• 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.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.
D

Deleted member 775630

Unconfirmed Member
Its amazing how many Xbox fans post religously about sustained frequencies and trying to understand. Its relentless.
I'm trying to properly understand how something works. Not just quote other people, but really understand how it exactly works so that it makes sense. I'm happy Crimson_Fate Crimson_Fate is taking the time to answer the questions I have so I can understand the differences of this approach and the one Microsoft took. Isn't that the whole idea of this topic? To discuss next-gen consoleS. There's a specific PS5 topic if you just want to discuss that and not discuss the technical aspects and differences of all consoles.

The Power budget is fixed such that heat is no longer a problem.

Think of it as having a Large cup of water and two smaller cups of water. You can't fill them both at the same time (LOAD) but you can move water from cup A to cup B . But you always have a fixed amount of water.
But Cerney said you actually could fill them up both, right? If they wanted, the console could run both the CPU and GPU at max clocks, this just never happens that often. Anyway thank you for explaining this in more detail with these lovely metaphors. :)
 

McHuj

Member
PS5 Pro is more than likely to happen, with the same butterfly chiplet or stacked this time of the same die. It's AMD's philosophy as well to have 36-40CU dies max and instead stack them, something makes PS5 Pro more predictable, and should be sold as much as release PS5 2020 but in 2023-2024. And it'll not use 5nm nor 3nm, it'll use the same one to gain more value out of it.

Sony would rather push for 8K gaming (dynamic) much faster as it benefits other branches (Sony TV's). And before anyone whines about 8K, 4K should hit 60-240fps more often to hit 8K@30-60fps (or reconstructed from 6K like that indie game that's been published with PS5 reveal). Should make everyone happy, if they're willing to upgrade instead of staying at base console and bitch about 30fps. :messenger_winking_tongue:


I mean, I'd buy it. I don't think many people would as I don't think they would be able to hit the same price as a launch PS5.

I think a pro that doubled the performance would be very expensive and power hungry. Based on the size, I think the PS5 will be coming in easily 200W+.

A 5nm shrink will only allow for a 30% power savings, so would Sony be willing to release a 300W Pro console? Waiting for a 3nm process that would bring down the power consumption at reasonable prices, I think will put us out to a 2025 date.
 

DrDamn

Member
The way I am reading this is that if you use variable frequencies you can go to a higher clock because you don't need to sustain that for a long time, so for short bursts you can go higher if the game needs this. The PS5 would never be able to keep the clocks this high for the CPU and GPU for a long time, right? Due to heating.

But this would also mean that if a 3rd party developer would be able to get everything out of the XSX with their more powerful GPU and CPU (barely), the game would need to be dumbed down for PS5 because it can't sustain those high clocks for a very long time which would be needed to follow the XSX. Right?

Heating is not just about clocks - it's about load, what those clocks are doing - i.e. power use. A fixed clock does not use a fixed amount of power. So this is why Cerny says the clocks can both run at the advertised speed at the same time *and* for the majority of the time. That's because most games won't put the full power load on the APU.

A need to reduce the clock is also not about time - i.e. how long it is under a specific load - it's just about power.

I don't see what this has to do with comparison to the more powerful APU of the XSX? Just compare to a PS5 without this technology. It's a system for maximising efficiency and performance of the PS5. But on that specific point again it can maintain high clocks, it's not about clocks, it's about power. As long as the load does not draw too much power then it maintains them. This is why it is predictable and consistent. External factors like ambient temperature do not effect the performance you will get.

So variable clocks is great mostly for first party games and 3rd party games that don't push the hardware too far?

No, it allows developers to push individual elements further than they would have been able to with fixed frequencies. It's a benefit to performance and efficiency.
 

geordiemp

Member
I'm trying to properly understand how something works. Not just quote other people, but really understand how it exactly works so that it makes sense. I'm happy Crimson_Fate Crimson_Fate is taking the time to answer the questions I have so I can understand the differences of this approach and the one Microsoft took. Isn't that the whole idea of this topic? To discuss next-gen consoleS. There's a specific PS5 topic if you just want to discuss that and not discuss the technical aspects and differences of all consoles.


