanwhile heterogeneous integration with 3D die stacking of multiple processing and memory dies in a single package (i.e. Intel Foveros) should increase the effectiveness of existings architectures. Then there will be more powerful design tools and manufacturing processes and that may allow true 3D circuits with multiple independent layers. Then there wouild be improved transistors that use quantum tunneling effects, on-die optical signal paths, non-binary (three- or four-state) logic, neural computing or quantum computing chips, etc.
I believe the end of the gaming console (and personal computing in general) is greatly exaggerrated, unless in 20 years everyone has a 1 Tbit/s Internet connection to the clould, and the whole Arizona Desert is covered with Google server farms - which come with trees of solar power panels (and resemble human growing fields from The Matrix for extra dramatic effect) and require 5 TW of power to simulataneously render 3 billion photo-realistic 16K 120Hz 3D streams and deliver them with 1 ns latency to each individual gamer's 100 inch wall-screen .
Even in the worst case scenario, if the scaling just stops at 3 nm, further performance requirements should be possible with NUMA nodes - so a typical enthusiast PC (or game console) of 2040s will probably follow the design of the CRAY supercomputer of 1980s, with a liquid-cooled fridge-sized 19" rack cabinet that hosts multiple processing blocks and consumes 20 kW of power. Less advanced users, who wouldn't be comfortable with having such industrial machinery at home, will probably use a single-processor tower case, though liquid cooling will still be required , but that's would only offer 'good enough' computing power - that is running holographic AI assistants for ordering food from local Amazon Stores, and going on AR trips to the Caribbeans as they were before the Global Flooding.
(Also, because these 3D stacked CPU cubes will be unable to exceed current 5 GHz frequency limit, thier performance will be rated by weight - so instead of bragging about 16-core 5 GHz processors, people will say 'I have 50 kg of processing power in my home PC' etc.)
OK, just kidding. I'm not really sure why I even post in this thread... guess it would be fun to re-read this after 15 more years pass by.
Meanwhile heterogeneous integration with 3D die stacking of multiple processing and memory dies in a single package (i.e. Intel Foveros) should increase the effectiveness of existings architectures. Then there will be more powerful design tools and manufacturing processes and that may allow true 3D circuits with multiple independent layers. Then there wouild be improved transistors that use quantum tunneling effects, on-die optical signal paths, non-binary (three- or four-state) logic, neural computing or quantum computing chips, etc.
I believe the end of the gaming console (and personal computing in general) is greatly exaggerrated, unless in 20 years everyone has a 1 Tbit/s Internet connection to the clould, and the whole Arizona Desert is covered with Google server farms - which come with trees of solar power panels (and resemble human growing fields from The Matrix for extra dramatic effect) and require 5 TW of power to simulataneously render 3 billion photo-realistic 16K 120Hz 3D streams and deliver them with 1 ns latency to each individual gamer's 100 inch wall-screen .
Even in the worst case scenario, if the scaling just stops at 3 nm, further performance requirements should be possible with NUMA nodes - so a typical enthusiast PC (or game console) of 2040s will probably follow the design of the CRAY supercomputer of 1980s, with a liquid-cooled fridge-sized 19" rack cabinet that hosts multiple processing blocks and consumes 20 kW of power. Less advanced users, who wouldn't be comfortable with having such industrial machinery at home, will probably use a single-processor tower case, though liquid cooling will still be required , but that's would only offer 'good enough' computing power - that is running holographic AI assistants for ordering food from local Amazon Stores, and going on AR trips to the Caribbeans as they were before the Global Flooding.
(Also, because these 3D stacked CPU cubes will be unable to exceed current 5 GHz frequency limit, thier performance will be rated by weight - so instead of bragging about 16-core 5 GHz processors, people will say 'I have 50 kg of processing power in my home PC' etc.)
OK, just kidding. I'm not really sure why I even post in this thread... guess it would be fun to re-read this after 15 more years pass by.
Last edited:
Similar threads
