PlayStation 5 System Architecture Deep Dive |OT| Secret Agent Cerny

Liabe Brave · Mar 27, 2020

Axel Stone said:
It is helpful, but this system doesn't eliminate fast max, you still can't go above 2.23ghz clock speeds, it just shifts it to a different clock speed that's only obtainable by reducing power elsewhere, no?

No. The point is it makes "fast max" and "heavy max" the same power draw, by dynamically reducing clockspeed in the former, where it doesn't matter. So no matter what power/cooling system you use, all of it provisions gameplay. Unlike fixed clocks, where a significant wattage must be reserved for basically no benefit.

Axel Stone said:
So this is what I don't get. If these drops are so unlikely, why do you need room to drop the clocks at all? Is that last 2% of performance really that important?

The benefit is more than 2%, which is just the raw TF gap. But you're also speeding up cache, the scheduler, texel and pixel fillrates, etc. The overall gain is still not huge, but why avoid grasping it? In edge cases, even small percentages matter: no tearing instead of minimal tearing; middle-distance enemies at full instead of half framerate; lots of particles instead of few particles.

anexanhume said:
If you take the difference between Arden and Navi 10, and then scale that based on CU count, it would suggest denser CUs and a 40CU chip could be as low as 293 mm^2.

My rough estimates appear to show that all elements are smaller in RDNA2, not juat CUs (PHY, I/O, cache, etc).

But this isn't certain, and I expect Sony may be a little bigger than 300mm^2. This is a good ballpark, though.

amstradcpc said:
I dont buy the 36 CUs was neccesary for BC at all, is not like the own PS4 pro deativate 16 CUs for doing it.

Yes it does (though the number of deactivated CUs is 18, not 16).

anexanhume said:
Yes, they deactivate when Boost Mode is not enabled.

Actually, even Boost Mode doesn't turn on the other CUs. It just causes the active 18 to run at full Pro clock, rather than PS4 clock.This is why the framerate bumps are relatively modest.

RCU005 said:
I don't think so. Microsoft said absolutely everything about the specs of the XSX, but they never mentioned SmartShift. IF they said everything else, why keep that to themselves? It's very likely the XSX doesn't have it.

Just because a platform holder hasn't mentioned something doesn't mean they don't have it. But in this case you're right they might not. Of course they can't balance CPU and GPU load by changing clocks. But I was thinking they could do so based on draw: if Zen2 is running cool, send more voltage to Navi to ensure more stable performance. But I forgot about Microsoft's Hovis Method of power matching. Their chips should already being fed the optimal voltage.

Axel Stone said:
Seems like the easier way to fix that would be to use a better cooling solution.

Easier perhaps, but it seems not affordable. Microsoft went all-out on cooling the XSX, with a novel airflow design, a huge fan and exhaust surface, and a vapor chamber with massive heatsink. And their clockspeed is notably slower anyway. Their SOC is bigger than Sony's, yes, but by a smaller percentage than the clock gap, even though Sony is necessarily also farther up the inefficiency curve.

Axel Stone said:
Fans that go crazy only seem to be an issue for Sony.

You're forgetting Xbox 360, the biggest overheating disaster consoles have ever seen.

thuway · Mar 27, 2020

Dave. said:
Brad Sams spewing misinformation, just so happens to pains the PS5 in a poor light...

When do TFlops Matter? - Thurrott.com

While Microsoft and Sony are taking similar paths with similar technology, there is a lot of comparison to previous consoles as well.

www.thurrott.com

This needs to be called out.

Betelgeuse · Mar 27, 2020

thuway said:
This needs to be called out.

He's consistently demonstrated a facile understanding of technology, but this is a new low. This especially:

This really isn't that big of an issue and it's a minor tradeoff to having higher sustained performance than boost performance. The PlayStation has the ability to push 10.3 TFlops of output but will typically operate in a state that is around 9.2 TFlops; the Xbox will operate consistently at 12 TFlops with no deviation in performance.

Put another way, PlayStation developers should be targeting games with the performance characteristics of a 9.2TFlop console and if a scene needs some extra muscle to keep frames stable, the console can briefly kick up the horsepower. Xbox developers will be targeting an environment with 12Tflops of performance and in the event that a scene drags frames down, they will need to either optimize the loading of assets or scale other features as there is no boost performance available.

"Boost" is a common feature in the CPU world, Intel has been using 'burst' that can overclock a CPU for increase performance and is a proven tactic for edging out a little bit more compute when needed. But, it's not sustainable for running for lengthy periods of time, otherwise, that would be the default clock speed.

is straight up false reporting.

Bunkles · Mar 27, 2020

Dave. said:
Brad Sams spewing misinformation, just so happens to pains the PS5 in a poor light...