But Cerney said you actually could fill them up both, right? If they wanted, the console could run both the CPU and GPU at max clocks, this just never happens that often. Anyway thank you for explaining this in more detail with these lovely metaphors. :)

No, Cerny said he expects max clocks for CPU and GPU most of the time, he did not say max frequencies does not happen often.

Go watch it again if you are genuinely interested, he gave the example of a map screen. Simple geometry heats up GPU in a loop on a static screen - so why do it ?.

If you want to predict that ps5 wont use the full 10 TF for map screens, now you understand.
 
Last edited:

IntentionalPun

Ask me about my wife's perfect butthole
No, Cerny said he expects max clocks for CPU and GPU most of the time, he did not say max frequencies does not happen often.

He said at or NEAR most of the time.

Which IMO.. implies one or the other is not maxed most of the time.

Which isn't some huge deal; he also implied it was mostly very small drops as a small drop in frequency saves quite a bit of power.
 
Last edited:
I'm trying to properly understand how something works. Not just quote other people, but really understand how it exactly works so that it makes sense. I'm happy Crimson_Fate Crimson_Fate is taking the time to answer the questions I have so I can understand the differences of this approach and the one Microsoft took. Isn't that the whole idea of this topic? To discuss next-gen consoleS. There's a specific PS5 topic if you just want to discuss that and not discuss the technical aspects and differences of all consoles.


But Cerney said you actually could fill them up both, right? If they wanted, the console could run both the CPU and GPU at max clocks, this just never happens that often. Anyway thank you for explaining this in more detail with these lovely metaphors. :)


Remember running at clock is not the same as running loaded at clock. so the CPU would need to have all 8 cores doing work 100% of the time to be using the max power
 

geordiemp

Member
He said at or NEAR most of the time.

Which IMO.. implies one or the other is not maxed most of the time.

And if you understand how a frame works, CPU and GPU are taxed at different times in a frame, CPU does work, GPU renders it and repeat. Clocks sensible with workload in a frame makes sense.

The important point is will it run normal gameplay code any slower, and the answer I believe is NO.

Will ps5 downclock on HZD map screen - I hope so. Maybe everytime the screen is static ps5 has a rest.....:messenger_beaming:
 
Last edited:

DrDamn

Member
But Cerney said you actually could fill them up both, right? If they wanted, the console could run both the CPU and GPU at max clocks, this just never happens that often. Anyway thank you for explaining this in more detail with these lovely metaphors. :)

Don't equate power to clock speed, that's the first step to understanding. It's obviously related but power is also linked to the load not just the speed.

So if your cooling system can handle X amount of power you can either consider forcing your clock speeds to be a set number where you know under full load it will keep within those limits, or you let it go faster but not under full load. Running faster but not under full load is beneficial (it's faster!) and it also covers the majority of use cases too - games rarely run the CPU/GPU under full load. This leaves you to balance power between CPU/GPU as required - it's added flexibility.
 

Thirty7ven

Banned
He said at or NEAR most of the time.

Which IMO.. implies one or the other is not maxed most of the time.

Which isn't some huge deal; he also implied it was mostly very small drops as a small drop in frequency saves quite a bit of power.

The most important bit is that instead of restraining all CPU/GPU workloads based on edge cases, they only restrain edge cases. These happen in singular scenes/moments. The frequency can change in milliseconds.
 

IntentionalPun

Ask me about my wife's perfect butthole
The most important bit is that instead of restraining all CPU/GPU workloads based on edge cases, they only restrain edge cases. These happen in singular scenes/moments. The frequency can change in milliseconds.
How is it an "edge case" if it is not at MAX most of the time?
 

DrDamn

Member
Variable performance is variable performance regardless of what ever round about way you try to down play it.

You are looking at it from the negative perspective though, variable down. From the perspective of the same APU with the same cooling at fixed clocks it is variable upward.

So they have the choice of a fixed profile where they can more consistently know the performance of the components. Easier to optimise for but leaving potential performance on the table. Or, they can put in more work to optimise for the variable performance which, as indicated above, is automatically more difficult than doing so for a fixed performance set up.

