- 10,956
- Canada
Nvm
Last edited:
In case of Forza, better CPU could afford more resources for AI simulation physics. Problem is, FM8 and FH5 will be crossgen and, most likely, crossplay titles that means improvements in this department will be difficult.Will GT (or Forza) really be able to improve the AI much with this next tech. Or will we be stuck with 'chasing the rabbit' forever?
As a long time gamer, I know the AI in games always needs some kind of advantage over the player to be competitive, but even just more
variables would be nice, like any car in the pack is capable of finishing first or last, or is that a pipe dream?
You're expecting a step change in the way game environments are made that isn't going to happen. It'll take time. This is part of the inertia problem.
PS5 games will not make full use of the ability to pull vast amounts of data continuously for a while either, for the same reason.
So in a way MS are being sensible. Once everyone starts to get to grips with what is possible, learns new tricks and can streamline production in a different way (the downside to having more detail is that you have to make it all and tie it all together), sure, then the older / lesser consoles will prove troublesome in trying to adapt a vision, so to speak.
It won't be until developers are really pushing up against the limits of the new hardware that we will actually see any real innovation anyway. That's just how people work: constraint drives originality.
In case of Forza, better CPU could afford more resources for AI simulation physics. Problem is, FM8 and FH5 will be crossgen and, most likely, crossplay titles that means improvements in this department will be difficult.
With the move to more racing online, I don't think AI may be that much of a priority any of them, as it has not seemed to have improved in line with the increase in processing power as much as it should have up to this point imho.I'm still curious how much racing games AI can change with these new consoles though.
I think the argument is the same for both consoles: we'll see a few very high-quality "top-tier" games from the start, with a gradual improvement across the board as experience grows. That's on both platforms, just to reiterate.I'm not saying world design will immediatley change so that it reaches the limits of the hardware right away. But Sony will see gradual innovation from the get go (because first party AAA won't have to design with the PS4 in mind).
We'll only see what MS first party studios can do with XBSX exclusives 2 to 3 years down the line.
There isnt any reason to do two versions, both games isnt struggle all that much from being part of XboxOne lineup.I don't know, a stripped down FH2 came out for the xbox 360, which didn't effect the Xbox one version.
So I have a feeling crossplay titles will hurt those 'left behind' if anything. Maybe they'll get another team to work on those or just throw them out for last gen like Project Cars 2 or Dirt Rally 2.0, which didn't have a negative effect on the pro or X verisons. I'm still curious how much racing games AI can change with these new consoles though.
This is incorrect AFAIK, getting late and also bored so thought I'd give an answer on how I think it works hoping this will bore me further off to sleep. (Sorry if I made any mistakes below)Since I'm a little bored (aren't we all?) I figured why not try and explain some stuff...that no one else is going to understand or care to remember.
If you have even a passing interest in any of this techno mumbo jumbo, you may have wondered "How are Teraflops calculated?" and maybe "Where are people getting the number of Stream Processors when neither company has explicitly revealed anything other than the number of Compute Units?"
Let's start with the latter question since it'll lead into answering the former. It's actually real simple how that's done. So RDNA uses an inner infrastructure called a Shader Engine, and each engine (up to 8 of them) has multiple Compute Unit clusters (also up to 8). With me so far? So let's illustrate that using my super professional skills.
To make it crystal clear, I'm in no way claiming this to be accurate, just that it's what I know based on how RDNA works.
Xbox Series X layout would look something like this:
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
So, the Series X features 8 Shader Engines, each with 7 Compute Unit clusters (works with 7 SEs and 8 CU clusters as well, but I figured this was cleaner). What's 8 x 7? 56. But, wait, the Series X features 52 CUs and not 56! Ah, that's where disabling comes in. Disable four of the clusters (doesn't matter where for the sake of the diagram) and you end up with — you guessed — 52.
Which would now look like this:
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU - CU
SE > CU - CU - CU -CU-CU-CU-CU
Within each Compute Unit is up to 64 Stream Processors, and taking that into account we wind up with 3,136 SPs per the 7 full clusters (64 x 49) and 3,328 SPs in total when adding in the additional 192 SPs from the cut down eighth engine. With that in mind, take that 3,328 and multiply it by the GPU clock frequency, 1825 MHz in this case, and multiply that by 2 (since the GPU is doing both vertex and pixel shader operations per clock cycle) and you end up with the peak figure of 12.14 TFLOPs.
The same holds true for the PlayStation 5, whose diagram would look something like this:
SE > CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU
SE > CU - CU - CU - CU - CU - CU
To shortcut the hell out of this: 6 SEs multiplied by 6 CUs comes out to 36 total CUs. Multiply that number by 64 SPs per, and you come out with 2,304 SPs in total. 2304 x 2230 x 2 is 10.3 TFLOPs (rounded up; the exact number is 10.27). I expect everyone that doesn't already have an interest in this to have glossy eyes and drool coming out of the mouths.
four SIMDs which each include 32 ALUs
Will GT (or Forza) really be able to improve the AI much with this next tech. Or will we be stuck with 'chasing the rabbit' forever?