When do TFlops Matter? - Thurrott.com

While Microsoft and Sony are taking similar paths with similar technology, there is a lot of comparison to previous consoles as well.

www.thurrott.com

AudiophileRS · Mar 27, 2020

cyenz said:
We don't know how many P states the consoles have, I doubt Xbox will run full power and speed for smaller indie games or watching media.

I agree, but that's not really the point I'm getting at. That's a matter of power efficiency in regards to eco-friendliness; as in, if you don't need the power, don't use it. I'm referring to power efficiency in regards to maximising performance; as in, if the power is available, can devs use it if they want to?

I'm simply stating that Sony's solution effectively makes all of its power available for every potential workload; this in turn allows them to maximise performance for each workload.

MS' conventional solution determines maximum power based on less common, edge case scenarios. By maintaining the same clocks with more common, less power hungry scenarios there is power left on the table which is not being used to maximise performance.

vivftp · Mar 27, 2020

Dave. said:
Brad Sams spewing misinformation, just so happens to pains the PS5 in a poor light...

When do TFlops Matter? - Thurrott.com

While Microsoft and Sony are taking similar paths with similar technology, there is a lot of comparison to previous consoles as well.

www.thurrott.com

I would expect better from him. bdsams would you like to comment on this?

KeRaSh · Mar 27, 2020

Betelgeuse said:
He's consistently demonstrated a facile understanding of technology, but this is a new low. This especially:

is straight up false reporting.

That's pretty embarrassing...

RingRang · Mar 27, 2020

Dave. said:
Brad Sams spewing misinformation, just so happens to pains the PS5 in a poor light...

When do TFlops Matter? - Thurrott.com

While Microsoft and Sony are taking similar paths with similar technology, there is a lot of comparison to previous consoles as well.

www.thurrott.com

What is the misinformation?

Betelgeuse said:
He's consistently demonstrated a facile understanding of technology, but this is a new low. This especially:

is straight up false reporting.

So what exactly is the baseline PS5 clock speed?

Brees2Thomas · Mar 27, 2020

vivftp said:
I would expect better from him. bdsams would you like to comment on this?

Deleted member 6539 · Mar 27, 2020

Betelgeuse said:
He's consistently demonstrated a facile understanding of technology, but this is a new low. This especially:

is straight up false reporting.

I don't understand how people can just blatantly lie like that and still expect to be taken seriously.

nib95 · Mar 27, 2020

RingRang said:
What is the misinformation?

So what exactly is the baseline PS5 clock speed?

Firstly he claims that that the PS5 will typically operate at 9.2 Tflops, presumably a false conjured figure he's decided on based on the Github leak, one that happens to be in direct contradiction with what Cerny himself stated.

"36 CUs at 2.23Ghz is 10.3 Tflops and we expect the GPU to spend most of its time at our close to that frequency and performance.

Similarly running the CPU at 3 GHz was causing headaches with the old strategy but now we can run it as high as 3.5 GHz, infact it spends most of its time at that frequency"

Cerny also says this.

"When that worst case arrives, it will run at a lower clockspeed, but not too much lower. To reduce power by 10%, it only takes a couple of percent reduction in frequency, so I'd expect any downclocking to be pretty minor"

So based on the above, we're told that both the CPU and GPU will spend most of their time at the subsequent max frequency clocks, and that only worst case power load scenarios will require a GPU downclock.

Brad compares the PS5's boost with other product boosts like CPU's etc, showing a fundamental misunderstanding of Sony's frequency clock variability, and how it differs from other competing boost implementations in other GPU's, CPU's etc. Eg that it isn't overclocked temporarily, and then throttled down based on thermal dynamics, but instead the PS5's CPU/GPU spend "most" of their time at their max clocks, and only need to downclock a tiny amount in a "worst case" scenario (a 2% downclock for 10% power back, such a drop would keep the GPU at 10 Tflops) not when a thermal threshold is reached, but instead a set power load limit.

Sekiro · Mar 27, 2020

Betelgeuse said:
He's consistently demonstrated a facile understanding of technology, but this is a new low. This especially:

is straight up false reporting.

Wow, i didn't know he had an exclusive interview with Cerny himself getting a heads up on how the variable clocks really work and also confirming that 9.2TF is real, that's amazing...

Jesus that is some shill behaviour right there, wasn't this guy suppose to be a professional or something? Kinda sad.

AudiophileRS · Mar 27, 2020

Betelgeuse said:
He's consistently demonstrated a facile understanding of technology, but this is a new low. This especially:

is straight up false reporting.

I think with a lot of these guys it's a result of the misconception that: all processors running at full clocks = operating at maximum teraflops. When in reality that's just the theoretical ceiling with a certain operation.

