Hendrick's
If only my penis was as big as my GamerScore!
I think SSD is important, but the difference in speed not so much. Ray tracing can also impact game design. Go watch the DF Minecraft RTX video and tell me it isn't a complete game changer.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
PS5 vs Xbox Series X ‘Secret Sauce’ – SSD Speed And Velocity Architecture
- Thread starter pasterpl
- Start date
LittleBigGamer
Banned
We know what more teraflops can bring, but we really don't know what more SSD speed brings.
So we can only guess. For example, you say speed will not matter that much. I say I really don't know.
If someone seems to be sure that this difference in speed between the two consoles is game-changer or nothing. He does not know what he is talking about. Because we literally haven't seen yet even one finished game designed with an SSD in mind.
This is something entirely new. So it is better for us to focus on what we know. TFs, memory etc.
Last edited:
Thirty7ven
Banned
It's a game changer for game developers (artists), because just like Minecraft shows, it's easier for artists. Which by the way only means anything if the game is developed from the ground up with Ray Tracing in mind. Otherwise it won't even help with development time.
Game design? Sorry but no, it's lighting and I don't care what people like Alex say about that end as they are bonafide pixel counters.
Last edited:
Hendrick's
If only my penis was as big as my GamerScore!
That was my point. We do know what Ray Tracing can bring.
LittleBigGamer
Banned
True. More or less for both RT and SSD
Hendrick's
If only my penis was as big as my GamerScore!
You're wrong. It's not a tool to make artists jobs easier, don't downplay it. It absolutely changes the way light and even sound reacts to the game world.
It is definitely the next big graphical feature that will turn heads and push boundaries, but it is also much more.
You can in real time as the player, interact with and affect change on light (and sound) in ways that would never have been possible before.
How do you not see the potential game design implications in that?
polybius80
Member
that is not how it works you actually have it backwards, PS5 variable speed only means it can adjust the cpu/gpu clocks to what it really needs, there is no reduction in speed in more demanding scenes(quite the contrary), if an area is not demanding the console will reduce the clock if the area is more demanding it will up clock, in games not every frame cost the same that is why in performance tests in youtube where the frame is not capped or vsync there are parts where they can have more frames per second than other parts of a scene, the variable speed its a mere optimization process similar o what cpu's do with repetitive checks, if it evaluates thousands of checks where the result is true or false and most checks are true then in the next checks it will asume is true and executes acordingly before the result is evaluated(to save time that otherwise is idle) and if its false then it rollback the changes and do the false branch that is a very good optimization(as long as you make checks that the code you assume is true is not making illegal things before you execute it
) for repetitive tasks another similar example is variable resolution, if a frame is predicted to take more time than expected then the game reduce resolution to prevent lower framerate and increase resolution when frame time is back to normal PS5 will increase clock when needed and reduce it when its not needed that is a very good optimizationtechnically speaking all consoles have all power available all the time but they realistically cant and wont use all of it, the reason is that the power metric is an amount of calculations over time(per second in this case) in a game you have to do many things and lot of them involve loading things, wait for a result and send work to other processors those actions take time and dont give "tangible results", any time you lose waiting for something to resolve is equal to an amount of that theoretical limit that second or time frame that wont be used, your cpu may say its busy but that doesn't mean there is no idle spaces during time frame, that is why optimization is way more important than bruteforce also is good to consider that the amount of use of a processor CPU, GPU or any other depends the game and what its doing
Last edited:
Thirty7ven
Banned
Oh man you really coming up with Alex’s idea for a puzzle that works around light? That’s it? Clearly coming from the dude(Alex) who couldn’t fathom how a SSD could change a game aside from loadings.
I know what Ray tracing is, it’s not new technology. It’s only now starting to become possible in games, but it’s old stuff.
Sure sound, that adds immersion specially in VR no doubt. But lighting and reflections, these are things that will make the game look better but they won’t change the actual games. Once developers can fully get on board, it will change development because devs get to see the changes they make in real time. For smaller studios? It will be a game changer. You can iterate so much faster!
Sure you could for example make a game sequence where you the player are in a hall of mirrors, and suddenly ray tracing is enabling a gaming moment that wouldn’t really be possible otherwise. But that’s a moment, it’s not a game.
For a graphics nerd, it’s the holy grail, but for game design? SSD, CPU, RAM, these are the things that enable you to do games not possible before.
Last edited:
Jon Neu
Banned
What it means is that the GPU and the CPU can't have the maximum perfomance at the same time for a significant period of time.
