I mean if you want to give him the benefit of the doubt. Devs will know what workloads will cause down clocking and to get the most performance possible you can manage the use of those. However I don't think that is really required.
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
- Status
Judging by the bans, it would seem so. Mods don't allow persistent targeting of any specific member. I'll be honest and say that I don't know (or necessarily agree with) where the line is on that here given the relevance of his commentary to the subject of the thread.
develop maybe, but optimize? No. It's way harder to optimize for PS5 than PS4, or Xbox series X.
This is due to the variable clocks. Cerny even said that devs need to optimize their engine based on those variable clocks. They would need to do this for a single platform. Which is highly unlikely for a multiplatform game.
With Xbox series X they won't need to do this, because Xbox series is sustained under any circumstances, this does not apply to PS5. Devs would need to do something different to optimize for PS5 and no, this is NOT simple or easy.
Before we knew the specs of either consoles, Tom Warren was on Twitter saying that the Xbox would be optimized better because "Microsoft has decades of experience on DirectX". Don't get why anyone would care about anything he has to say past that point, dude really thought Microsoft's experience with APIs for game development would differentiate them from the company that made the PlayStation consoles. Clearly those guys don't know how to make an API !
I don't believe Codemasters have stated resolutions for the 120fps mode.
They didn't mention the resolution on XSX too. I expext something between 1600p and 1440p at 120 fps for consoles and this is good.
You forgot about follow up interview where he clarified his statement.
See, I was thinking about this.....but even reading the quote in the other post.....Cerny explains what he meant right after that bolded part. I'm still re reading it to see what other ways it could be interpreted. Because he mentions power consumption in that quote too.
This does clarify it tho.
User Banned (Permanent): Modwhining and doubling down on a ban for hostility; long history of modwhining and hostility
Yup I got banned for calling him out on his "journalism" with baseless research. Spouting nothing but fud.. I got banned like 15 mins after. Don't question "verified" members I guess...
That's not true at all. The PS5 devkits have different profiles to sustain the clock rate of the processors to ease the optimization of the code (if needed at all). You can use the same working flow using CPU or GPU profilers for optimization like for the PS4 or XBone. The variable clock rate doesn't have to be considered by the developers at all. Since the whole SoC (i.e. CPU and GPU) knows when the code is running (flipping bits and consuming more power) it will clock the core to its set maximum. If no activity is detected it will decrease the clock rate to save power.
Thus, the varying of the clockspeed is deterministic, because it is based on the activity of the code. Intel for instance also uses boost clocks and throttlings, but it solely based on whether the cooling system can handle the current power output.
Some clarification would be great.
well it seems you can discuss his tweets but u can't call him a warrior. I checked all four bans and they were all accusing him of being a console warrior.
The rest of the posts discussing the contents of his posts didn't get banned. Seems fair enough.
Tbh, I've been called a warrior before and told im stupid, dumb and that i should piss off from this very thread and no one got banned for antagonizing me.
Thats hilarious....
Anyway, the other part that was left out in that quote was this:
That came directly after.
This too. This is all in the same DF article.
Know what came right before all this in the article?
PlayStation 5 uncovered: the Mark Cerny tech deep dive
On March 18th, Sony finally broke cover with in-depth information on the technical make-up of PlayStation 5. Expanding …
www.eurogamer.net
Feedback from developers...questions about how it all works...
I know we like discussing things in articles, but sometimes it would be good to include stuff before and after the point you want to discuss. It helps the discussion IMO.
Aegon i see you play TLOU 2 but you notice how impressive the clothes parts were in that game.
It crazy to think what ND will do on next gen .
Try posting some FUD on Twitter. May get you a different result.
*shrug*
Lol actually i didn't. I was so blown away by the foliage, the rain and lighting that i ignored everything else.
Fog + fire + volumetric lighting = best visuals ever.
It funny cause it so impressive that people did not notice how impressive it is since it look normal .
When characters take off clothes and stuff like that they don't normal show it because it hard to do with clipping , funny animation etc etc .
ND made it look so normal .
No? He is a member of this board and as such he is not to be questioned.... otherwise you are agonizing that user....
yeah i saw you guys talking about the cutscene where she takes her shirt off. i didnt even think about how it had no clipping.
lol
brb writing a poem.
The question was already answered and didn't require your input.
Just because it's deterministic doesn't mean it is clear at the outset. Yes, when a developer sees the results, those will be the same on any console. But they may find some workloads resulting in lower clocks and therefore lower performance, and may not know that until they try.
however - if Cerny's description is accurate this will be rare, and most workloads should run at full clocks.
so Warren's comments aren't necessarily inaccurate, but I do feel the 'variable clocks' are often posted as though they're more like regular PC clocks, which they aren't. It would be nice to see More balanced reporting on that, vs what often does appear to be more fear mongering
however - if Cerny's description is accurate this will be rare, and most workloads should run at full clocks.
so Warren's comments aren't necessarily inaccurate, but I do feel the 'variable clocks' are often posted as though they're more like regular PC clocks, which they aren't. It would be nice to see More balanced reporting on that, vs what often does appear to be more fear mongering
One, was being sarcastic and agreeing with you.
Two, i wasn't that far with reading this thread.
Three, sorry that i responded and quoted you. Good to know where my place is. Thank you my master.
This place is becoming more horrible and vile by the minute.
This myth that code normally executes in trivially predictable amounts of time keeps getting perpetuated by these discussions. It might be true on a DSP, or something as ancient as a 6502, but on modern high-performance processors and operating systems you always have to test code to determine how cache misses, branch mispredictions, context switching, virtual memory interactions, and many other factors impact your code's performance. There's a good reason why measurement is the gold standard for figuring out where and why your code performs the way it does. This isn't new to even moderately skilled developers. Stop this madness and listen to what developers are focused on, not casual observers.
Knowing where 9tf and 9.2tf comes from....c'mon. That person is directly refuting what Cerny said.
Why those specific numbers and not, say, 10.0?
If a 2% reduction in frequency can bring down the power load by 10%, described by Cerny as "the worst case game," how often, then, would any realistic game code require an even bigger reduction in power?
If it's so rare, then describing that scenario as the "real performance" is nothing but baseless nonsense.
Now, if there's a credibly sourced write up saying otherwise then I'd love to read it.
If it's so rare, then describing that scenario as the "real performance" is nothing but baseless nonsense.
Now, if there's a credibly sourced write up saying otherwise then I'd love to read it.
Such circumstances should be even rarer than you think, especially for game code. Modern computers already use "accelerator" processors to outsource such tasks to DSPs for sound or modulated data signals of your WLAN, Ethernet or Bluetooth. The other juggernaut using a very high activity for a long sustained time would be encryption and the PS5 is using a custom unit for that to free up resources for the CPU. The PS4 on the other hand didn't sacrificed all its CPU power for encryption. So that shouldn't be a real problem to begin with.
Plus the PS5 has a 8 CPU cores and 36 GPU CUs. I can't think of any code to use everything to it's fullest for a any time and I'm not a real coder though.
The whole variable clock concerns are complete nonsense. Lowering the power limit by 10% even on a less efficient pascal gpu results in a 76mhz drop. Even with RTSS up I would struggle to see a performance drop on my system. The whole thing is totally overblown.
I could possibly see the system going below 2GHz when it gets a really stressful menu/map screen like what Cerny was talking about, which of course would go unnoticed. But during actual meaningful gameplay? I don't think we'll see drops of more than maybe 5%, and even those would be brief outliers.
That's literally how it's always worked.
You can't call something an optimization until after you've profiled/measured it. And it really goes for all code we write, you don't really know how something runs until it's been tested.
Different hw will of course have different performance characteristics - but optimization methodologies don't change here.
And it's good to remember that the time frames Cerny was talking about mean that CPU and GPU might be shifting frequencies multiple times within a single frame - juice up as needed, chill when not critical.
That would be a >10% clock drop and likely a >30% power draw drop. I'd be asking questions if you're a dev and you've overshot your goals by 30%.
Hey, my game requires a lot of 256 bit instructions!
(But really I'm not sure this is the right way to think about it.).
Like I said, I was specifically talking about the stuff Cerny mentioned where menu screens and map screens and the like, due to their simplicity, tend to pointlessly run the GPU at maximum efficiency and cause the power draw and heat dissipation to go crazy. Cerny did also say that running the PS5 GPU at a fixed 2GHz was looking infeasible, and I'm guessing this is the reason why since, indeed, regular gameplay really should not need such large drops.
I'm just surprised as how there's still a drought of concrete information. I thought the June presentation would have kicked off a trickle of news, maybe a tear down or showing off the ui maybe new features. There's been nothing unless they just plan to cram everything in all at once in August
It certainly doesn't tell the whole story. What does a typical instruction mix look like? What's the mix of integer vs. floating point vector work? What's your cache hit rate like? Is this typical of the instruction mix on all threads, or is the task single-threaded? Are the relevant threads scheduled across all cores, or is there affinity for a subset (and indeed can you express affinity with the API you have to work with?)
lmao. i just came to post this. i thought the floodgates wouldve opened by now.
i really need to see next gen gameplay. no offense to ratchet and gt7, but i need more than what they showed off.
I think TLOU2 is going to fuck up next gen for me since most of the game cross gen.
Was watching some of the new AC game and all i thinking what the fuck is this animation .
Was watching some of the new AC game and all i thinking what the fuck is this animation .
I think we could end up seeing that teardown sometime soon. The console is in production
Drought of information? On the console?
Outside of the PS5 price and confirmed list of launch games, we pretty much know everything important about the PS5.
What kinda info. are you specifically looking for?
I was about to post that I forgot next gen exists thanks to part 2.
- Status