Predict: Next gen console tech (10th generation edition) [2028+]

[davis.anthony]

It might be that consoles start having more than one pool of RAM again.

But it would obviously depend on cost and design complexity Vs just increasing a single pool but it might be an option.

Especially as we now have a BVH which can consume 1GB+ depending on the game.

 

[fehu]

Is it still so pricey?

GDDR6 VRAM Prices Plummet: 8GB of Memory Now Costs $27

Shouldn't 16GB come standard on cards costing $400 or more?

www.tomshardware.com

 

[Tkumpathenurpahl]

@Shifty Geezer is there any chance this could be split off into its own thread, please? I think a "memory capacity and configuration of next gen consoles" could be quite an interesting thread in its own right.

 

[Wardious]

For the ps6, i predict a 3nm apu with 24gb gddr7 and gpu perf similar to a 4090.

For the cpu, 8 zen 5c clocked at 4.5 ghz without avx 512.

It will be released between 2028 and 2030.

 

[Seanspeed]

For the ps6, i predict a 3nm apu with 24gb gddr7 and gpu perf similar to a 4090.

For the cpu, 8 zen 5c clocked at 4.5 ghz without avx 512.

It will be released between 2028 and 2030.

You think a 2028 console would be using 2024 CPU architecture?

I do think 'c' cores could make a lot of sense for consoles, assuming this sort of paradigm is still around by then.

Personally, if we're talking late 2028, I'd expect 2nm process, 32GB of GDDR7, 12+ Zen 6/7 cores(depending on pace of Zen advancements), RDNA6 GPU architecture. Monolithic. $550-600 without disc drive.

2nm might be the bolder 'guess' there, but if it's debuting in 2026, I can see sense for it to be used in mainstream devices by late 2028. Even if costly, pushing maximum yields per wafer and better efficiency still has benefits, and I think they can push consumers to take on the burden of a higher price tag as a result. Also, Intel will probably at around that same process level before TSMC and could be an option if Intel can build up scale and helps with pricing through competition somewhat. Samsung another longer shot option maybe if they somehow get everything together and take advantage of their headstart with GAA, though I feel they'll be behind on any High NA EUV.

 

[Globalisateur]

For PS6:

CPU: still 8 cores with unified L3 cache. Maybe 3d cache to improve cache size / CPU performance without increasing APU size.

GPU: I don't expect it would be based on any RDNAX. Historically I expect AMD will release a brand new GPU architecture along with next-gen consoles. more accelerated RT, upscaling and AI units and probably more exotic Cerny stuff like in PS5 GPU like cache scrubbers / ID Buffer etc.

Both of them will be clocked the highest they can with improved dynamic clocks still based on max number of instructions. The APU should be even smaller than PS5 launch APU (< 300mm²). They likely won't need chiplets.

I/O: Targeting double average compressed speeds than PS5 I/O with even lower latencies.

Memory: 32GB, 256 bit bus.

Releasing in 2028.

 

[Shifty Geezer]

I/O: Targeting double average compressed speeds than PS5 I/O with even lower latencies.

How much difference would that actually make? PS5's IO isn't being pushed even now, and lower latencies will still be too long to operate as virtual RAM. I kinda feel 16 GBs RAM, same IO, with just chunky lots more CPU and especially GPU would achieve more. Of course, you'll take the natural improvements so if twice the IO BW is only a buck or two more, you'd spend that. Putting it another way, adding more RAM to PS5 won't make much, if any, difference. Increasing PS5's IO speed won't make any difference. Increasing CPU will get some smoother framerates and upping the graphics will make everything prettier.

About the most exotic idea I can think of that's reasonable - a second pool of slow DRAM between RAM and IO. Could do something like 4-16 GBs to supplement the IO stack to get those latencies down to RAM speeds, though not main RAM speeds. Though the added complexities of more buses might make it not cost effective. I kinda feel RAM topology is the only area with room to play. Fast IO -> DRAM cache -> GDDR -> 3D cache on GPU

I think the key issues moving forwards will be getting lots of ML power that actually gets used, upscaling tech, and address the issues of BVH maintenance. Design hardware around those and they'll prop up everything else very effectively.

 

[Wardious]

You think a 2028 console would be using 2024 CPU architecture?

I do think 'c' cores could make a lot of sense for consoles, assuming this sort of paradigm is still around by then.

Personally, if we're talking late 2028, I'd expect 2nm process, 32GB of GDDR7, 12+ Zen 6/7 cores(depending on pace of Zen advancements), RDNA6 GPU architecture. Monolithic. $550-600 without disc drive.

2nm won't see the light of day before 2026 or 2027 on iphone.
Sounds complicated, not to mention the high cost of the new node.
Maybe samsung 3nm Gaa could be significantly faster than 3nm tsmc finfet and cheaper.

