• 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.

Balancing clocks for max power draw is a different animal than variable clock throttling

PaintTinJr

Member
Jan 30, 2020
380
962
385
Whether the PS5 APU is variable based on temperature or power draw, can we all agree on one thing?

12 Sustained Tflops > 9 - 10 Variable Tflops
Greater in what way? Greater at puttinga small electrical capacitance in all the transistors that don't actually flop per clock on narrower workloads? Yeah, it will likely be significantly better at that, and therefore better at heating the surface of its APU, which inturn creates further resistance in those transistors, require greater power draw to maintain the clock - for no extra work done.
 
Last edited:

Pizdetz

Member
May 5, 2020
193
218
230
If the frequencies were fixed at max values, then it would overheat correct?
Also does the developer need to be aware of how to move the power from GPU or CPU or does the PS5 do it automatically?
If it's automatic does it mean that it somehow "knows" what's coming up? Is this feasible for most games?
For CPU intensive games, would it mean you're mostly going to get closer to 9 TF from this thing (eg. Ass creed?).
Just curious.
 
  • Like
Reactions: darrenskywalker

geordiemp

Member
Sep 5, 2013
8,578
5,387
790
UK
If the frequencies were fixed at max values, then it would overheat correct?
Also does the developer need to be aware of how to move the power from GPU or CPU or does the PS5 do it automatically?
If it's automatic does it mean that it somehow "knows" what's coming up? Is this feasible for most games?
For CPU intensive games, would it mean you're mostly going to get closer to 9 TF from this thing (eg. Ass creed?).
Just curious.
I cant take anymore stupid, I am out.

 
Last edited:

LazyParrot

Member
Mar 19, 2019
665
1,091
335
If the frequencies were fixed at max values, then it would overheat correct?
That depends entirely on the cooling. Pretty much all consumer GPUs can run at their maximum clock speeds indefinitely. Mine did for almost an entire month after Windows update fucked with my power settings.
 

Shin

Member
Feb 4, 2013
5,832
4,480
995
If the frequencies were fixed at max values, then it would overheat correct?
Might overheat regardless, electronics go through various stress tests, but it's an electronic and everything and anything can go wrong no matter how much you test it.
Fixed would would be more prone to overheating hence why the other camp went with 1750 or whatever it (in line with my expectation of what's a safe bet for 7nm).
I'm having issues (yet would love to see them pull it off as it would be an amazing feat.) with how high the clocks are, from the top of my head I can't remember a GPU running so high.
On top of that you'd think that the actual GPU vendors that have way more years of experience in this field would comp up with the idea long ago and someone that throws an idea together and run with it (for a lack of better word).
 

Rikkori

Member
May 9, 2020
395
543
290
It's crazy how much Sony bullshitters are trying to spin things, and yet the saddest thing is - what is it for anyway, because the difference even in a throttle scenario would be small?

It's like lying about 0.01" of height, it's just dumb.

Thermals still matter, and unless you're paying $100 for your cooling solution then it's still something to account for quite seriously (and let's remember - consoles aren't hardware you overspend on, it's something you penny pinch on, because you're gonna sell 10s-100s of millions of them. every cent counts). And why are they waiting an extra year compared to Xbox to show a breakdown? If they're this confident in their solution & how little thermals mean now then why aren't they shouting it from the rooftops?

Show, don't tell.
 

Jon Neu

..he neus nothing
Jan 21, 2018
2,653
4,613
675
Republic of Catalonia
That the only thing you have ever posted that almost makes sense abut you still got it wrong,
The clocks of the XsX doesn’t have anything to do with what I said.

But the problem is that you have your nose so far up Sony’s ass you can’t stop to jump to the Sony is best narrative at every corner, and the best defense is to make a comparison with the Xbox clocks that are lower, so therefore it must be because they can’t reach the PS5 clocks without using the same PS5 solution, right? Impressive argumentation.

It’s not like, you know, they prefer to have 12 teraflops instead of 10 and therefore they prefer to sacrifice clock speed.

Sony, like MS, compromises in one aspect and is trying it’s best to make those compromises the smaller they can -that’s why Cerny is so happy to claim that is only a 2% downclock-, but still are compromises.
 
May 10, 2018
110
176
260
CA
Greater in what way? Greater at puttinga small electrical capacitance in all the transistors that don't actually flop per clock on narrower workloads? Yeah, it will likely be significantly better at that, and therefore better at heating the surface of its APU, which inturn creates further resistance in those transistors, require greater power draw to maintain the clock - for no extract work done.
As a layman consumer, why should I care whether the APU is being worked to capacity 100% of the time even when it's not required, or whether it is able to throttle according to it's needs?

Unless I was counting pennies on the amount I could save on my electricity bill each month or a heating issue became apparent, it makes no difference to the average consumer. Knowing developers will have more theoretical raw power to make great playing and looking games is what I care about.
 
Last edited:

geordiemp

Member
Sep 5, 2013
8,578
5,387
790
UK
The clocks of the XsX doesn’t have anything to do with what I said.

But the problem is that you have your nose so far up Sony’s ass you can’t stop to jump to the Sony is best narrative at every corner, and the best defense is to make a comparison with the Xbox clocks that are lower, so therefore it must be because they can’t reach the PS5 clocks without using the same PS5 solution, right? Impressive argumentation.

It’s not like, you know, they prefer to have 12 teraflops instead of 10 and therefore they prefer to sacrifice clock speed.

Sony, like MS, compromises in one aspect and is trying it’s best to make those compromises the smaller they can -that’s why Cerny is so happy to claim that is only a 2% downclock-, but still are compromises.
Ps5 has higher clocks, do you know what that does other than texels ?

Show, don't tell.
Let me show you, now its your turn to show

 
Last edited:

Bojanglez

The Amiga Brotherhood
Apr 16, 2007
1,630
131
1,270
Liverpool, England
It's crazy how much Sony bullshitters are trying to spin things, and yet the saddest thing is - what is it for anyway, because the difference even in a throttle scenario would be small?

It's like lying about 0.01" of height, it's just dumb.

Thermals still matter, and unless you're paying $100 for your cooling solution then it's still something to account for quite seriously (and let's remember - consoles aren't hardware you overspend on, it's something you penny pinch on, because you're gonna sell 10s-100s of millions of them. every cent counts). And why are they waiting an extra year compared to Xbox to show a breakdown? If they're this confident in their solution & how little thermals mean now then why aren't they shouting it from the rooftops?

Show, don't tell.
Maybe they prefer people to see their games than the internal of their console.

A teardown will come.
 

Major_Key

Banned
Apr 12, 2020
97
279
360
Theoretical based on RDNA1 mumbers, we dont have ROPS yet and all numbers are max potential including TF.

Triangle rasterisation is 4 triangles per cycle.

PS5:
4 x 2.23 GHz ~ 8.92 Billion triangles per second

XSX:
4 x 1.825 GHz - 7.3 Billion triangles per second

Triangle culling rate is twice number triangles rasterised per cycle.

PS5:
8 x 2.23 GHz - 17.84 Billion triangles per second

XSX:
8 x 1.825 GHz - 14.6 Billion triangles per second

Pixel fillrate is with 4 shader arrays with 4 RBs (render backends) each, and each RB outputtting 4 pixels each. So 64 pixels per cycle.

PS5:
64 x 2.23 GHz - 142.72 Billion pixels per second

XSX:
64 x 1.825 GHz - 116.8 Billion pixels per second

Texture fillrate is based on 4 texture units (TMUs) per CU.

PS5:
4 x 36 x 2.23 GHz - 321.12 Billion texels per second

XSX:
4 x 52 x 1.825 GHz - 379.6 Billion texels per second

Raytracing in RDNA2 is alleged to be from modified TMUs.

PS5:
4 x 36 x 2.23 GHz - 321.12 Billion ray intersections per second

XSX:
4 x 52 x 1.825 GHz - 379.6 Billion Ray intersections per second
AND ?
 

loktaiextatus

Member
Apr 12, 2020
265
233
220
If the frequencies were fixed at max values, then it would overheat correct?
Also does the developer need to be aware of how to move the power from GPU or CPU or does the PS5 do it automatically?
If it's automatic does it mean that it somehow "knows" what's coming up? Is this feasible for most games?
For CPU intensive games, would it mean you're mostly going to get closer to 9 TF from this thing (eg. Ass creed?).
Just curious.
TF is the rating of the CUs with their specific paramaters X clock speed for how many floating point operations the GPU is capable of
per SECOND. If the game is CPU intensive- meaning it needs more CPU than GPU- and the CPU would be the bottleneck what are you
asking about... because the CPU is going to likely run at full clocks alongside the GPU, we dont have any details on power consumption.
Cerny stated a few percent, I would say the most you'll drop to is 10 BASED ON what he said- If it drops int the 9s who knows.
Here is another way to phrase your question-

If a Game is CPU intensive and not optimized for heavy use of the GPU- and is not programmed to use enough compute units
because its programmed for a common denominator / common platform for instance, the PS5 or a 5700XT with fewer than 40 CUs,
what would the XBOX's TFLOP rating be? Lets do the math. If a game is not optimized to use many CUs lets say it uses 40 which is more
than on the PS5.... AND LETS say the xbox is clocked at, what is its clock 1850? that is Minus 12 CUs.... 12 CUs is 9TF. Meanwhile Sony's system only has
36 CUs so it is missing 4 to hit how many the Xbox is using. That means its actually crunching at its full 10.3tf... Why is that?
Because per Cu the Sony console has a much higher clock speed.
 
  • Like
Reactions: polybius80

MarkMe2525

Banned
Jul 16, 2017
230
238
230
34
The power draw between the CPU and GPU is variable correct? It goes where needed. So it doesn't throttle GPU when it gets hot, it throttles GPU when CPU needs power and vice versa.
 

Shin

Member
Feb 4, 2013
5,832
4,480
995
Do you?
Your timestamp highlights limitless nanite and geometry, so what are you trying to say here as the engine runs on both NG consoles.
Genuinely asking because your post didn't make sense, back forwarded even to hear the guy address the scene(s).
 

loktaiextatus

Member
Apr 12, 2020
265
233
220
Knowing developers will have more theoretical raw power to make great playing and looking games is what I care about.
Really? Then you should care about the actual games. Because we have the example of the XBox one X and the PS4 pro today and thats a huge gap
compared to the new consoles.... And if you're talking about developers making great playing and looking games I dont think the PS5 is going to have
much of a problem.
 
Last edited:

geordiemp

Member
Sep 5, 2013
8,578
5,387
790
UK
Do you?
Your timestamp highlights limitless nanite and geometry, so what are you trying to say here as the engine runs on both NG consoles.
Genuinely asking because your post didn't make sense, back forwarded even to hear the guy address the scene(s).
Higher frequency and faster IO has benefits,

You linked a tech demo for a 3rd party engine that is available across all platforms and where the engineers have already confirmed said demo is possible on XSX. How is that a valid response?
The demo is also possible on the ps4 and xb1. If you think XSX will run that demo better, keep wishing.

Show dont tell.,
 
Last edited:

Jon Neu

..he neus nothing
Jan 21, 2018
2,653
4,613
675
Republic of Catalonia
Let me show you, now its your turn to show

But we will finally reach Motorstorm graphics or not?

Asking for a friend.
"Variable frenquencies is better than Sustainable frequencies"

The funny thing is that we know that if the consoles were reversed, the ones claiming variable frequencies are better would be claiming the opposite.
 
May 10, 2018
110
176
260
CA
Really? Then you should care about the actual games. Because we have the example of the XBox one X and the PS4 pro today and thats a huge gap
compared to the new consoles.... And if you're talking about developers making great playing and looking games I dont think the PS5 is going to have
much of a problem.
Exactly, the gap between PS4 PRO and X1X is the reason I was able to play the best console version of Red Dead 2 and other 3rd party titles. Plus, I'm a die hard Halo fan so there's that.
 
Last edited:

Thugnificient

Member
May 29, 2020
221
526
270
Theoretical based on RDNA1 mumbers, we dont have ROPS yet and all numbers are max potential including TF.

Triangle rasterisation is 4 triangles per cycle.

PS5:
4 x 2.23 GHz ~ 8.92 Billion triangles per second

XSX:
4 x 1.825 GHz - 7.3 Billion triangles per second

Triangle culling rate is twice number triangles rasterised per cycle.

PS5:
8 x 2.23 GHz - 17.84 Billion triangles per second

XSX:
8 x 1.825 GHz - 14.6 Billion triangles per second

Pixel fillrate is with 4 shader arrays with 4 RBs (render backends) each, and each RB outputtting 4 pixels each. So 64 pixels per cycle.

PS5:
64 x 2.23 GHz - 142.72 Billion pixels per second

XSX:
64 x 1.825 GHz - 116.8 Billion pixels per second

Texture fillrate is based on 4 texture units (TMUs) per CU.

PS5:
4 x 36 x 2.23 GHz - 321.12 Billion texels per second

XSX:
4 x 52 x 1.825 GHz - 379.6 Billion texels per second

Raytracing in RDNA2 is alleged to be from modified TMUs.

PS5:
4 x 36 x 2.23 GHz - 321.12 Billion ray intersections per second

XSX:
4 x 52 x 1.825 GHz - 379.6 Billion Ray intersections per second
That's contingent on both relying on a 4 shaders engine system though.
 
Last edited:
  • Like
Reactions: geordiemp

Major_Key

Banned
Apr 12, 2020
97
279
360
You want to know what's actually loopy?

You creating this thread based on a tweet you failed to even read properly.

While you're here, why don't you also add this to the OP? :

PS5 is BEAST console.
 

Rikkori

Member
May 9, 2020
395
543
290
Oh man, if only AMD managed to put this magical "thermals don't matter anymore" technology in their desktop CPUs & GPUs as well, maybe Nvidia wouldn't wipe the floor with them. But I guess they're just that good friends with Sony, they'd rather tank their company than make use of it. #GoodGalAMD

And what's heat density anyway? Sounds spooky. Then again, thermals don't matter anyway because they "design for power". Why didn't anyone think of this before!

I guess we'll never find out. 🤷‍♂️
 

loktaiextatus

Member
Apr 12, 2020
265
233
220
Exactly, the gap between PS4 PRO and X1X is the reason I was able to play the best console version of Red Dead 2 and other 3rd party titles. Plus, I'm a die hard Halo fan so there's that.
So a Halo fan. Then you certainly need an XBOX. And yes.. the "best console version" thats still inferior to PC and by best console version you mean
crisper.. sure. The 50% performance increase gave you a crisper image. The 15% ... Well if you think faster transitions/loading means better gameplay too
then it'll be a tradeoff by my estimation. I have both of those systems I mentioned and Red Dead is the one xbox guys ALWAYs have to bring up because it was
def one of the ones that DF showed a big difference in, maybe the biggest difference. I dont mean to downplay it- because crisp is good- But if I am honest
I would not trade Red Dead Redemption 2 for the Sony exclusives, ok ... actually I'll trade it outright for the last of us 2 and I'll throw in Knack since Im feeling generous.
 

geordiemp

Member
Sep 5, 2013
8,578
5,387
790
UK
Both consoles is beast in their own way, so all good bro geordiemp geordiemp

Next gen is crazy !
I dont mind talking about how they are doing the variable frequency, its interesting. Its the mindless FUD thats annoying.

Both are good consoles, XSX will do some things better, Ps5 will do somethings better. It will average out allot.

Anyone thinking of any great superiority next gen will be the disapointed ones.
 
Last edited:

loktaiextatus

Member
Apr 12, 2020
265
233
220
Oh man, if only AMD managed to put this magical "thermals don't matter anymore" technology in their desktop CPUs & GPUs as well, maybe Nvidia wouldn't wipe the floor with them. But I guess they're just that good friends with Sony, they'd rather tank their company than make use of it. #GoodGalAMD

And what's heat density anyway? Sounds spooky. Then again, thermals don't matter anyway because they "design for power". Why didn't anyone think of this before!

I guess we'll never find out. 🤷‍♂️
Or you'll find out when AMD releases this tech to the PC market later this year when RDNA2 becomes available?
 

Greeno

Neo Member
Jun 14, 2020
5
5
75
The power draw between the CPU and GPU is variable correct? It goes where needed. So it doesn't throttle GPU when it gets hot, it throttles GPU when CPU needs power and vice versa.
This is exactly what I understood. It seems to mean that for the GPU to be at its top game, the CPU needs to take a hit. For the CPU to be at its top game, the GPU takes a hit. Which is not ideal to be honest.
 
Mar 27, 2020
2,988
7,890
555
And why are they waiting an extra year compared to Xbox to show a breakdown? If they're this confident in their solution & how little thermals mean now then why aren't they shouting it from the rooftops?

Your just impatient that's all. Sony will have a breakdown and they will show us the cooling system. They have already promised that to us so they are confident in the systems design. If they weren't they wouldn't break it down for us.

Just because Microsoft did it earlier in the year doesn't mean that Sony has to satisfy impatient people like yourself. They will reveal it when they want to. I remember there were talks about Sony not having a reveal event because they were going to delay the system or that Sony was afraid of the XSX which is why they wouldn't have a reveal.

In the end Sony has a timeline that they follow and it isn't based off what Microsoft does.

You will get your breakdown and reveal of the cooling system.

Please be patient.
 
Last edited:

OliverKahn

Banned
Mar 22, 2020
116
444
395
Theoretical based on RDNA1 mumbers, we dont have ROPS yet and all numbers are max potential including TF.

Triangle rasterisation is 4 triangles per cycle.

PS5:
4 x 2.23 GHz ~ 8.92 Billion triangles per second

XSX:
4 x 1.825 GHz - 7.3 Billion triangles per second

Triangle culling rate is twice number triangles rasterised per cycle.

PS5:
8 x 2.23 GHz - 17.84 Billion triangles per second

XSX:
8 x 1.825 GHz - 14.6 Billion triangles per second

Pixel fillrate is with 4 shader arrays with 4 RBs (render backends) each, and each RB outputtting 4 pixels each. So 64 pixels per cycle.

PS5:
64 x 2.23 GHz - 142.72 Billion pixels per second

XSX:
64 x 1.825 GHz - 116.8 Billion pixels per second

Texture fillrate is based on 4 texture units (TMUs) per CU.

PS5:
4 x 36 x 2.23 GHz - 321.12 Billion texels per second

XSX:
4 x 52 x 1.825 GHz - 379.6 Billion texels per second

Raytracing in RDNA2 is alleged to be from modified TMUs.

PS5:
4 x 36 x 2.23 GHz - 321.12 Billion ray intersections per second

XSX:
4 x 52 x 1.825 GHz - 379.6 Billion Ray intersections per second
You must be drunk at this point
 
  • LOL
Reactions: Jon Neu

TLZ

Member
Oct 20, 2015
9,106
11,580
865
Sydney
Isn't he essentially saying the term variable clocks shouldn't really be applied to PS5 because traditionally it had been associated with throttling due to temperatures and is unpredictable. Whereas the PS5 is the "different animal" in a good way?
Yes. That's what I understood. Sounds good and also in line with what Cerny said too. I guess he's probably just reiterating what Cerny said. I hope this works really well.
 

Shin

Member
Feb 4, 2013
5,832
4,480
995
Someone should ask that ex Sony employee why he's not long with them, since they are trying to westernize and shit.
You'd think they want more western building blocks, not like others on the game left the company after completing the PS4.
Yet this guy is always making noise about PS5 (even defensive ones as seen in the reply to the French person), kinda strange.

Passion about your work I can get, but then don't say anything at all - it's not your job, Sony has a marketing department for that.
 

TLZ

Member
Oct 20, 2015
9,106
11,580
865
Sydney
Someone should ask that ex Sony employee why he's not long with them, since they are trying to westernize and shit.
You'd think they want more western building blocks, not like others on the game left the company after completing the PS4.
Yet this guy is always making noise about PS5 (even defensive ones as seen in the reply to the French person), kinda strange.

Passion about your work I can get, but then don't say anything at all - it's not your job, Sony has a marketing department for that.
Just like Penello I guess. No longer at MS but roots for them still.
 
  • Like
Reactions: Bo_Hazem

geordiemp

Member
Sep 5, 2013
8,578
5,387
790
UK
They are talking about throttling a GPU. They havent released any RDNA2 gpus yet. Arent we talking about the GPU throttling for workload here or
not? Does the Zen2 Have an RDNA2 GPU in it?
We have not seen this technology before correct. Most boosts work of multiple temps spread across the die, and if there is a hot spot away from the sensor that is not good, so you have to be conservative.

If you can predict heat and thermals from workload predictively its better not just in the above hot spot case but also PID control and timing. Reducing heat generated is better than acting when things are already hot.

The real question is how good is the modelling and prediction logic, and I am sure we will get endless presentations in due course.
 
Last edited: