For now we should let it be as it is. Both consoles in one thread. Once they're officially announced, we can split.
-
Hey Guest. Check out your NeoGAF Wrapped 2025 results here!
For now we should let it be as it is. Both consoles in one thread. Once they're officially announced, we can split.
Kazekage1981
Member
Perhaps the Zen6/RNDA5 architecture with AI integration will create breakthroughs in graphics that will not be possible on current gen. It's up to AMD/Sony/MSFT to serve that to us. I am down with both PS6 and Helix this time because they are both going off different directions without losing backwards compatibility. Still on Xbox One X, and have yet to finish a lot of games but at the same time want something new...
LiquidMetal14
hide your water-based mammals
Other than for some personal reason it's not something I still agree with as it will not create the most conductive conversation and personally I don't really want to mix both as it's already hard enough to keep up with all the other tech and then you're talking about the same technologies in one thread. It's not some picky thing it's just that people are too busy to keep up with all the little things and then they're going to cross the streams. But I can see the reasoning until there's more technical stuff but I think stuff has already been revealed that you can already speculate and start to compare a PC or something just for the sake of it since we have very little information as of now
Kazekage1981
Member
i remember back in 2020 the next gen thread was amazing with the meltdown over TFLOPs. The hype over PS5 SSD. Ah memories...
Romulus
Member
Honestly I'd take PS3/XB360 hardware with more RAM as a new generation so long as we could have those developers back again over the best modern hardware money can buy.
I'll play at shitty resolutions at this point just to have good, unique games again.
ResilientBanana
Member
AI isn't great, that's not to say it won't get better, but it's a gimmick for gaming, unless it's more interactive and used for decision making a speech. But people hate this and will shit on any game that utilizes AI.
Agreed, but we do have XSX / PS5 and they are more than capable of having games optimized for them and give use both high resolutions and creative gameplay. It's possible that modern game engines are the issue. I bet a UE4 game developed for these current consoles would be a high fidelity experience.
Last edited:
Isn't 9070XT (PS6) and 5080 (Helix) raster numbers from last year a bit odd? 15-20% perf difference seems quite low if Helix is clocked higher.
Shouldn't Helix be closer to a 4090 than a 5080?
Keeping an eye on early PS6 exclusives should provide a glimpse of what nextgen will deliver. But it's a shame crossgen will extend well into 2030s.
Shouldn't Helix be closer to a 4090 than a 5080?
They could use it for NTC on feedback, but the speedups weren't great when Compusemble tested them a few months ago. Closer to on on sample (AI decode) than on load (BCn fallback).
Should go well beyond AI considering how big the architectural change RDNA 5 is: A trinity of PT, ML, and work graphs. I'm most interested in the last one TBH.
Keeping an eye on early PS6 exclusives should provide a glimpse of what nextgen will deliver. But it's a shame crossgen will extend well into 2030s.
Last edited:
SupremeJ
Member
There are no actual figures (i dont believe in the leak comparison numbers), but there is something. We know that the Helix will be premium hardware and that the PS6 is more cost-effective due to mainstream pricing. There will probably be a significant performance difference between the two consoles. I expect more than what was the case with the One vs PS4.
However, it is also possible that there will be two SKUs, which further complicates the calculation.
Last edited:
viveks86
Member
K KeplerL2 is the only one who could answer that. The reference points for both came from him after all.
I think NTC on feedback will remain purely as an academic curiosity, just like SFS. Inference on sample will be for those who dare to make the leap and inference on load for those playing it safe or for cross gen support.
For real speedups on AI decode, next gen architectures are the only solution. Even Blackwell isn't sufficient to fully exploit LinAlg due to register pressure. Don't know what Nvidia has cooking for Rubin, but Neural arrays seem like the right direction
Last edited:
technically you need at least double the memory to get real graphical and gameplay jumps. I dont think the issue is developers, remember the jumps in memory used to be over 4x back in the days, games have gotten more complex while compute and memory jumps have slowed down. I think memory doubling is a must inorder to even have anything worthwhile tbh.
ResilientBanana
Member
While I agree, I also disagree. Double the RAM these days is a fairly significant amount compared to previous generations. I wouldn't say games are any more complex than they've ever been in the past as much as they're simply bloated. They're also only optimized enough to work with frame generation and up-scaling, never native frames or resolutions. I'd love to see games truly optimized to the hardware for a more native experience.