As a long time gamer, I know the AI in games always needs some kind of advantage over the player to be competitive, but even just more
variables would be nice, like any car in the pack is capable of finishing first or last, or is that a pipe dream?
They scrapping game worlds that costs millions to build every 2-3 years, sure this would cost alot. Industry is inefficient, thats why R* making one game this gen.The only real concern we should have is how much longer game development will take, powerful hardware is nice and exciting, but pushing for even higher production values will require even bigger teams and more time, for instance from 2006-2013 Rockstar has released 7 games for PS360, and I'm not even counting the games made for portable platforms and expansions made during this time, from 2013-2020 they've released just 1 game - RDR 2. I think the industry is running head first into a wall where AAA development becomes unsustainable despite the straightforward hardware and much better tools than in the past. I much preferred the variety and breadth of games produced in the PS2 days, the issues started to appeard during 7th gen and they're getting worse as expections keep growing. I wouldn't be surprised if we see some games take 10 years to develop even with smooth production and experienced teams just because of the amount of high fidelity assets required.
That's what I've been thinking lately - AI assisted development, but so far I haven't seen much talk about from developers, doesn't seem like the concept is doable now. There's only so much meat you can throw in the grinder, R* had more than 1000 workers on RDR2, back in the PS2 days they had a fraction of that and made all sorts of games quickly. Now GTA has skipped a whole generation and is not even on the horizon, given that they usually announce them at least a couple years before release, even if they show it now, it's a 2022 game at the earliest, but so far we got nothing, it could very well be a 10+ years long project.Next step is asset production with modern Ai tools.
The only real concern we should have is how much longer game development will take, powerful hardware is nice and exciting, but pushing for even higher production values will require even bigger teams and more time, for instance from 2006-2013 Rockstar has released 7 games for PS360, and I'm not even counting the games made for portable platforms and expansions made during this time, from 2013-2020 they've released just 1 game - RDR 2. I think the industry is running head first into a wall where AAA development becomes unsustainable despite the straightforward hardware and much better tools than in the past. I much preferred the variety and breadth of games produced in the PS2 days, the issues started to appeard during 7th gen and they're getting worse as expections keep growing. I wouldn't be surprised if we see some games take 10 years to develop even with smooth production and experienced teams just because of the amount of high fidelity assets required.
That's what I've been thinking lately - AI assisted development, but so far I haven't seen much talk about from developers, doesn't seem like the concept is doable now. There's only so much meat you can throw in the grinder, R* had more than 1000 workers on RDR2, back in the PS2 days they had a fraction of that and made all sorts of games quickly. Now GTA has skipped a whole generation and is not even on the horizon, given that they usually announce them at least a couple years before release, even if they show it now, it's a 2022 game at the earliest, but so far we got nothing, it could very well be a 10+ years long project.
I agree that the amount of work put into RDR2 is far beyond their previous work, it's a masterpiece by all accounts. But if I don't like cowboys I'm **** out of luck. In the PS2 days you would have several GTA games set in different time periods, racing game with Midnight Club, some games for NDS and PSP, Manhunt, Bully, Max Payne and others. The simpler graphics allowed a diversity in games produced in a small window of time. You know, me and you are not getting younger, so waiting 10 years for a single game is a bit sad, and it may not be the one I'm interested in. By the time next GTA comes out I will be pushing 30, and then the next one I might be over 40 years old. Yeah these games are impressive displays of modern tech and ambition, but the trade off is too much for me. It would be fine if others were picking up the slack, but they don't, there's nothing like Bully or Midnight Club or Manhunt on the market, and nobody else has Dan Houser as a writer. I always enjoyed his work but now he's retired, so who knows if next GTA will even have his style of writing and humour.RDR2 has more detail and world complexity than probably all their PS2 games combined though. You're simply looking at the number of games, instead of looking at how much work each game is.
I agree that the amount of work put into RDR2 is far beyond their previous work, it's a masterpiece by all accounts. But if I don't like cowboys I'm **** out of luck. In the PS2 days you would have several GTA games set in different time periods, racing game with Midnight Club, some games for NDS and PSP, Manhunt, Bully, Max Payne and others. The simpler graphics allowed a diversity in games produced in a small window of time. You know, me and you are not getting younger, so waiting 10 years for a single game is a bit sad, and it may not be the one I'm interested in. By the time next GTA comes out I will be pushing 30, and then the next one I might be over 40 years old. Yeah these games are impressive displays of modern tech and ambition, but the trade off is too much for me. It would be fine if others were picking up the slack, but they don't, there's nothing like Bully or Midnight Club or Manhunt on the market, and nobody else has Dan Houser as a writer. I always enjoyed his work but now he's retired, so who knows if next GTA will even have his style of writing and humour.
Not sure about refined, more corporate and safe than ever for sure. But even with 10 year dev cycles games are shipping with big day 1 patches, bugs and glitches, incomplete content and DLC planned for years in advance. Back before solid online connectivity you had to ship a finished game, polished to absolute perfection because you had no other option, in case of a particularly terrible bug you'd have to recall the entire stock and reprint the game. These days most games if not all games have less polish because of that ability to update later.On the plus side the game industry is far more refined (less shovelware) and there is a thriving indie scene.