However, what takes it from misunderstanding to something that resembles either misinformation or poor journalism is when he says "the console can briefly kick up the horsepower", which is in direct contention what Cerny has stated. If anything, it seems it's the reverse.

As someone under contract with Sony and effectively acting as a representative of them, Cerny will no doubt have to tow the company line and frame things in as positive a light as possible. But, he's yet to set any precedent for outright lies.

Betelgeuse · Mar 27, 2020

RingRang said:
What is the misinformation?

So what exactly is the baseline PS5 clock speed?

3.5 GHz for the CPU, 2.23 GHz for the GPU.

I know you're thinking "no, I'm asking how low does it go", but I'm saying the above numbers are what make up the baseline. That's because it's where the PS5 sits, and only downclocks from there in certain limited circumstances - i.e., only specific workloads - and for limited durations.

It's really important to understand this distinction from traditional boost clocks, which is how Sams has erroneously framed this. PS5 doesn't sit at a lower clock/TF figure and only boost thereabove for limited durations. It's the opposite. The fact he phrases PS5's thermal management in terms of traditional PC-style boost clocks shows he fundamentally misunderstands how this all works.

In other words, PS5's baseline is its max clocks. It downclocks from there when needed. In the PC space, the baseline is a lower clock, with brief upclocks happening until a thermal limit is reached, at which point clocks fall back down to the baseline.

The 9.2 TF figure is misleading as well, if we are to take Cerny's remarks at face value. He said "to reduce power by 10 per cent it only takes a couple of percent reduction in frequency". Let's say that's a 3% downclock - that's still good for 10 TF. So unless Cerny is really downplaying the extent of downclocking that will be necessary to remain within the TDP budget, 9.2 TF is not likely to be reached in real world scenarios.

For further reading, I'd defer to Liabe Brave's posts. He has a far more thorough understanding about all of this than I.

Bunkles · Mar 27, 2020

Sams also makes it seem like the flops between PS5 and Series X is a big deal but we should ignore the flops number on Lockhart because it's not a big deal. Pretty hilarious how he spun this article.

Adookah · Mar 27, 2020

Brad Sams is an Xbox fanboy and he knows shit about gaming hardware.

jroc74 · Mar 27, 2020

Foorbits said:
Sams also makes it seem like the flops between PS5 and Series X is a big deal but we should ignore the flops number on Lockhart because it's not a big deal. Pretty hilarious how he spun this article.

See, Lockhart is different tho.

It's comical how easy it is to spot biased folks. New gen must be around the corner.

Is he basing it off github? Pretty sure the PS5 sits above 2.0...

Like an earlier post said, the baseline is what Cerny said, not lower.

Xeontech · Mar 27, 2020

That dude is a just another bitch ass ingnorant.

Watch the media take his words for gospel though.

nib95 · Mar 27, 2020

Foorbits said:
Sams also makes it seem like the flops between PS5 and Series X is a big deal but we should ignore the flops number on Lockhart because it's not a big deal. Pretty hilarious how he spun this article.

The amusing thing about that, is that we don't actually know the prices of these systems yet. If the PS5 does come in cheaper, it'll further diminish the credibility of that lopsided comparison, because the justification for Lockharts much lower performance is a more affordable price.

anexanhume · Mar 27, 2020

Xeontech said:
That dude is a just another bitch ass ingnorant.

Watch the media take his words for gospel though.

I know we can do better than this.

RingRang · Mar 27, 2020

nib95 said:
Firstly he claims that that the PS5 will typically operate at 9.2 Tflops, presumably a false conjured figure he's decided on based on the Github leak, one that happens to be in direct contradiction with what Cerny himself stated.

"36 CUs at 2.23Ghz is 10.3 Tflops and we expect the GPU to spend most of its time at our close to that frequency and performance.

Similarly running the CPU at 3 GHz was causing headaches with the old strategy but now we can run it as high as 3.5 GHz, infact it spends most of its time at that frequency"

So based on the above, we're told that both the CPU and GPU will spend most of their time at the subsequent max frequency clocks.

Cerny also says this.

"When that worst case arrives, it will run at a lower clockspeed, but not too much lower. To reduce power by 10%, it only takes a couple of percent reduction in frequency, so I'd expect any downclocking to be pretty minor"

Brad compares the PS5's boost with other product boosts like CPU's etc, showing a fundamental misunderstanding of Sony's frequency clock variability, and how it differs from other competing boost implementations in other GPU's, CPU's etc. Eg that it isn't overclocked temporarily, and then throttled down based on thermal dynamics, but instead the PS5's CPU/GPU spend "most" of their time at their max clocks, and only need to downclock a tiny amount in a "worst case" scenario (a 2% downclock for 10% power back, such a drop would keep the GPU at 10 Tflops) not when a thermal thread hold is reached, but instead a set power load limit.