 

[Globalisateur]

How much difference would that actually make? PS5's IO isn't being pushed even now, and lower latencies will still be too long to operate as virtual RAM. I kinda feel 16 GBs RAM, same IO, with just chunky lots more CPU and especially GPU would achieve more. Of course, you'll take the natural improvements so if twice the IO BW is only a buck or two more, you'd spend that. Putting it another way, adding more RAM to PS5 won't make much, if any, difference. Increasing PS5's IO speed won't make any difference. Increasing CPU will get some smoother framerates and upping the graphics will make everything prettier.

About the most exotic idea I can think of that's reasonable - a second pool of slow DRAM between RAM and IO. Could do something like 4-16 GBs to supplement the IO stack to get those latencies down to RAM speeds, though not main RAM speeds. Though the added complexities of more buses might make it not cost effective. I kinda feel RAM topology is the only area with room to play. Fast IO -> DRAM cache -> GDDR -> 3D cache on GPU

I think the key issues moving forwards will be getting lots of ML power that actually gets used, upscaling tech, and address the issues of BVH maintenance. Design hardware around those and they'll prop up everything else very effectively.

We are talking about a next-gen console, not an improved PS5 like PS5 Pro. And I'd disagree about I/O not being pushed. At the very least we know they save a lot of CPU ressources based on a few PC ports. Dedicated I/O blocks is the future anyways

 

[Shifty Geezer]

We are talking about a next-gen console, not an improved PS5 like PS5 Pro.

I know. And I'm saying for next gen, there's aspects that may be ceilinged at this gen. eg. IO - faster load times might not contribute anything over what we have now. As discussed for PS5's IO, is a game that maxes out the IO feasible and cost effective? To date, no. Far slower SSDs are pretty much just as capable. Going forwards, will faster than PS5's IO make much difference? Probably about as much difference as a 4 GB/s drive now makes versus PS5's. Early showcases of PS5's IO power like the UE5 showcase turned out to be eminently doable on much slower storage with Nanite's requirements being very conservative. As all streaming largely is.

And I'd disagree about I/O not being pushed.

PS5's SSD has been replaced with a completely gimped SSD at 1.7 GB/s versus PS5's 7 GB/s and it still performs fine. Until we get a game that keels over when PS5's SSD is replaced with something slower, it's not being pushed. And until that game exists, there's little justification for even faster storage next gen.

Given a limited budget to spend on next-gen hardware, I think extending the IO stack one of a lowest priorities.

 

[Charlietus]

How do you sell a new generation console? I think next gen will be the ai generation.
So having enough "tensor cores" to effectively assist developers in stuff like the Nvidia ai demo, ai assisted animation or even texture up sampling and real time asset creation (lots of repeated ai in my post, sorry).
DLSS like reconstruction and frame generation, of course, with the first probably coming soon enough on the pro.

GPU: 35-40 teraflops RDNA 6 (something like 2x for the GPU compared to a ps5 pro, in 2029 it should be feasible) with tensor cores and dedicated rt cores

CPU: Zen 6c with 8 cores and 16 threads at 4,5 GHz

RAM: 24GB GDDR7 dedicated to games with 4GB DDR for the OS

SSD: 1 Terabyte at 7GB/s

Games using ai for gameplay will be very few, like 1 or 2 games a year, and most games will be cross gen for at least 3 years. Generations will continue to exist, but transitional periods will get longer and longer as time goes on.

 

[fehu]

BTW in 2027 PCIe 6.0 will be widespread, a faster ssd bus is inevitable, a faster ssd really likely

 

[SlmDnk]

How about a dedicated NPU (Neural Processing Unit)? Would that be feasible?

 

[cheapchips]

How about a dedicated NPU (Neural Processing Unit)? Would that be feasible?

MS were certainly thinking about it.

 

[neckthrough]

There's a single point in ARMs favor when it comes to CPUs, in that all instructions are the same length, which makes it much easier to make a "wider" CPU (one that can execute more instructions at the same time) than with x86's variable length instructions, which start to become really complicated to handle at a point we arrived at for modern CPU designs just a few years ago.

Not sure what you mean here in terms of complexity. Variable length ISAs are ugly, no question. But the dataflow backend of a modern superscalar CPU works on decoded microinstructions, which are clean. Making it wider is problematic for the wakeup/issue logic because it scales quadratically, but that is unrelated to the variable-length ISA. Are you talking about decoder complexity? Sure, that's a known quantity but I'm not aware of any inflection points that CPU decoders arrived at recently. It could be that I'm out of touch.

 

[Shifty Geezer]

BTW in 2027 PCIe 6.0 will be widespread, a faster ssd bus is inevitable, a faster ssd really likely

Sure. My point is more about choosing the right cost. If it's no more expensive to add faster IO, go for it! General tech advances might enable that. However, if it costs most and you can get 7 GB/s notably cheaper, I don't think that'd hamper a next-gen console. Hence my question what would a faster SSD enable? Is it really necessary?

 

[dobwal]

I know. And I'm saying for next gen, there's aspects that may be ceilinged at this gen. eg. IO - faster load times might not contribute anything over what we have now. As discussed for PS5's IO, is a game that maxes out the IO feasible and cost effective? To date, no. Far slower SSDs are pretty much just as capable. Going forwards, will faster than PS5's IO make much difference? Probably about as much difference as a 4 GB/s drive now makes versus PS5's. Early showcases of PS5's IO power like the UE5 showcase turned out to be eminently doable on much slower storage with Nanite's requirements being very conservative. As all streaming largely is.


PS5's SSD has been replaced with a completely gimped SSD at 1.7 GB/s versus PS5's 7 GB/s and it still performs fine. Until we get a game that keels over when PS5's SSD is replaced with something slower, it's not being pushed. And until that game exists, there's little justification for even faster storage next gen.

Given a limited budget to spend on next-gen hardware, I think extending the IO stack one of a lowest priorities.

I imagine Sony still wants to take advantage of the SDD memory costing saving without any of the computational burden on the CPU/GPU that the current system offers. Plus PS5 IO speeds is based on the performance of the decompressors as well as the SDD. The decompressors just needs to able handle the max bandwidth of whatevers current in 2028.

 

[McHuj]

I think 2nm process is a must for nextgen consoles, even if it means waiting longer than 2028 to launch. Using this article as a source (https://www.anandtech.com/show/1883...e-power-delivery-in-2026-n2x-added-to-roadmap) and normalizing PPA table to 7nm as the reference for power and perf you can get this (using the most favorable numbers in the table):

Process	7nm	5nm	3nm (N3E)	2nm
Power	1.0	0.7	0.448	0.3136
Perf	1.0	1.15	1.357	1.56

I don't think 3nm provides enough perf and power scaling to give us a good generational bump.

 

[Frenetic Pony]

Not sure what Sony and MS are doing exactly, but I know what I'd advise them to do:

Sony: Mobile PS5/PS5 Pro Switch, essentially. Keep backwards compatibility for both implicit for a long while, PS5 base BC in handheld and PS5 Pro BC in docked. People want handheld devices, and care less and less about each console generations "upgrades". Sure you can have upgrades for the future, 18-24gb of ram, a GPU with high end matrix multiplication and formats, maybe even a dedicated high end NPU. Note: on current devices these are for constant low end AI calculations while GPU is for intensive workloads. However NPUs are more power efficient here, and for a mobile console a large one that can do constant intensive work might make sense power wise.

Plus you get to take the controllers off the side and turn the screen/processing into a VR headset for cheap. Maybe Apple Vision comes along and makes AR/VR super popular in a year or two (2025, 26?). Sure this makes the base unit more expensive, you need a 4-5k screen and extra cameras and watnot. But this way you can sell a VR kit not for $549 but $199, way cheaper. And at $699 total you're probably selling a high end AR/VR headset as cheap or cheaper than anyone else is.

MS: You've got the money, split the market. For low income/mobile wanting market segment sell a mobile console at $349-399 (whatever you can get it down to). It needs to have more ram than the Series S, devs want to drop that asap. But otherwise it just needs to be something BC with the Series S that devs would rather target while mainting BC with Series X/PS5. 1080p 8" screen, only need a few upgrades other than 16gb of ram minimum that will be cheap by 2027/8. Basically a target for games to hit at 1080p 30fps.

And then for the high end one you target 4k+, 120hz? Near future 3d? (Lightfields! Think if holograms were real). Sure it's a home console but you can charge $599 for this easy. 15gbps(+?) SSD; true games find it hard to use more than 8 cores, but super high clocks and a large cache. Go all out, get as much of a chiplet based GPU arch as you can and target 5080 levels of performance, or 5090 even. Anything the mobile one runs at 1080p 30 this can run at 4k60 upframed to 4k120.


Either way any mobile console will need a few things ones today don't have, which is a cheap but useful 5g modem. An increasing number of people are mobile only, their console needs to accomodate that without the need for wifi.

 

[arandomguy]

Regarding memory I think it's worth mentioning the only publicly known roadmap so far is from Micron which shows GDDR7 being 24Gb (3GB) at 36 Gbps from 2026 through to 2028. I don't believe there is any publicly known comment on 32Gb (4GB) GDDR7 yet.