The speculation got more crazy xDFinally all the dumb rumours and insiders will be put to bed, the speculation has become ridiculous in the last few months.
Going from HDD speed to SSD should be like Christmas for racing devs. Games will blow us away.I think tracks details will be a complete revolution. No more Sainte Croix track (emptiness and ugly textures) .
Nope, the xbox has dedicated hardware for audio & decompression, aswell.Also everyone, don't immediately discount it based on TFlops vs Xbox. All the optimizations in the surrounding hardware, plus all the coprocessors, may actually make it faster in practice than the xbox
I think the SSD itself is 825 and it'll be less for the end user. Such a fast SSD is expensive.It'll be a 1TB SSD, but after over provisioning, system requirements, and other stuff that's needed that's what'll be available for the end user - or at least that's my guess
MS already stated Xbox Series X doesn't have variable clockspeeds. While the PS5 has them.Sure, but the point is that systems usually aren't actually capable of running at their to speed all the time, as they are bottlenecked by the surrounding hardware. Sony seems to have really focused on removing those bottlenecks, so they may be able to run at their peak performance for a higher percentage of time
The State of Decay Demo isn't using any of the benefits of the Velocity engine. It's just the Xbox One game running on the Series X.Video of Series X showing a side by side comparison of State fo Decay loading
I would bet my house that the PS5 would leave Series X in the dust with that.
Don't worry. Devs will use them imo, because XSX supports RT Audio, 3D audio and their own engine aswell. Ninja Theory is very excited about it. Obviously their audio engine isn't Sony tempest engine, but both are pushing audio & habe dedicated hw for it. imo third party devs will take advantage of them.What makes me worry a little too is the possibly more specialized design, especially audio, means it will be taken advantage of if the devs put alot of effort into it. Where as the series x could just run more standard coding with a little more grunt.
Both consoles have dedicated hardware for sound and so on to minimise or remove bottlenecks all together.That dedicated 3D audio processor is intriguing. It'll basically free up (half) a CPU core for other stuff, including graphics. At some point, excess GPU horsepower isn't any benefit if you can't throw enough of anything interesting at it to process, so having some CPU power in hand might be a real advantage
His example was an HDD game. Although PS5 has an fairly big advantage in bandwidth, the Xbox SSD is an huge jump from HDD, too. Game design will chance on both consoles. With PS5 probably offering a bit more. I could see teleportation working in PS5, but not without drawbacks on Series X.Only first party studios will be able to explore the faster hardware of the PS5, since 3rd parties have to develop and design for multiple platforms and that means bottlenecks in world building, as Cerny illustrated yesterday (with a level from an old game).
This is false. The PS5 & Series X don't have multiple Zen 2 cores for I/O. What they said is their dedicated hardware frees up CPU cores worth of multiple ZEN 2 cores. But there aren't hidden Zen cores... The new consoles have 8. Not more, not less.it's I/O unit has multiple Zen 2 processors
Should be fine, if you bought the game already.I'm worried about titles that have been removed from PSN like Driveclub though - will I still be able to redownload it and all the DLCs in the future?
This interview was made in 2019 and published in 2020. Him saying 1-2 years of cross gen games means holiday 2021 should end the period.We'll only see what MS first party studios can do with XBSX exclusives 2 to 3 years down the line
They have AMD CU's custom designed for specific purposes to relieve the main CPU from carrying out those purposes.This is false. The PS5 & Series X don't have multiple Zen 2 cores for I/O. What they said is their dedicated hardware frees up CPU cores worth of multiple ZEN 2 cores. But there aren't hidden Zen cores... The new consoles have 8. Not more, not less.
His example was an HDD game. Although PS5 has an fairly big advantage in bandwidth, the Xbox SSD is an huge jump from HDD, too. Game design will chance on both consoles. With PS5 probably offering a bit more. I could see teleportation working in PS5, but not without drawbacks on Series X.
This was fast travel in the Spiderman demo and State of decay starting the game. If we want to see the difference in load times, we should wait for an third party game. Otherwise it's not a fair comparison. Obviously PS5 should be much faster there. But not 0,8 vs 10 seconds faster.Remember that Spiderman comparison between PS4 Pro and PS5?
Why's that? Developers are already seem used to building on 2+ different pieces of hardware, especially this late in the console' life span - First party and Third party alike. What would make it so drastically different?My point is that, because devs have to develop for the base Xbox during the first 2/3 years, you won't see anything (game and level design related, not loading times of performance) taking advantage of the SSD on XBSX as you'll see on the PS5 much sooner.
Why's that? Developers are already seem used to building on 2+ different pieces of hardware, especially this late in the console' life span - First party and Third party alike. What would make it so drastically different?
Agree with your argument about First Party and cross gen titles. But i disagree with this, because building an PC for 500$ (rumoured Xbox price) with the same performance will be impossible in 2020 and even looking at 2021/22 i don't think an pc stronger than an series x will be "much cheaper" than the price of the Xbox Series X.On top of that, if MS don't release XBSX console exclusives, there's no point in buying it if you have a decent PC in 2 or 3 years from now (which will be much cheaper).