And in game development, the CPU is the one who is going to be sacrificed most of the time in favour of the GPU.
Shmunter
Member
They can be at maximum MHz all day and all night. Only when the power draw threshold overreaches they get throttled to maintain a power ceiling.
The crazy thing is that power draw runs rampant when there is not much on screen, as mentioned in example; map screen on Horizon. No I don’t get this either, it’s counter intuitive.
Jon Neu
Banned
It's easy though.
When GPU (or CPU) stalls most of the time? Waiting for memory.
So the question becomes, why there is a lot less waiting for memory in simple scenes?
To understand that you need to remember how memory is fetched. To get most of the memory bandwidth it should be fetched by small blocks (use all the memory controllers) and predictable pattern (memory controllers have a lot of requests, to serve them optimally it needs to predict well where the next access will go).
Simpler scenes answer to both: you have not a lot of data and the computation complexity is small, i.e. it's easy for the controllers to predictably fetch stuff.
The most complex stuff (computationally) is RT, it's so bad at utilization, that you need a special fixed path hardware to make some of the computations. Otherwise it will just eat all the bandwidth and all the GPU cores for nothing.
pawel86ck
Banned
SDD will not help PS5 if developers on PS5 will be forced to turn off RT completely.
thisgengaming.com
Ex-Sony Dev: PS5 Will Struggle With Big Third Party Open Worlds; Can’t do Real Time Ray Tracing - ThisGenGaming
Whilst the official specs are out in the wild, it’s still unclear just how big the power difference is between the Xbox Series X and PS5. Since the reveal of the PS5 specs, we’ve seen multiple Sony developers come out and say that the PS5 performs better than the specs suggest and the power...
thisgengaming.com
Shmunter
Member
Just realised this is the guy with the audacious tweet repeating his grievances again. Man Sony must’ve touched him in a bad place..SDD will not help PS5 if developers on PS5 will be forced to turn off RT completely.
Ex-Sony Dev: PS5 Will Struggle With Big Third Party Open Worlds; Can’t do Real Time Ray Tracing - ThisGenGaming
Whilst the official specs are out in the wild, it’s still unclear just how big the power difference is between the Xbox Series X and PS5. Since the reveal of the PS5 specs, we’ve seen multiple Sony developers come out and say that the PS5 performs better than the specs suggest and the power...
Jon Neu
Banned
He is not lying, though.
The XsX is much more prepared to do Ray Tracing than the PS5.
The fact that all the talk Sony does is about the SSD and audio while MS has already shown Ray Tracing demos, already tells everything.
Shmunter
Member
PS5 can’t do real-time raytracing? Will struggle with open world?
Quick someone call Sony before it’s too late, they’re doing it wrong! Lol
That’s not what he is saying ... he states that Series X is more prepared for ray tracing and he is right.
He is not saying PS5 can’t do ray tracing, it can, but it looks like it will be a little less capable than Series X.
Like Series X a little bit less capable in the SSD department.
Last edited:
Shmunter
Member
Don’t make me dig up the tweet or listen to that click bait titled video.
What you say is legit however.
pawel86ck
Banned
I can say the same thing about Ali Salehi, yet diehard playstation fans still believed what Ali has said in regards to PS5 power.
Thirty7ven
Banned
Third party games are right around the corner. If the PS5 is supposed to struggle, it will be apparent pretty quickly.
I’m sure some will keep delaying “Its because devs still aren’t doing it right!” aka, the lazy devs routine, but the facts will be made physical.
That ex killzone 2 developer dude is a tosser and I’m sure there’s a reason he no longer works in the industry and nobody paid mind to the bs he spouted before.
The Crytek dude looked like an idiot too, and he had to take a step back. This idiot though? Still trying to squeeze himself into the conversation.
Sony hasn’t shown RT but they haven’t shown anything so saying they are avoiding showing something is console warrior talk.
MS should be praised for coming forward and hype their base, but they really do need to talk about next gen.
I’m sure some will keep delaying “Its because devs still aren’t doing it right!” aka, the lazy devs routine, but the facts will be made physical.
That ex killzone 2 developer dude is a tosser and I’m sure there’s a reason he no longer works in the industry and nobody paid mind to the bs he spouted before.
The Crytek dude looked like an idiot too, and he had to take a step back. This idiot though? Still trying to squeeze himself into the conversation.
Sony hasn’t shown RT but they haven’t shown anything so saying they are avoiding showing something is console warrior talk.
MS should be praised for coming forward and hype their base, but they really do need to talk about next gen.
polybius80
Member
no actually it was specifically addressed
"There's enough power that both CPU and GPU can potentially run at their limits of 3.5GHz and 2.23GHz, it isn't the case that the developer has to choose to run one of them slower."
both can run at their maximum speed if needed nobody has to control anything in particular, it can upclock/downclock as needed, the amount of work varies from game to game and frame to frame in XSX this also happens the difference is that it will be idle for the time its not being used while PS5 will also be idle but can also reduce the clocks and manitain lower clocks if the workload is not demanding, considering that the CPU this generation is considerable less powerful than the CPU in the next consoles it will down clock in the PS5 because it only needs part of its performance to do the same works as the jaguar CPU during the same frametime that is why it was mentioned the CPU down clocked when running PS4 games in PS5 because the CPU is too powerful for the requirements of those games per frame, the same will probably happens with cross-generation games unless there is a difference like running at higher framerate or things like that that can require more CPU but if its not the case then CPU can downclock it will make the console run more cool and wont affect the game, for example fighting games are usually more simpler for current CPU you can expect the next street fighter or soul calibur to run at lower clock CPU for example
Last edited:
Lort
Banned
What is actually said was technically they both can run at the highest speed so long as they weren’t running power hungry instructions. SIMD and other instructions require more power and hence cannot be used at the highest clock speeds.
Brings new meaning to the term Reduced Instruction Set Computer I guess lol.
The simple obvious fact is if the ps5 could sustain both running at those clock speeds they wouldn’t need to be throttled due to heat/ power usage.
polybius80
Member
actually that is not what was said, wrongly some site interpreted it that way that is why Cerny did a follow up interview
"Developers don't need to optimise in any way; if necessary, the frequency will adjust to whatever actions the CPU and GPU are performing. I think you're asking what happens if there is a piece of code intentionally written so that every transistor (or the maximum number of transistors possible) in the CPU and GPU flip on every cycle. That's a pretty abstract question, games aren't anywhere near that amount of power consumption. In fact, if such a piece of code were to run on existing consoles, the power consumption would be well out of the intended operating range and it's even possible that the console would go into thermal shutdown. PS5 would handle such an unrealistic piece of code more gracefully."
to put this more clearly if you run code that use XSX at maximum speed all the time forcing it to run all calculations possible during an extended period of time it will run into thermal problems and will have to throttle back, shutdown or run a procedure for that event like other procesors do, nothing to worry as games dont work that way
a code like that will also cause the same in PS4, Xbox one, Xbox one x, PS4 pro and older consoles, also the same in PS5 but you already knew that, its just that for some reason you appear to think that games use every part of a hardware(in a literal sense) all the time and appear to think other consoles are immune to that
the simple obvious fact is that there are lot of people that don't understand tech and tech interviews( assuming they actually read them) and likes to twist quotes and pretend to know how games work
Last edited:
Panajev2001a
GAF's Pleasant Genius
Not all workloads are the same, so you could run both at their highest clockspeed depending on the work required. He was mostly talking about AVX-512. Some other consoles need to disable SMT, some may need to downclock a tiny bit in some rare instances running particular workload that dev tools should make possible to get a handle on very early on and over time people will find better and better balance out of the components (thanks fixed HW specs).
On Xbox 360 and PS3 ensuring you were not running into macro-coded instructions and got bitten in the ass by when moving data across registers https://www.gamasutra.com/view/feature/132084/sponsored_feature_common_.php).
Last edited:
Dory16
Banned
My issue with Cerny is that he often contradicts himself. It gives the impression that he thinks that only intellectually inferior people are listening to him and it's a very insulting feeling. I'll give you 2 examples.
In the very quote that you posted hes' telling us that developers do not need to optimise in anyway. Yet later in the same interview, he says the below:
So they have put feature in dev kits to facilitate something for the developers that they do not need to do in any way? When you imagine how much it costs to produce hardware even for dev kits,you start to suspect that you are being intentionally misled.
Other example is simply what he said about higher clocks vs more CUs in his PS5 presentation. To summarize, according to Cerny it's almost impossible to keep a high number of CUS efficiently fed, and you're much better off just having fewer CUS that can do a lot well, in other words clocked higher. Multiple benchmarks have proven that what he said does not happen in practice. You get more performance in resolution and framerate with more CUs, not with higher clock rates and fewer CUs. Not only that, but more CUs is actually the precise solution to the CU inefficiency problem that he was describing and that is detailed in this article:
Sony May Be Overselling Aspects of the PS5's Hardware Performance - ExtremeTech
The Sony PS5 and Xbox Series X will both be powerful, but Sony might be overselling how well the PS5 will compare to Microsoft's console, at least by a little.
www.extremetech.com
Cerny has the right to defend and promote the technology that he designed for his employer. It's even his job to not just be truthful but marketing efficient.