As of now, they have all kinds of "cheats" to get them over that finish line.
Depending on how far they're willing to go architecturally SW and HW wise with the nextgen consoles there should be an additional scaling factor, similar to how NVMEs had an impact for current gen for pregen assets. This time however It'll be pretty much universal.
NTC even if only using the BCn fallback (on load) acts as a IO speed multiplier in addition to new encryption surpassing Oodle Kraken. So even more aggressive data management than on PS5 should be possible.
If new approaches are found that're much faster than the current implementations for NTC and the HW is stronger then partial or complete ML decode (on sample) should be viable.
There's procedural stuff to reduce the pregenerated asset size. Possible that's only for non-photorealistic games.
Rn the GPU scratchpad are extremely bloated and must account for worst case scenario. Work graphs can address that and procedural assets.
The entire approach to RT BVH stack is a joke rn and can be cut down a lot.
12.5 GB -> 26.5 GB for PS6, unless they nerf it to 24GB, seems significant and I'll be interested to see the implications.
Nextgen PS6 exclusives should be an early glimpse into the post crossgen era. Entirely reasonable to expect a larger forward than current gen brought.
Kazekage1981
Member
I think for Helix the stacking of:
Variable Rate Shading
Neural Texture Compression
Sampler Feedback Streaming (updated version)
Shader Model 6.10
Direct X Raytracing 1.2
Work Graphs
AMD Advanced Shader Delivery
Mesh Shaders
FSR Diamond
among other tools/techniques/APIs I can't think of (along with updated hardware) is really going to change things....
Variable Rate Shading
Neural Texture Compression
Sampler Feedback Streaming (updated version)
Shader Model 6.10
Direct X Raytracing 1.2
Work Graphs
AMD Advanced Shader Delivery
Mesh Shaders
FSR Diamond
among other tools/techniques/APIs I can't think of (along with updated hardware) is really going to change things....
Last edited:
Clintizzle
Lord of Edge.
Pair all of that with devs who dont know how to implement and Playstation being the lead development platform. Will basically be irrelevant.
Ashamam
Member
Less true this time around. Sony has clearly stated a lot of the tech they are championing with AMD will not be exclusive. So expect texture compression to be ubiquitous for example, one of the advancements with more potential. FSR4.x will also clearly be a straight forward substitution as compared to PSSR. Main difference being PSSR tuned and optimised for a few fixed SKU's. It may be that devs don't use some of the features in order to cater to lowest common denominator (Steam machine/Series S/Switch 2 for example..), but I'm expecting that to impact both Helix and PS6 in that case more often than not.
Last edited:
Kazekage1981
Member
Next Gen Helix games from my understanding will be a PC game, not a console game. It *may* make console gaming irrelevant
Ashamam
Member
The vibe I'm getting from MS is that it will be a console to users and a PC to developers essentially. Probably more locked down on the user side than people expect actually. I think with the way costs are going they will be forced to forgo margin in lieu of future marketplace spend. Which means they will be heavily incentivised to keep you in their marketplace. So, no Steam basically and likely no other marketplaces either. But they'll stick with 99% PC dev environment to ensure all devs have to do is a final package for the Xbox store which hooks into achieves and voice chat etc.
Thats my current read, but its such a moving target who knows in a months time. You can only speculate at this point. Wouldn't surprise me if they have only just rescoped all that in the very recent past, some aspects might even still be in flux. (but that would be leaving it really late)
Last edited:
Kazekage1981
Member
They could not get PC gaming and OS into a console form factor with hardware/technical limitations 25 years ago. They have finally made it, and they are trying to make PC/MSFT gaming (with xbox marketing/branding) ubiquitous. In addition, they got NVIDIA, Intel, Qualcomm, and OEMS with their own technological innovations that is propelling MSFT and their ecosystem forward. Their ecosystem cannot be shackled down by a console with hardware limitations any longer, it would also be foolish and a disastrous money hemorrhaging decision to do so if they follow Sony or even Nintendo's business model. They have to leave this path behind once and for all and make a major change. The traditional console gaming path works well for Sony and Nintendo and that is fine. Helix as a traditional console is DOA no matter how much better it is than PS6 technically speaking. Helix is a hybrid console/PC can be used as some sort of leverage and a path forward for gaming within their ecosystem. Time will tell what MSFT's vision will be for gaming with Helix and moving beyond Helix.