Betelgeuse said:
3.5 GHz for the CPU, 2.23 GHz for the GPU.

I know you're thinking "no, I'm asking how low does it go", but I'm saying the above numbers are what make up the baseline. That's because it's where the PS5 sits, and only downclocks from there in certain limited circumstances - i.e., only specific workloads - and for limited durations.

It's really important to understand this distinction from traditional boost clocks, which is how Sams has erroneously framed this. PS5 doesn't sit at a lower clock/TF figure and only boost thereabove for limited durations. It's the opposite. The fact he phrases PS5's thermal management in terms of traditional PC-style boost clocks shows he fundamentally misunderstands how this all works.

In other words, PS5's baseline is its max clocks. It downclocks from there when needed. In the PC space, the baseline is a lower clock, with brief upclocks happening until a thermal limit is reached, at which point clocks fall back down to the baseline.

The 9.2 TF figure is misleading as well, if we are to take Cerny's remarks at face value. He said "to reduce power by 10 per cent it only takes a couple of percent reduction in frequency". Let's say that's a 3% downclock - that's still good for 10 TF. So unless Cerny is really downplaying the extent of downclocking that will be necessary to remain within the thermal budget, 9.2 TF is not likely to be reached in real world scenarios.

For further reading, I'd defer to Liabe Brave's posts. He has a far more thorough understanding about all of this than I.

I'm going to reply to both of you as you've both responded with similar takes.

There is one thing about this whole tiny little downclocking thing I'd like for everyone to consider. If the PS5 could consistently maintain 10tf, why the hell didn't they just lock it there and call it a day? It seems so much simpler. Devs don't ever have to even think about it, Cerny doesn't have to explain it, consumers/media don't have to speculate about it, etc. It's the obvious choice.

That is unless sustaining 10tf wasn't possible, in which case you realize it just looks better for marketing reasons to go with this variable option as 10.3tf looks a whole lot better on a spec sheet.

I'm just applying occam's razor here, and this is the simplest explanation for why we've got this variable clocks thing going on.

Foorbits said:
Sams also makes it seem like the flops between PS5 and Series X is a big deal but we should ignore the flops number on Lockhart because it's not a big deal. Pretty hilarious how he spun this article.

The Lockhart is a budget option that will play the same games as the Series X but at a lower resolution and without ray tracing. There is no reason to bring up the Lockhart in a conversation about the PS5 as they are not competing products. It's not that complicated.

jroc74 · Mar 27, 2020

nib95 said:
The amusing thing about that, is that we don't actually know the prices of these systems yet. If the PS5 does come in cheaper, it'll further diminish the credibility of that lopsided comparison, because the justification for Lockharts much lower performance is a more affordable price.

Yup.

If Lockhart stays at 4tf, the price of the PS5 and Lockhart will be very interesting.

Afrikan · Mar 27, 2020

Dave. said:
Brad Sams spewing misinformation, just so happens to pains the PS5 in a poor light...

When do TFlops Matter? - Thurrott.com

While Microsoft and Sony are taking similar paths with similar technology, there is a lot of comparison to previous consoles as well.

www.thurrott.com

These types are easy to spot after watching their previous videos.... I didn't bother watching this one after his specs "breakdown" from a week ago.

Muppet of a Man · Mar 27, 2020

I wish we could see some first party games/demos taking advantage of the custom SSD and I/O pipeline customizations. I think people are going to be surprised by what this enables.

This actually makes me want a true modern day ROM cartridge based console. I'd just like to be able to experience the new game design and interaction possibilities that affords. I think PS5 will function close to that sort of hardware architecture in practice.

Adder7806 · Mar 27, 2020

RingRang said:
. Devs don't ever have to even think about it,

Maybe Devs don't have to think about this solution either?

Doctor Avatar · Mar 27, 2020

We need some game demonstrations stat.

These things are out in six months, so they have game footage. SHOW IT.

nib95 · Mar 27, 2020

RingRang said:
I'm going to reply to both of you as you've both responded with similar takes.

There is one thing about this whole tiny little downclocking thing I'd like for everyone to consider. If the PS5 could consistently maintain 10tf, why the hell didn't they just lock it there and call it a day? It seems so much simpler. Devs don't ever have to even think about it, Cerny doesn't have to explain it, consumers/media don't have to speculate about it, etc. It's the obvious choice.

That is unless sustaining 10tf wasn't possible, in which case you realize it just looks better for marketing reasons to go with this variable option as 10.3tf looks a whole lot better on a spec sheet.

I'm just applying occam's razor here, and this is the simplest explanation for why we've got this variable clocks thing going on.