It deosn't prevent us however from submitting what we're hearing to the truth test and that's also necessary when it comes out of the mouth of the likes of Phil Spencer or other MS spokespeople by the way.
Taking PR talks as the gospel and not questioning what is easily questionable (even accusing others of not understanding when they do) does not advance the cause of consumer protection which is well needed in the video game industry.
Last edited:
polybius80
Member
I think I understand your problem, it seems you dont undertand what dev kits are and what they are for and of course the features appear nonsensical to you when compared to what the final system is supposed to do
they put this kind of features in a devkit because the devleopers intention is to make a lot of different tests in the system in order to understand the capabilities of the systems and the different gains and problems of the algorithms they come up with to resolve diffent problems, not all devkit are the same, there are devkits specially old ones that require a computer that conect to a "cartridge port" or a special "cartridge" to upload code and run and not much else those are usually known as test units(consoles that run unsigned code), others have a conection tho check different stats in the console like check its ram for debugging , there are kits that use network adapters to transfer executable files to test into the unit despite the final console wont do that, there are devkits that include a pc interconected to different parts of a moded console to see stats of different parts of a hardware(the ps2 "tool" for example) and of course you may be need to control the clock and check what effects have or have a clear picture of the impact of certain codes check memory leaks and a lot of other things maybe to test what what happens with pal and ntsc systems
for example what if they want to test a very heavy process but want it to run in the slower clock posible of a console like PS5 or like nintendo switch that can run at different speeds, PS5 is automatic but I want it to be manual to see the real impact of my code in switch there are profiles I cannot simple run in high profile and "asume" the impact when running in undocked I need to know what specific thing cause performance problems, in Ps5 ideally I want to make my game as good as I imagine without compromises but at the same time I want it to run very lightweight as possible so you ideally prefer the console to use lowerclocks as long as not compromise what you want in the game specially if you are indie, in future games I will push further or can work giving advice for other programmers , and what if I intend to run a complex game at 8k? of course Ill want to know exaclty the impact of my algorithms, there is nothing weird for PS5 devkits to have that ability
I made some tests in opengl for example years ago to make functions to make text in screen, my computer was very fast and ran my demo at near 200 fps but with text it ran at 100 fps it was acceptable for my PC but I intended to run the game 60 fps vsync in a netbook(the first ones) and it ran bad so I had to modify my code, that day I learned about drawcalls and how they impact performance I fixed my code creating a 3d model with quads for each glyph , every quad changed its UV maps to the coordinates of a glyph in a texture then when the entire paragraph was ready I send a drawcall before that I made a quad for each glyph and made one drawcall per character that caused the problem with the new method I ran at very high framerat in a netbook and could write more text than needed than same code worked for 2d tiles later
Dory16
Banned
Thanks for sharing your experience of using devkits, that was insightful. You explained very well why developers DO NEED to optimise on the PS5 like on every other console and the fact that fixed clocks are needed for that is actually besides my point. It's just one of those statements that make you think that Cerny consciously oversells his product: "Developers don't need to optimise". That's car salesman language. Like "the real mileage of this car is way below what the clock shows".
You left the second glaring contradiction in Cerny's pitch completely unaddressed. I don't blame you, I actually respect that you didn't insult my intelligence by attempting to dispute that one. Cerny's praise for the superior performance of smaller GPUs with higher clocks simply has no substance in physics.
Last edited:
polybius80
Member
I dont think was necessary as again is misconception in your part, everything Cerny said was basically that he see better to have fewer CU more occupied than having lot of CU, that is true as an optimization but its not really that relevant to performance, actually it has a lot of physics involved XD is the classic example of having certain amount of workers, and doing the same work with half the workers as long as they work twice as fast, its more efficient but its the same work so is not that relevant performance wise, were is more relevant is the cost of the GPU as you can make more GPUs per slice if your GPU is smaller, you are basically making a storm in a glass of water
I recommend you leave your fanboy hat in house, you may not like that Cerny doesnt work in your favorite brand of consoles but he has an impressive record in this industry when it comes to extracting performance from consoles I am not saying he is perfect or take the absolute best design choices, but when it comes to discuss tech and how it works, if you start a discusion saying "Mark cerny contradict himself" and the reason you give is because you couldnt think of a good reason to put a clock control in a devkit for a console that can vary its clock, chances are you are not ready for that kind of discussions, better make question instead or arrogantly point fingers at professionals with far more experience than you
Last edited:
Ascend
Member
1)
If the console could handle both the CPU and GPU at max load, why would developers have to choose a profile to throttle the CPU to ensure the GPU runs at 2.23 GHz? Timestamped quote by DF....