Last edited:
A lot of what made gen on gen leaps was because of memory. PS1 to PS2 was over 12x more memory, PS2 to PS3 was similarly around 10x increase in memory. And I would say developer advocacy to convince Sony to double PS4 memory to 8GB essentially saved the system in terms of offering compelling gen defining experiences. So the jump in memory isnt as large as before and this has implications on what developers can do considering they have fixed resources in terms of money, time and human creativity. If the PS6 doesnt get enough physical memory and commensurately enough high memory bandwidth there is no practical amount of optimization that can deliver what consumers expect.
Wrt the difference between "native" and upscaled really there is no going back to "native" its all approximations at the end of the day even what you call native. ML upscaling is the most optimal way moving forward despite some of the headwinds for which I agree there have been instances where its really unoptimal. As well in all this I do agree games have gotten bloated as you said. Thats more creative, management issues for example if you compare the bloated games created by Ubisoft to what IO or Rockstars or Netherrealm have been making then it becomes clear its really down to managing those 3 things(time,human resource and capital) appropriately to meet the requirements.
You can add a couple of points to both. They also said DirectX Next so that could imply SM 7.0+.
Reiterating old K KeplerL2 info RDNA 5 goes beyond Blackwell architecturally.
Should know more by GDC 2027 if it repeats lastgen cadence. DX12U at GDC 2020.
The problem is also old PC and PS5 inertia. Handheld incentivizes max TAM complancency and will prolong crossgen. For a while the only truly nextgen games will be PS6 exclusives.
We'll just have to see how much farther they can push it past DLSS 4.5 Preset L, but long term the ray denoising will be more important considering the accelerating adoption of RT and PT in games.
The framegen on the market rn won't be the same as 2-3 years from now.
PlayStaytion
Member
That GPU clock right there for emulators which is what I mostly need from PCs nowadays.
I don't fully agree about ram. Most of the assets are streamed in like meshes and textures so i can see them having a fixed budget like nanite. I guess animations might require more ram but they are fairly small and its not practical to make game where core resident animations are really expansive.
Also 30gb ram and 200gb game which would already be blasted as uncompressed would 15% of total game size in terms of amount of ram. ps3 gta v was 512mb and 18gb game which was 3%. I know its bit odd example but sort of highlights the scale.
Streaming and better engine abstractions help but they're not a substitute for enough physical memory and bandwidth. The resident working set in memory is really what matters most and besides those two you mentioned there is a need for physical memory storage for visible and near visible assets, AI state, render targets, frame history for reconstruction, GPU buffers, etc, So really it comes down to what Developers need to achieve they're creative visions early on in the gen as soon as possible. And that requires a commensurately large amount of physical memory, how that is determined is through telemtry and back and forth between Sony/Xbox and the devs. So when the devs tell Sony they need a doubling of memory they know exactly what they are talking about. Thats what happened with the PS4.
If you want denser open worlds, more persistent simulations, richer animations, more unique assets, improved NPC behaviour, better destruction, higher quality lighting and more advanced ML pipelines, the physical memory budget needs to meet or exceed developer requirements. So I dont see Sony going below 30GB of memory for the PS6.
Now consider that with GTA 6 running on a PS6 popin may become a thing of the past since there is a larger active working set of memory available for near visible assets. A lot of storage especially these days is compressed data packaged for streaming into the memory with only specific data needed at any one moment. Especially now that compression plays a larger part since games are not being loaded from discs but from disk storage. So the game can depend on the virtual memory address space and page in what it needs from fast SSD storage. So the important question is not what percentage of the whole game fits in ram, but how large the active working set can be to run the game without stutters, compromises or aggressive asset eviction. And from the example you gave of 30GB memory and 200GB GTA 6 game, we could really see popin disappear.
Shogmaster
Ask me about my package
1. Sony/MS dont pay street prices (unlike Valve lolol).
2. RAM prices in late 2027 wont be RAM prices in mid 2026.
PS1 to PS2 to PS3 to PS4 was exactly 16x each time.
Very true, I was just being conservative