The performance is predictable and consistent for a given game though. Put it this way fixed clocks would still have moments of spikes in CPU or GPU load. A variable clock approach can shift some power to where it's needed and manage those spikes in one or the other better no? Then where both are over loaded you have the option to optimise the CPU or GPU processes.

It's a different approach to what's come before, but it's about efficiency and an elegant solution to getting higher performance for your power budget.
 

IntentionalPun

Ask me about my wife's perfect butthole
Well edge cases for the purpose of this discussion and indeed for developers are cases where the frequencies have a noticeable drop, and not just an almost academic one that has next to no impact on performance.
Ah gotcha.

Locked designs are set up to go above "normal power" though. Those edge cases aren't sustained, so they are setup to handle brief surges up to a point.

So during "edge cases" it's likely a locked design benefits that code.

It's a give and take really.
 

IntentionalPun

Ask me about my wife's perfect butthole
The performance is predictable and consistent for a given game though..
Not exactly; only a given workload.

So only really predictable for really predictable/linear games... you aren't going to get predictable clocks with an open world game where what's happening on screen has a huge variety.

Although you also don't get predictable performance on a locked design for those games; but the clocks won't lower.
 

Brudda26

Member
The way I am reading this is that if you use variable frequencies you can go to a higher clock because you don't need to sustain that for a long time, so for short bursts you can go higher if the game needs this. The PS5 would never be able to keep the clocks this high for the CPU and GPU for a long time, right? Due to heating.

But this would also mean that if a 3rd party developer would be able to get everything out of the XSX with their more powerful GPU and CPU (barely), the game would need to be dumbed down for PS5 because it can't sustain those high clocks for a very long time which would be needed to follow the XSX. Right?

So variable clocks is great mostly for first party games and 3rd party games that don't push the hardware too far?
That's if the PS5 was throttled by heat but as per cerny talks that's not the case. The cooling system is designed by the maximum heat output and instead the clocks are adjusted by the available power at hand. The console can run its maximum clocks at all times, but if a job is more intensive and exceed the power budget clocks will be lowered by a small percentage. Basically at fixed clocks the console will be at those clocks no matter the job even if it doesnt need it generating unwanted heat and can cause the fan to ramp up for no reason. That already happens on current gen and why some ps4 and xbox consoles fans go absolutely bat shit crazy for no reason in areas you perceive as not demanding. The variable clocks on ps5 which arent the traditional variable clocks you find on PC. Basically manage what's needed for the job at hand therefore managing power limits. You can have a 2 jobs both at 2.23ghz but one job can be less power intensive. So that then unwanted power for the job can be diverted to say the CPU if its needed to help a job on its side.

They both have advantages and disadvantages.

Your more likely to have excess heat generated with fixed clocks as it the jobs could cause it to exceed power budget on both CPU and GPU even if the jobs dont need the power as it cant manage it on the fly.
 

Neo Blaster

Member
I don't think anyone here was claiming the PS5 was going to outperform the Series X just because it has higher clocks

Genius up top was claiming fixed clocks are objectively better than variable clocks. Which is obviously not true since a PS5 with fixed clocks and its limitations wouldn't perform equally or better than the PS5 with its variable clocks
Gosh, even before Cerny's talk we've been hearing these machines are very close, but somewhat Xbox fans hear that and in some weird way think everyone is saying PS5 is more powerful and will crush XSX.
 
T

Three Jackdaws

Unconfirmed Member
It is too early to talk about mid gen options, especially when it will come out that PS5 can effectively perform as fast and strong as XSX even with apparent power deficit that will silence xbox fans, what would be the reason to release a powerful console if it still targets 4K?
Yeah, only reason PS4 Pro and One X were released was because there was a high demand for 4K gaming, I don't see the purpose of a mid-gen refresh for the PS5 or Series X, both consoles will be more than capable of 4K gaming and I don't see 8K gaming on the horizon anytime soon.

Sony doesn't need to release a mid-gen refresh either, just to meet a small deficit in compute performance, doesn't make sense to me. We have true generational leaps with next-gen consoles, including the CPU/GPU/SSD.

But this is just my 2 cents, we'll have to wait and see what happens.
 

Thirty7ven

Banned
It is too early to talk about mid gen options, especially when it will come out that PS5 can effectively perform as fast and strong as XSX even with apparent power deficit that will silence xbox fans, what would be the reason to release a powerful console if it still targets 4K?

Indeed. Penello is right either way, and midgen refreshes would need either a breakthrough like 3D Stacking or a more specialized approach to silicon like custom processors for RT and A.I.

Which brings me to another thing, do we know if the PS5 is a chiplet design? I ask this because 5 months ago there was this thread on reddit:


Sony’s Next Generation PS5 contains world’s first “chip-let” based APU. This process splits the traditional “monolithic” APU into four (4) tightly components. Resulting in more performance: reticle limit on the silicon package, inverse-exponential reduction in yield with die size, and more flexibility due to silicone configuration.

Proposed specification on the chip-let design:

  • 10.3 Teraflops (Navi)
  • 40 compute units (8 disabled / 48 total)
  • Clocked at 2Ghz (2010Mhz) due to reduction in yield with die size
  • 16GB GDDR6 + 4GB DDR4 (background task)
  • 512GB/s memory bandwidth
  • Hardware-based ray tracing & VRS (Variable Rate Shading)
  • Hybrid SSD at 5-7GB/s
  • Price & launch date: Unknown

This was pre road to PS5. It's speculated elsewhere too, and I wonder if it's true and if so what it means for a possible PS5 midgen refresh.
 
Indeed. Penello is right either way, and midgen refreshes would need either a breakthrough like 3D Stacking or a more specialized approach to silicon like custom processors for RT and A.I.

Which brings me to another thing, do we know if the PS5 is a chiplet design? I ask this because 5 months ago there was this thread on reddit:




This was pre road to PS5. It's speculated elsewhere too, and I wonder if it's true and if so what it means for a possible PS5 midgen refresh.
That is very close to actual PS5 specs... maybe the chiplet design will allow Sony to make PS5 cheaper than expected?
 

Neo Blaster

Member
I did not see any retail units out there for any demo. The dev kits runs at fixed clocks so we haven't seen real world performance yet. Epic has optimized like hell for the PS5 and ignoring all other platforms. The PS5 got its own io rewrite ffs.
I remember reading about a devkit version indeed running with fixed clocks, that being the reason why we heard about heating issues. However, later versions had variable frequency and ran much cooler.
 

Dodkrake

Banned
I am not moving any goal posts. You posted an article about a variable frequency technique with the intention to conserve power, e.g. downclock your CPU if it is not needed. It is just something that is completely different; in how it works and in what it is intended to do. And obviously neither of these techniques is bad, as it does exactly what it intends to do: In this case reduce power consumption.

The only similarity to the PS5's system is that both techniques can adjust CPU/GPU frequency somehow. It is like saying everything that produces energy is a nuclear plant and is bad for the environment.

It would have helped if you read the implementations

2nKJdCq.png


Namely, AMD's Cool'n'Quiet for CPUs and AMD PowerTune.

Sony's implementation is an hybrid of existing implementations by both AMD and Intel.
 

DrDamn

Member
Not exactly; only a given workload.

So only really predictable for really predictable/linear games... you aren't going to get predictable clocks with an open world game where what's happening on screen has a huge variety.

Although you also don't get predictable performance on a locked design for those games; but the clocks won't lower.

Yes, sorry my point was that there isn't unpredictability between individual PS5s or depending on external conditions.

Even in the situation of an unpredictable game though the performance will be more consistent than fixed lower clocks, because the variance can manage spikes in GPU or CPU load better.
 

DrDamn

Member
Variable frequencies are worse, which is why every desktop and laptop grade CPU uses them to some capacity.

Just checking - is that phrased correctly? Variable frequencies are worse so "every desktop and laptop grade CPU uses them to some capacity"?

Ah sarcasm ... sorry, this is why you are getting the replies you are getting. People are misunderstanding your intent and point.
 
Last edited:
2 power skus upfront by Microsoft rule out yet another sku mid gen I agree. But nothing here says no PS5 Pro. Indeed, I’d bet on it.

Its more than likely PS put out a Pro version because of the advancements to their I/O tech & register their patents. Plus probably aiding in their collab with AMD.

I rather we get a new PSP with updated PS NOW that has back compat of PS1-4 games. I would eat that up.
 

Handy Fake

Member
And if you understand how a frame works, CPU and GPU are taxed at different times in a frame, CPU does work, GPU renders it and repeat. Clocks sensible with workload in a frame makes sense.

The important point is will it run normal gameplay code any slower, and the answer I believe is NO.

Will ps5 downclock on HZD map screen - I hope so. Maybe everytime the screen is static ps5 has a rest.....:messenger_beaming:
Like wiping your brow between pushes during an excessively painful shit.
 

DrDamn

Member
Namely, AMD's Cool'n'Quiet for CPUs and AMD PowerTune.

Sony's implementation is an hybrid of existing implementations by both AMD and Intel.

Regarding AMD PowerTune - interesting article linked from Wikipedia - https://semiaccurate.com/2013/12/16/amds-powertune-2-0/

Seven years ago but still on point ...

A lot of people have a fundamental misunderstanding of AMD’s PowerTune 2.0 which is unfortunate because the tech is pretty simple. The biggest problem is that people don’t even try to understand what it does or why.

The basic premise is that fixed clocks speeds are a thing of the past, if they are not pointless at the moment they will be soon. Almost no one out there actually understands why this is a necessity much less how it functions, then work themselves up in to a frothy lather debating misunderstandings.
 
D

Deleted member 775630

Unconfirmed Member
Remember running at clock is not the same as running loaded at clock. so the CPU would need to have all 8 cores doing work 100% of the time to be using the max power
Don't equate power to clock speed, that's the first step to understanding. It's obviously related but power is also linked to the load not just the speed.

So if your cooling system can handle X amount of power you can either consider forcing your clock speeds to be a set number where you know under full load it will keep within those limits, or you let it go faster but not under full load. Running faster but not under full load is beneficial (it's faster!) and it also covers the majority of use cases too - games rarely run the CPU/GPU under full load. This leaves you to balance power between CPU/GPU as required - it's added flexibility.
Got it, thanks guys!
 

kensama

Member
<5% attach rate even with heavy discounts. With all of the hype around VR during 2015-2017 it was expected to sell a lot better than it has, especially considering the cost of R&D.

And yet House declared that PS VR was sold beyond Sony expectation during the first months.
I hope they don't expect a 50% ration 5% is already a great job.
And we have this:

 
Last edited:

Thirty7ven

Banned
And yet House declared that PS VR was sold beyond Sony expectation during the first months.
I hope they don't expect a 50% ration 5% is already a great job.
And we have this:


Kinect sold over 30 million units, for comparison.

The kind of bet Sony is placing on VR, from hardware to software, is backed by the expectation that VR is a market waiting to burst. It's nonsense really, and VR in gaming isn't going to happen in any significant fashion for at least the next ten years.

Half Alyx happened and the market barely felt it.
 

geordiemp

Member
I think the PS5 will be whatever TF count that's required by a game. Why use the full 10.28 TF count on a simple 2D side scroller? You definitely don't need that kind of power for a game like Cuphead.

Power is by definition a maximum, think of it like horse power on your car. Its available power when needed. TF will be what it needs to be every nanosecond.

As long as the CPU and GPU meet the workload in the frame time, they can chill.

The whole point is IF Ps5 has the 10.28 TF when it needs that power in a frame, that is all that matters.
 
Last edited:

kensama

Member
Kinect sold over 30 million units, for comparison.

The kind of bet Sony is placing on VR, from hardware to software, is backed by the expectation that VR is a market waiting to burst. It's nonsense really, and VR in gaming isn't going to happen in any significant fashion for at least the next ten years.

Half Alyx happened and the market barely felt it.

Kinect and VR are not comparable.
Plus Kinect take advantage of Wii Hype and was not as costly as VR for consumer.
 
<5% attach rate even with heavy discounts. With all of the hype around VR during 2015-2017 it was expected to sell a lot better than it has, especially considering the cost of R&D.

It's one thing to play armchair business analyst, it's another to to make absolute statements based on literally no facts. What were the expectations how much (if any) did it fall to be considered a failure despite a successor being produced? What was the cost of R&D and how much revenue or profit does Sony expect in future to mitigate these costs?

Kinect sold over 30 million units, for comparison.

A Honda Civic sells ten times more than a Ferarri. It's almost as if there is different margins and expectations for certain products.

It's almost as dumb bringing up the Nintendo DS, which sold 3x much as an Xbox One as some type measurement of what success or failure is. "But you can't compare a handheld that cost significantly less than a home console!" Yeah, how about camera sensor that costs three times less than a VR unit?
 
Last edited:

Bo_Hazem

Banned
I mean, I'd buy it. I don't think many people would as I don't think they would be able to hit the same price as a launch PS5.

I think a pro that doubled the performance would be very expensive and power hungry. Based on the size, I think the PS5 will be coming in easily 200W+.

A 5nm shrink will only allow for a 30% power savings, so would Sony be willing to release a 300W Pro console? Waiting for a 3nm process that would bring down the power consumption at reasonable prices, I think will put us out to a 2025 date.

The PS4 Pro formula worked perfectly for Sony, and it won't be very expensive at all, and $500 is easily achievable by 2023-2024 with 36+36CU chiplet. Making an all-new apu/die is money wasting and you'll make it more complicated. PS4 Pro and the Slim are using nearly the same (doubled) die so the same would make base PS5 even more profitable over time just like PS4 now.
 

Thirty7ven

Banned
It was 150$, Kinect wasn't exactly cheap. But if you wanna be stingy about that, go on right ahead. There's a lot more investment being put into VR, a lot more software is available and it still isn't getting any sort of traction that would lead anyone to believe it's going to be a big hit anytime soon.

PS VR sold little over 5 million units in three years. This isn't a success and if you think it is, well ok that's cool.
 

DaGwaphics

Member


I would expect MS to have new hardware in 3 or 4 years time. However, probably not a mid-gen refresh in the same form as the 1x. They will probably just roll forward with a new iteration of hardware that will replace the existing hardware. If they are serious about ending generations, that's the way you'd need to do it. Each hardware iteration would be supported for the same 6 or 8 years as it has been, but without the complete user base restarts. You'd only lose half the install base worse case scenario, and you've had 3 or 4 years to sell a new console to your users before then.
 

ToadMan

Member
This I don't understand, why do you say this? If the PS5 would have fixed clocks at their current max variable clocks why would it have lower performance?

The PS5 couldn't achieve such high clocks if it used a traditional fixed clock design. It wouldn't be practical for Sony to deliver a consumer electronic device with fixed clocks at these speeds using current silicon technology.

MS chose a clock more in line with standard AMD GPU specs - they leave a power/cooling/performance buffer and that buffer of performance remains largely unused.

Sony made extra work for themselves to innovate the PS5 design to introduce system power control and variable clocks. This means Sony can allow devs to code to whatever the system will render without worrying about finding in subsequent QA their code will cause systems to fail. That allows devs to extract all the power from the PS5 - no more considering power drain or temperature issues overloading the cooling system by running above TDP for longer than they're supposed to.

Sony could just have built a PS4Pro+ but instead they chose to try and get more performance through innovation.
 

FeiRR

Banned
2 power skus upfront by Microsoft rule out yet another sku mid gen I agree. But nothing here says no PS5 Pro. Indeed, I’d bet on it.
We don't even know if Lockhart is a console. Maybe it's a Surface aimed at gamers? When I read this tweet, I noticed weird wording. Sounds like "Xbox for PC", which is an app in Windows. I think Microsoft can release new hardware more often because they want to maximize subs.

 

kensama

Member
It was 150$, Kinect wasn't exactly cheap. But if you wanna be stingy about that, go on right ahead. There's a lot more investment being put into VR, a lot more software is available and it still isn't getting any sort of traction that would lead anyone to believe it's going to be a big hit anytime soon.

PS VR sold little over 5 million units in three years. This isn't a success and if you think it is, well ok that's cool.


150$ compared to 400$ there is a large gap in term of price.
I never said it was a success i said what house said at launch of PS VR. Even Sony seems not confident on VR but with 950k of PS VR sold in the first months it surpassed Sony Expectation and sold more than other competitor.
For me it's not failure except if you have big ratio expectation.
 
Status
Not open for further replies.
Top Bottom