Because if you can gain the extra performance the bulk of use at the risk of a situation that only arises on rarer occasion, why not?

It's essentially a case of having more performance the majority of the time and a bit of added consideration or performance loss in worst case scenarios, vs having less performance the majority of the time in order to avoid the rare thermal heat or power spike. Sony decided the former was the better proposition, Microsoft went with the latter, and many GPU manufacturers/brands go for something between the two (fixed clocks but with thermal dynamics based overclocks).

Sony's take on this with the whole set power limit to determine clocks thing is actually pretty unique, so we don't know how successful it will or will not be.

Gofreak also extrapolated with more points.

Gofreak said:
The proposition is that:

1) Heat varies with workload, not just clock

2) Often - but not always - workloads would safely allow higher clocks than what they fixed in previous consoles. Safely, meaning, they could run those worloads at higher clocks and stay within the power/thermal limit of the console

3) Based on that, why not run those workloads at a higher frequency when you can?

The downside of a fixed frequency is that you're fixing it for an assumed worst case workload - such that the system won't break when a very intensive game comes along - and potentially leaving clockspeed on the table for a range of other workloads that could have sustained something higher without breaking the machine's heat dissipation limits.

Bobbyleejones · Mar 27, 2020

Betelgeuse said:
3.5 GHz for the CPU, 2.23 GHz for the GPU.

I know you're thinking "no, I'm asking how low does it go", but I'm saying the above numbers are what make up the baseline. That's because it's where the PS5 sits, and only downclocks from there in certain limited circumstances - i.e., only specific workloads - and for limited durations.

It's really important to understand this distinction from traditional boost clocks, which is how Sams has erroneously framed this. PS5 doesn't sit at a lower clock/TF figure and only boost thereabove for limited durations. It's the opposite. The fact he phrases PS5's thermal management in terms of traditional PC-style boost clocks shows he fundamentally misunderstands how this all works.

In other words, PS5's baseline is its max clocks. It downclocks from there when needed. In the PC space, the baseline is a lower clock, with brief upclocks happening until a thermal limit is reached, at which point clocks fall back down to the baseline.

The 9.2 TF figure is misleading as well, if we are to take Cerny's remarks at face value. He said "to reduce power by 10 per cent it only takes a couple of percent reduction in frequency". Let's say that's a 3% downclock - that's still good for 10 TF. So unless Cerny is really downplaying the extent of downclocking that will be necessary to remain within the TDP budget, 9.2 TF is not likely to be reached in real world scenarios.

For further reading, I'd defer to Liabe Brave's posts. He has a far more thorough understanding about all of this than I.

Wait how is cap a baseline, that makes no sense

hey it's that dog · Mar 27, 2020

Bobbyleejones said:
Wait how is cap a baseline, that makes no sense

Some of the confusion may be that baseline can colloquially mean either minimum or modal value. The PS5's 10.3 TF is not the minimum, obviously, but is, according to Cerny, the standard level of performance to which the occasional downclocks should be compared.

Iwao · Mar 27, 2020

Dave. said:
Brad Sams spewing misinformation, just so happens to pains the PS5 in a poor light...

When do TFlops Matter? - Thurrott.com

While Microsoft and Sony are taking similar paths with similar technology, there is a lot of comparison to previous consoles as well.

www.thurrott.com

Why are there so many Microsoft/Xbox gaming media proponents spreading misinformation about these consoles?

Betelgeuse said:
He's consistently demonstrated a facile understanding of technology, but this is a new low. This especially:

is straight up false reporting.

That's some questionable reporting, almost like he's not doing any research, or speaking with other people in the industry to fact check. How ridiculously unprofessional.

tutomos said:
Wasn't there already XBX for that?

Xbox One X will offer that same description for Lockhart, but lower frame rates. I think…. maybe…

Bobbyleejones · Mar 27, 2020

hey it's that dog said:
Some of the confusion may be that baseline can colloquially mean either minimum or modal value. The PS5's 10.3 TF is not the minimum, obviously, but is, according to Cerny, the standard level of performance to which the occasional downclocks should be compared.

I'm just wondering why he didn't just give a range of performance. Like the PS5 is most likely to operate at the 10.23 to 10.15 range 99% of the time

Betelgeuse · Mar 27, 2020

RingRang said:
There is one thing about this whole tiny little downclocking thing I'd like for everyone to consider. If the PS5 could consistently maintain 10tf, why the hell didn't they just lock it there and call it a day? It seems so much simpler. Devs don't ever have to even think about it, Cerny doesn't have to explain it, consumers/media don't have to speculate about it, etc. It's the obvious choice.

That is unless sustaining 10tf wasn't possible, in which case you realize it just looks better for marketing reasons to go with this variable option as 10.3tf looks a whole lot better on a spec sheet.

These are good questions, but they're above my pay grade so I don't really know. Lady Gaia has shared her informative experience as regards developer effort.

I will say that I heavily disagree with the notion that their variable clock implementation can be attributed to boosting PR. This undoubtedly was a massive engineering effort to get this to work. It wasn't just for image; they clearly decided there was a significant technical advantage that justified the time, cost, and energy to do this.

Bobbyleejones said:
Wait how is cap a baseline, that makes no sense

The term isn't ideal, because most people think of 'baseline' as being a minimum, such that performance can go above the minimum.

But it makes sense if you think of 'baseline' instead as meaning "where the machine's performance spends most of its time". In this case the baseline is max clocks, with downclocking being performed as needed.

CrispyGamer · Mar 27, 2020

Xeontech said:
That dude is a just another bitch ass ingnorant.

Watch the media take his words for gospel though.

Not helpful

tutomos · Mar 27, 2020

RingRang said:
The Lockhart is a budget option that will play the same games as the Series X but at a lower resolution and without ray tracing. There is no reason to bring up the Lockhart in a conversation about the PS5 as they are not competing products. It's not that complicated.

Wasn't there already XBX for that?

Bobbyleejones · Mar 27, 2020

Betelgeuse said:
These are good questions, but they're above my pay grade so I don't really know. Lady Gaia has shared her informative experience as regards developer effort.

I will say that I heavily disagree with the notion that their variable clock implementation can be attributed to boosting PR. This undoubtedly was a massive engineering effort to get this to work. It wasn't just for image; they clearly decided there was a significant technical advantage that justified the time, cost, and energy to do this.

The term isn't ideal, because most people think of 'baseline' as being a minimum, such that performance can go above the minimum.

But it makes sense if you think of 'baseline' instead as meaning "where the machine's performance spends most of its time". In this case the baseline is max clock, with performance decreasing as needed.

There has to be a reason to not state the base clock then no? Because lets say the PS5 does stay in 10.05 at the baseline, why not just state that. The reason being stating a max rate leads to many interpetations of how low the console itself can go

JaseC64 · Mar 27, 2020

Any expectations as to when Sony will reveal the box and maybe OS and some features demoed?

They can hold the games and price for later but I need more info now.

hey it's that dog · Mar 27, 2020

Bobbyleejones said:
I'm just wondering why he didn't just give a range of performance. Like the PS5 is most likely to operate at the 10.23 to 10.15 range 99% of the time

I'd imagine he can't give an accurate figure independent of the software being run so he didn't.

I'd also imagine that information isn't really useful to anyone except people arguing about bullet points for each console.

nib95 · Mar 27, 2020

RingRang said:
The Lockhart is a budget option that will play the same games as the Series X but at a lower resolution and without ray tracing. There is no reason to bring up the Lockhart in a conversation about the PS5 as they are not competing products. It's not that complicated.

Except it might be very relevant, because it's all relative. Imagine the following scenarios.

Lockhart, 4 Tflops - $299
PS5, 10.28 Tflops - $399
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $100, whilst the XSX would get you 18% extra for $100. Suddenly the PS5 is looking like a steal, and Lockhart a bit of a dud. The XSX's performance advantage is also looking like a pretty bad deal.

Potentially more realistic pricing.

Lockhart, 4 Tflops - $349
PS5, 10.28 Tflops - $449
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $100, whilst the XSX would get you 18% extra for $49. The PS5 is still looking like a steal, and Lockhart still a bit of a dud. XSX's performance advantage is looking like better value proposition.

And then the worst case for Sony, but that is still a possibility.

Lockhart, 4 Tflops - $349
PS5, 10.28 Tflops - $499
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $149, whilst the XSX would get you 18% extra for no added cost. Lockhart is looking like a better proposition now, but $149 extra for so much more performance from the other two still seems worthile, the XSX especially. Notably if buyers expect to get a 4K screen to replace their 1080p one, any time in the next several years.

Betelgeuse · Mar 27, 2020

Bobbyleejones said:
There has to be a reason to not state the base clock then no? Because lets say the PS5 does stay in 10.05 at the baseline, why not just state that. The reason being stating a max rate leads to many interpetations of how low the console itself can go

I don't know. Maybe I'm being overly charitable to Cerny, but it's possible the "base" clock varies quite a bit from game to game/workload to workload, so a single figure might not be very representative of all games.

I think for much of this discussion we'll have to wait until games come out and outlets like DF are able to report hard numbers. That's going to be really interesting to see what's really happening under the hood.

Axel Stone · Mar 27, 2020

JaseC64 said:
Any expectations as to when Sony will reveal the box and maybe OS and some features demoed?

They can hold the games and price for later but I need more info now.

I think it will definitely happen some time this year.

CrispyGamer · Mar 27, 2020

PS5 is a 10.28 TF machine and people seem to have a hard time dealing with that lol these people obviously don't plan to purchase one because if they did then why would they want it to be weaker? Are people deliberately mixing up the 10% reduce power to a 10% downclock? They can't let gitub go lol it's not a 9.2TF machine deal with it! In the end we all win right?

HalStep · Mar 27, 2020

i could be compltely wrong here (i know my numbers will be) but my understanding is:

10.3 is the base as in where it usually runs,that's 100% of the power being drawn,50% for cpu,50% for gpu.

but at some point for a few seconds the gpu might need to do something that need MOAR POWER so it will need 60% of the power leaving 40% for the cpu which will lower the clock and in turn lower the TF number for those few seconds.

so you can't say minimum speed or TF because it will only dip there shortly as needed depending on what devs do.

perhaps i'm completely wrong and that's all nonsense,i'll shut up and go back to just reading the thread.

RingRang · Mar 27, 2020

nib95 said:
Because if you can gain the extra performance the bulk of use at the risk of a situation that only arises on rarer occasion, why not?

It's essentially a case of having more performance the majority of the time and a bit of added consideration or performance loss in worst case scenarios, vs having less performance the majority of the time in order to avoid the rare thermal heat or power spike. Sony decided the former was the better proposition, Microsoft went with the latter, and many GPU manufacturers/brands go for something between the two (fixed clocks but with thermal dynamics based overclocks).

Sony's take on this with the whole set power limit to determine clocks thing is actually pretty unique, so we don't know how successful it will or will not be.

The fact is there is just so much we don't know here. If these thermal heat or power spikes were so rare why wouldn't they just be avoidable by the developers? Again, why would you design the consoles clocks around these "rare" instances? Why can't these rare instances just be avoided if they're predictable?

I think there is a reason this approach is so unique, and I dont think it's because it's so brilliant no other GPU/CPU maker has thought of it.

Betelgeuse said:
These are good questions, but they're above my pay grade so I don't really know. Lady Gaia has shared her informative experience as regards developer effort.

I will say that I heavily disagree with the notion that their variable clock implementation can be attributed to boosting PR. This undoubtedly was a massive engineering effort to get this to work. It wasn't just for image; they clearly decided there was a significant technical advantage that justified the time, cost, and energy to do this.

If there was a significant technical advantage surely we would see AMD implement this in their upcoming CPU/GPU coming out this year.

tutomos said:
Wasn't there already XBX for that?

The XB1X is part of the prior generation, and while Microsoft will support it for the early part of this generation, it will eventually be left behind.

Adum · Mar 27, 2020

RingRang said:
I'm going to reply to both of you as you've both responded with similar takes.

There is one thing about this whole tiny little downclocking thing I'd like for everyone to consider. If the PS5 could consistently maintain 10tf, why the hell didn't they just lock it there and call it a day? It seems so much simpler. Devs don't ever have to even think about it, Cerny doesn't have to explain it, consumers/media don't have to speculate about it, etc. It's the obvious choice.

That is unless sustaining 10tf wasn't possible, in which case you realize it just looks better for marketing reasons to go with this variable option as 10.3tf looks a whole lot better on a spec sheet.

I'm just applying occam's razor here, and this is the simplest explanation for why we've got this variable clocks thing going on.

I know you're probably not going to do this, but reading the past couple pages of this thread would give you a much better idea on why Sony chose to go with this method instead of fixed clocks. I also know you think you're trying to be reasonable here, but completely ignoring everything Mark Cerny said in the presentation and applying your own logic to it like this just makes you look a warrior. Mark Cerny -the guy who designed the console- gave out actual numbers in an official Sony presentation that was aimed at developers (no fluff marketing speak there) and your response is basically "BUT WHAT IF HE'S LYING THO. WHAT IF THOSE AREN'T THE ACTUAL NUMBERS"

RingRang · Mar 27, 2020

nib95 said:
Except it might be very relevant, because it's all relative. Imagine the following scenarios.

Lockhart, 4 Tflops - $299
PS5, 10.28 Tflops - $399
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $100, whilst the XSX would get you 18% extra for $100. Suddenly the PS5 is looking like a steal, and Lockhart a bit of a dud. The XSX's performance advantage is also looking like a pretty bad deal.

Potentially more realistic pricing.

Lockhart, 4 Tflops - $349
PS5, 10.28 Tflops - $449
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $100, whilst the XSX would get you 18% extra for $49. The PS5 is still looking like a steal, and Lockhart still a bit of a dud. XSX's performance advantage is looking like better value proposition.

And then the worst case for Sony, but that is still a possibility.

Lockhart, 4 Tflops - $349
PS5, 10.28 Tflops - $499
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $149, whilst the XSX would get you 18% extra for no added cost. Lockhart is looking like a better proposition now, but $149 extra for so much more performance from the other two still seems worthile, the XSX especially. Notably if buyers expect to get a 4K screen to replace their 1080p one, any time in the next several years.

I made no argument on the value proposition of the Lockhart. I only said it's irrelevant in a discussion around the Series X and PS5.

nib95 · Mar 27, 2020

RingRang said:
I made no argument on the value proposition of the Lockhart. I only said it's irrelevant in a discussion around the Series X and PS5.

Except it isn't, because if there's a price difference between the XSX and PS5, the exact same argument exists, only the PS5 ends up looking better all around, and Lockhart might end up seeming decidedly worse value proposition even if it is cheaper than the other two.

georgemichael · Mar 27, 2020

People still think lockhart exists

gothmog · Mar 27, 2020

nib95 said:
Except it might be very relevant, because it's all relative. Imagine the following scenarios.

Lockhart, 4 Tflops - $299
PS5, 10.28 Tflops - $399
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $100, whilst the XSX would get you 18% extra for $100. Suddenly the PS5 is looking like a steal, and Lockhart a bit of a dud. The XSX's performance advantage is also looking like a pretty bad deal.

Potentially more realistic pricing.

Lockhart, 4 Tflops - $349
PS5, 10.28 Tflops - $449
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $100, whilst the XSX would get you 18% extra for $49. The PS5 is still looking like a steal, and Lockhart still a bit of a dud. XSX's performance advantage is looking like better value proposition.

And then the worst case for Sony, but that is still a possibility.

Lockhart, 4 Tflops - $349
PS5, 10.28 Tflops - $499
XSX, 12.15 Tflops - $499

PS5 would get you 157% more performance for $149, whilst the XSX would get you 18% extra for no added cost. Lockhart is looking like a better proposition now, but $149 extra for so much more performance from the other two still seems worthile, the XSX especially. Notably if buyers expect to get a 4K screen to replace their 1080p one, any time in the next several years.

Sony has to be loving being the middle man. People tend to buy the middle choice since it makes them feel like they're getting the best price vs performance option.

Adum · Mar 27, 2020

RingRang said:
If there was a significant technical advantage surely we would see AMD implement this in their upcoming CPU/GPU coming out this year.

Do you really think that dedicated GPUs/CPUs have the same design goals as home consoles? They aren't as constrained when it comes to thermals and cooling designs. What Sony designed was the result of compromising between price, thermals and wanting 100% BC with the PS4. I doubt AMD has to think that hard about that stuff when designing their own chips.

the-pi-guy · Mar 27, 2020

RingRang said:
There is one thing about this whole tiny little downclocking thing I'd like for everyone to consider. If the PS5 could consistently maintain 10tf, why the hell didn't they just lock it there and call it a day?

Because 10.3 TF 99% and 10 TF 1% of the time offers more performance than 10 TF 100% of the time.

We don't really know for sure what the performance range is.

Consider these scenarios. Say we have similarly performing systems A, B, C, D, E, F that are mostly capable of performing 9 TF, 9.5 TF, 10 TF, 10.3 TF at some point.

TF A B C D E F
9 100.00% 90% 5% 1% 1% 0.20%
9.5 0 9% 90% 4% 4% 0.30%
10 0 1% 4% 90% 5% 0.50%
10.3 0 0 1% 5% 90% 99%
Total 9 9.055 9.503 9.985 10.24 10.2935

All of these systems have a minimum performance of 9 TF. That's their "baseline".
Compare System A, which is locked to 9 TF, and system B which is locked to 10 TF or below.

In order to achieve a locked frequency, we would need to go with a performance like system A.

System F has "unreliable" performance, but for all intents and purposes, it performs nearly identical to a 10.3 TF machine. This machine has 14% more performance than simply guaranteeing constant performance.

It's just a question of whether gaining that 14% performance boost is worth the 1% inconsistency.

PlayStation 5 System Architecture Deep Dive |OT| Secret Agent Cerny

Professionally Enhanced

Attempted to circumvent ban with alt account

I left my heart on Atropos

Alt account banned

Contains No Misinformation on Philly Cheesesteaks

Attempted to circumvent ban with alt account

Contains No Misinformation on Philly Cheesesteaks

Alt account banned

Contains No Misinformation on Philly Cheesesteaks

Enlightened

Contains No Misinformation on Philly Cheesesteaks

Alt account banned

Alt account banned

Contains No Misinformation on Philly Cheesesteaks