"More than one developer has told us that they are running the CPU throttled back, allowing for excess power to pour into the GPU to ensure a consistently locked 2.23 GHz."
2)
If developers prefer non-variable clocks for optimization, why have variable clocks if the console can reach the max clocks at max workloads at all times anyway?
Last edited:
Rightisright
Banned
Exactly. If there is no issue with running both CPU and GPU at full clocks, just lock them in at full speed.
TBiddy
Member
This is an honest question. If the PS5 can run both the CPU and the GPU at full clocks, at all times.. why the heck did they introduce variable clock rates then, and spend so much time on it?
polybius80
Member
developer are choosing a profile in devkits, is good for making optimization they also mention it in the article
here is direct quote of Mark Cerny
"The CPU and GPU each have a power budget, of course the GPU power budget is the larger of the two. If the CPU doesn't use its power budget - for example, if it is capped at 3.5GHz - then the unused portion of the budget goes to the GPU. That's what AMD calls SmartShift. There's enough power that both CPU and GPU can potentially run at their limits of 3.5GHz and 2.23GHz, it isn't the case that the developer has to choose to run one of them slower."
sorry but between a comment from DF that lack context and with a possibility of misinterpretation vs a quote from the system architect that specifically say both can run at max speed I stay with the system architect comment
AMD SmarthShift
you are mixing 2 different situations
when you are optimizing you may prefer a non-variable clocks because you want to see how much of a processor is being used, you don't need max cpu speed for optimization you only need for it to not suddenly improve its capacity, when a dev optimize something it doesnt necesarily mean runs the whole game, just a small program and see how it behaves increasing the loads, also low clock sound a good option when optimizing given that the profiler tools analyze less clocks so the bars look more clearly not tiny bars indicating a spike in performance that you may miss
variable clocks is very similar to how dynamic resolutions vary the frame size depending the workload when your game is doing something that require max clocks then you will use them IF you require them and both IF YOU REQUIRE BOTH because higher graphics doesnt mean high CPU usage for example minecraft RTX or quake RTX are not going to require more AI or physic calculations than what their non RTX counterparts use, the expensive parts is the raytracing and that is GPUs jobs, we dont know how much CPU new games requires, but if DX12 and Vulkan proved something is that a good chunck of the CPU was dedicated to drawcalls given that drawcalls are not a problem in consoles(and also improved in PC) and physics now are made in GPU(in current consoles ) there is a lot of CPU performance gains in jaguar but we dont know how much is needed in newer games so we dont really know how much CPU will be required in newer games but in my opinion not much in the first 2 years as there will be lot of cross-generation games
why not run at high clock all time? in my opinion if you can downclock and not compromise performance then DO IT, when frames dont require much computation and energy power down clock will reduce the heat generated, when that heat is dissipated then the cooling system is more cool by the time the clocks get high again also is good because if not much CPU is used in the game then downclocking it will make it generate less heat so more cooling for the GPU the algorithm for the clock is not based in temperature is based in work load but it will improve thermals as only what is needed most generate most heat that is great idea
Last edited:
Ascend
Member
"Potentially". And you're confusing clocks with workloads. I can lock my CPU at max frequency in Windows. Even if I let the PC sit idle, it will run at its max clock, but will barely consume any power.
If I put the CPU workload is at 100%, the power consumption will rise significantly.
The same can be done for the GPU.
And more importantly, Cerny mentioned they had trouble maintaining locked 3 GHz on the CPU and locked 2 GHz on the GPU, doing things the traditional way. But somehow the console can handle both the GPU and CPU at max workloads and clocks at the same time? Yeah right.
That does not address the question in the slightest.
The reality is this. I'll try to break it down as simple as possible;
- The PS5 has a max power budget based on the PS5 cooling capabilities.
- That power budget needs to be divided between the GPU and the CPU.
- The max clocks can be reached for both as long as that power budget is not exceeded.
- The actual power used by the system depends on the workload of the components.
- It is rare that both the CPU and GPU are at 100% workload at all times.
- If the power budget is at risk of being exceeded, the clocks of the hardware that has the lower workload, be it the CPU or the GPU, is lowered in order to not exceed the max power budget.
- The system cannot handle both the CPU and the GPU having max workload, but it can handle both of them having max clocks.
- Developers can choose to limit the workload on one to max out the clock speed of the other.
Last edited:
polybius80
Member
yes "potentially" as in "potential" is there if you want to use it
you can have your pc idle or you can have your pc idle with less clocks
you can do something small with full clocks you can do the same small task with less clocks
at the end of the day we intend to run a game and they run in frames
do you have the link?
in question 1 you ask about profiles, those are for devkits
in question 2 you are mixing two things, when developers make optimization they dont want variable clocks because want to test how many clocks it takes a algorithm the devkit have profiles for that, when developer dont need to optimize the can simply run the game without a profile and the system will just automatically adjust itself for what they intend to use to save power, if they dont use much CPU then it can be reduced if they require more CPU the it will use more they can test and run in a profile with specific clocks or just as the system run normally that depends what they need to optimize or check
games have a tendency to use more GPU than CPU
that is why AMD do this
SmartShift Gaming Power
AMD SmartShift technology allows the processor and the graphics to consume power from a shared power budget by dynamically shifting power depending on the task at hand. In the above gaming performance example, the power has been shifted from the processor to the graphics to enable improved gaming performance.
- The PS5 has a max power budget based on the PS5 cooling capabilities.
Inside the processor is a power control unit, constantly measuring the activity of the CPU, the GPU and the memory interface, assessing the nature of the tasks they are undertaking. Rather than judging power draw based on the nature of your specific PS5 processor, a more general 'model SoC' is used instead. Think of it as a simulation of how the processor is likely to behave, and that same simulation is used at the heart of the power monitor within every PlayStation 5, ensuring consistency in every unit.
"The behaviour of all PS5s is the same," says Cerny. "If you play the same game and go to the same location in the game, it doesn't matter which custom chip you have and what its transistors are like. It doesn't matter if you put it in your stereo cabinet or your refrigerator, your PS5 will get the same frequencies for CPU and GPU as any other PS5."
"As for the details of the cooling solution, we're saving them for our teardown, I think you'll be quite happy with what the engineering team came up with."
- That power budget needs to be divided between the GPU and the CPU.
"The CPU and GPU each have a power budget, of course the GPU power budget is the larger of the two. If the CPU doesn't use its power budget - for example, if it is capped at 3.5GHz - then the unused portion of the budget goes to the GPU."
- The max clocks can be reached for both as long as that power budget is not exceeded.
if required they will run at their fastest clocks and yes that also means using it for calculations
- The actual power used by the system depends on the workload of the components.
- If the power budget is at risk of being exceeded, the clocks of the hardware that has the lower workload, be it the CPU or the GPU, is lowered in order to not exceed the max power budget.
- It is rare that both the CPU and GPU are at 100% workload at all times.
- The system cannot handle both the CPU and the GPU having max workload, but it can handle both of them having max clocks.
- Developers can choose to limit the workload on one to max out the clock speed of the other.
"Regarding locked profiles, we support those on our dev kits, it can be helpful not to have variable clocks when optimising. Released PS5 games always get boosted frequencies so that they can take advantage of the additional power."
Last edited:
FranXico
Member
Since when is marketing praise worthy though?
Sony is taking too long to show games, it's as simple as that.
Dory16
Banned
First of all I am not the only one who questions Cerny claims. I referenced several articles where equally respectable professionals raised the same contradictions. What defines a fanboy is not that they question what console makers say. It’s that they stick their heads in the sand, rant and name call when presented with rational arguments.
I dont think was necessary as again is misconception in your part, everything Cerny said was basically that he see better to have fewer CU more occupied than having lot of CU, that is true as an optimization but its not really that relevant to performance, actually it has a lot of physics involved XD is the classic example of having certain amount of workers, and doing the same work with half the workers as long as they work twice as fast, its more efficient but its the same work so is not that relevant performance wise, were is more relevant is the cost of the GPU as you can make more GPUs per slice if your GPU is smaller, you are basically making a storm in a glass of water
I recommend you leave your fanboy hat in house, you may not like that Cerny doesnt work in your favorite brand of consoles but he has an impressive record in this industry when it comes to extracting performance from consoles I am not saying he is perfect or take the absolute best design choices, but when it comes to discuss tech and how it works, if you start a discusion saying "Mark cerny contradict himself" and the reason you give is because you couldnt think of a good reason to put a clock control in a devkit for a console that can vary its clock, chances are you are not ready for that kind of discussions, better make question instead or arrogantly point fingers at professionals with far more experience than you
Nowhere did I say that I couldn’t think of a good reason to have clock control in a dev kit. Cerny said it makes sense for optimisation and I completely agree. Too bad he said a paragraph later that devs don’t need to optimise on his console. You consciously choose to be hookwinked and I don’t. It’s a simple as that.
Finally read what you said in bold. At least Cerny doesn’t maje completely irrational things up when he sells his product. He leaves that to soppy fanboys on forums. What is the purpose of optimisation if not to improve performance and how can something be relevant in optimisation but not in performance?
Respectfully, you’re a fraud sir.
polybius80
Member
I didnt see any quotes to several articles in your post, there was one article to extreme tech where they supposedly interviewed a dev from Oxide, Dan Baker and was quoted but it was taked out of context by the extremetech guy dan was speaking how small triangles can be more efficiently procesed in more CU opposed to less but faster CU
“Small triangles are indeed inefficient,” said Baker, “Because you have to partially shade fragments that are ultimately discarded. However, this inefficiency is largely in CU execution because the CUs are being asked to compute more work, so you’d want more CUs to offset the inefficiency.
but then Dan Baker mention that this is not a problem in defferred rendered engines and also mentions that most industry use them
“However,” he continued, “This is specific to the type of renderer. In deferred renderers, which make up most of the market today, most of the shading computation is done in screen space, where the small triangle problem is minimized. Only the material setup really pays the cost for small triangles. For Oxide’s decoupled shading rendering technology, neither the setup nor the shading efficiency is affected by the size of the triangle, so we are impacted even less.”
the guy from extremetech mention a vague small latency problem without mentioning where or how much and provides no quotes from Dan Baker to back it up or to check
sorry but that is not a contradiction to Cerny actually is an example of bad journalism which sadly is the norm in the industry
are you referring to this?
But what if developers aren't going to optimise specifically to PlayStation 5's power ceiling? I wondered whether there were 'worst case scenario' frequencies that developers could work around - an equivalent to the base clocks PC components have. "Developers don't need to optimise in any way; if necessary, the frequency will adjust to whatever actions the CPU and GPU are performing," Mark Cerny counters. "I think you're asking what happens if there is a piece of code intentionally written so that every transistor (or the maximum number of transistors possible) in the CPU and GPU flip on every cycle. That's a pretty abstract question, games aren't anywhere near that amount of power consumption. In fact, if such a piece of code were to run on existing consoles, the power consumption would be well out of the intended operating range and it's even possible that the console would go into thermal shutdown. PS5 would handle such an unrealistic piece of code more gracefully."
he is talking about optimization for the frequency adjust particular in PS5
Last edited:
FranXico
Member
No. He said developers don't need to worry about/do anything to account for the clock behavior of retail units when optimizing. Not that they don't need to optimize their engines at all.
No. You are the one trying to woodwink others.
Not the first time you misquote or distort what people say.
Last edited:
Dory16
Banned
I'm actually probably the only person on this forum that uses screenshots and links for every article that I quote. Not that I think it would deter a soldier like yourself to accuse me of distortion.
And later:
Any reader of good faith (so probably not you) should raise an eyebrow here and at least estimate that more coherent wording should have been used.
It's about credibility. Look up that word.
Dory16
Banned
You're ranting away from evidence as always. I am referring to the below and you know it very well:
They asked Dan Baker specifically about Cerny's claim that the smaller the GPU the better and they said the opposite (a larger GPU) is the solution to the problem.
Have a good life. You're too dishonest for my time.
FranXico
Member
Why paste the text or expect people to know how to read (which is what people do in good faith).I'm actually probably the only person on this forum that uses screenshots and links for every article that I quote. Not that I think it would deter a soldier like yourself to accuse me of distortion.
And later:
Any reader of good faith (so probably not you) should raise an eyebrow here and at least estimate that more coherent wording should have been used.
It's about credibility. Look up that word.
No, no, no. We can't have that. Instead, let's crop select sentences and highlight them out of context. "They don't need to optimize in any way" refers to accounting for underclocking, of course. Any intelligent person who read the entire interview understands that.
You are the one posting in bad faith and were proven to do so already before.
Your hipocrisy taints this very forum.
Dory16
Banned
And the goal of the fixed clock profiles still according to Cerny is... is precisely to get underclocking out of the equation when optimising.
Questioning your intelligence (which I have genuinely struggled to encounter at any point) is unnecessary. Cerny's self contradiction is so glaring that no intellect is required to acknowledge it. Just literacy and fair mindedness.
I honestly think that the assessments of what taints this forum should not be left in the hands of individual religiously and emotionally invested in the reactionary defense of a device. Also it's spelled hypocrisy but thanks for your input.
Last edited:
hyperbertha
Member
Its not about 'filling up' the memory its about clearing the memory to show new data in the next second of gameplay. 2.5 GB of new data doesn't equal 5 Gb.
Thirty7ven
Banned
It's working and exciting their fanbase. If it gets them excited and they feel engaged, that's gotta be worth something no?
pawel86ck
Banned
I wonder how many memory resources developers can spend for streaming if they only have 13GB available.
Streaming pool wasnt big in killzone shadow fall. Maybe PS5 games will change that, but developers will really use over 5 GB just for streaming? Maybe MS engineers have thought about that and came to the conclusion they will not need more 5GB (compressed) for streaming purposes considering XSX will only have 13GB available to developers.
Last edited:
hyperbertha
Member
There's no point having a big streaming pool in shadowfall as its streaming from an HDD at 50 mb/s. Its pointless. But when you have an SSD at 5 gb/s the bigger the streaming pool the better.
Xbox's streaming pool is about 5 gb and ps5's is about 8-9 gb after calculations. Both companies designed the SSD exactly around available streaming memory. And do you really think Cerny overlooked this critical factor in designing the SSD that amounts to the highest BOM in his console? Its designed exactly as per required.
pawel86ck
Banned
I can ask the same question. Do you really think MS engineers overlooked how many memory developers will be able to use for streaming purposes with only 13GB available to them?
Last edited:
Thirty7ven
Banned
Kumomeme
Member
yeppi another SSD thread!
joke aside, people should stop using current gen game as comparison how ssd will impacted the game..current gen game always designed with 50mb/s minimum speed..even spiderman is based on 20mb/s ...we need to compare with new game that created from ground up with minimum ssd speed in mind...or you wont get answer by looking at 1 sec different of loading time
also you wont see the different merely by graph or benchmark...you cant even measure it that way..even if there is number given by devs during gdc as example, its not something that everyone would understand...some benefit is during development..how you gonna to see it? we might be only can hear word from devs mouth but some people probably not gonna 'like' what they said....probably only way to see how much impact ssd had is by looking at game that being build up with it in mind..ssd supposed to improve world streaming, better lod, bigger world scale, seamless transition, no need to mask loading by tight section corridor, no/less pop in texture etc. so the impact will be translated through that way...just wait for next gen game and see ourselves how it will fare
same goes with cpu, some people argue that jaguar is enough for next gen game...we only need to see how next gen game A.I, physic, object perscreen and object destruction is compared to now...how to measure these things aside by looking at the game itself?
both console had ssd...xsx ssd surely not fast as ps5, but it already fast enough which is surely will benefit for next gen games
joke aside, people should stop using current gen game as comparison how ssd will impacted the game..current gen game always designed with 50mb/s minimum speed..even spiderman is based on 20mb/s ...we need to compare with new game that created from ground up with minimum ssd speed in mind...or you wont get answer by looking at 1 sec different of loading time
also you wont see the different merely by graph or benchmark...you cant even measure it that way..even if there is number given by devs during gdc as example, its not something that everyone would understand...some benefit is during development..how you gonna to see it? we might be only can hear word from devs mouth but some people probably not gonna 'like' what they said....probably only way to see how much impact ssd had is by looking at game that being build up with it in mind..ssd supposed to improve world streaming, better lod, bigger world scale, seamless transition, no need to mask loading by tight section corridor, no/less pop in texture etc. so the impact will be translated through that way...just wait for next gen game and see ourselves how it will fare
same goes with cpu, some people argue that jaguar is enough for next gen game...we only need to see how next gen game A.I, physic, object perscreen and object destruction is compared to now...how to measure these things aside by looking at the game itself?
both console had ssd...xsx ssd surely not fast as ps5, but it already fast enough which is surely will benefit for next gen games
Last edited:
pawel86ck
Banned
Only real games (and developers) can provide an answer to this question. If developers will end up using less or more than 5GB for streaming purposes we will know whitch company has smarter engineers. Right now we can only ask some obvious questions. For example we know for a fact GPU alocate the most memory for non streaming textures in order to use them right away, and streaming memory pool should be a lot smaller like Killzone memory usage analysis shows.
Last edited: