Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]

[MrFox]

If fixed clocks of 2.0 GPU and 3.0 CPU were hitting a power wall, how is it possible for the variable approach to constantly keep both near 2.23 and 3.5? Doesn't add up.

They were not hitting a power wall, they never said that. It was causing cost/performance issues. They can make a 1000W console any time they want if they have no cost limit on the cooling. It's about the balancing act of where to invest a limited BOM.

 

[shiznit]

They were not hitting a power wall, they never said that. It was causing cost/performance issues. They can make a 1000W console any time they want if they have no cost limit on the cooling. It's about the balancing act of where to invest a limited BOM.

I get what you're saying but it still doesn't make sense to me, unless you're referring to absolute worst corner case. Say clocks of 2.0/3.0 were consuming 200W under typical load, how is 98% of 2.23/3.5 going to consume less? They are claiming to be able to sustain higher clocks than the fixed version.

 

[MrFox]

I get what you're saying but it still doesn't make sense to me, unless you're referring to absolute worst corner case. Say clocks of 2.0/3.0 were consuming 200W under typical load, how is 98% of 2.23/3.5 going to consume less? They are claiming to be able to sustain higher clocks than the fixed version.

Yes, and he explains why the average is higher than a fixed clock, because they don't have to design the entire power delivery and cooling system just for the massive outlier peaks, which may or may not even happen. The normal gaming demand however is "most of the time".

It's about averages, and peak being crazy exponential outliers.

We have the same thing on gpus with "gaming clocks" which are much higher than the guaranteed base clock. If you have to guarantee any sequence at a fixed clock, you have to use the base clock. Which is what ps4/xb1 mostly did.

 

[3dilettante]

A few questions for anyone who know... things....

1. Is AVX256 disabled on xbsx to do 3.6/3.8 ?

That might be one way to make clocks more predictable, or perhaps given what we know about the power delivery for the Series X that means Sony's console may run at lower consumption.
There are also other ways to reduce consumption that are like throttling clocks, like instruction issue limits or warm-up periods. Duty cycling could keep the "fixed clock" even though it halts some activity on a routine basis.

2. Considering 2.23ghz is a parametric yield, does it mean a super expensive cooling solution could sustain that clock for any load? Or would it be already beyond the heat spreader capabilities (die-to-surface thermal resistance)?

We'd need to know more about the hot-spot properties of the chip and process, and the ability of the chip and board to deliver adequate power. The power-based modeling and Vdroop compensation AMD has can make some of these problems more readily handled, especially since temperature-based tracking can lag too much at these scales, whereas activity counters or voltage drop can highlight high consumption in a handful of cycles.

I am guessing yes on the first. As for the second, I would suppose that parametric yield for ps5 have a relatively high test voltage attached, to keep a high number of passing dies.

It's possible that Sony pulled an AMD and left voltages higher than necessary for many of its chips. In that scenario, having a dynamic clock method like AMD's can enable higher consumption than would be possible otherwise, since some of the chips that need the higher voltage could have failed more stringent limits. I think Sony would want a better handle on power consumption, but then again if their choice in CU count forced them to higher clocks--which the clocking method makes possible--then they might be worse off in terms of overall power than they would have been otherwise.

If fixed clocks of 2.0 GPU and 3.0 CPU were hitting a power wall, how is it possible for the variable approach to constantly keep both near 2.23 and 3.5? Doesn't add up.

Cerny outlined a scenario where the question was whether they would bind the platform clocks to what they estimated was the worst-case consumption scenario. The kinds of operations and the overall utilization can wildly change power consumption, and the designers would need to make a prediction about all the software the chip would ever run, and hope they guessed right.
Should a design cap its frequencies at a level that is safe for some worst-case they cannot predict, and what can they do if they guess wrong. A clocking method that instead catches when they meet a high-consumption scenario and allows higher clocks at all other times would be tempting.

 

[shiznit]

Yes, and he explains why the average is higher than a fixed clock, because they don't have to design the entire power delivery and cooling system just for the massive outlier peaks, which may or may not even happen. The normal gaming demand however is "most of the time".

It's about averages, and peak being crazy exponential outliers.

We have the same thing on gpus with "gaming clocks" which are much higher than the guaranteed base clock.

They could have still throttled under 2.0/3.0 for the outliers, I'm not referring to the benefits of a power cap, I get that. What doesn't make sense is their claim of sustained clocks above 2.0/3.0 using less power than 2.0/3.0 fixed.

 

[goonergaz]

I just cannot imagine what ground up games for these things are going to look like. Zen2's have ~5x more performances then Jaguars they replace, this is going to be absolutely insane.

People are going on about texturing, streaming etc. and I know devs will benefit most from fast SSD, but interactivity and physics jump we are going to get will be absolutely insane.

Last gen it was all about prittier pixels. Next gen we'll get much more.

I don’t get the excitement over the CPUs, to be honest I’m disappointed- last gen CPUs were criticised for being too weak so 5x isn’t much of an upgrade.

They could have still throttled under 2.0/3.0 for the outliers, I'm not referring to the benefits of a power cap, I get that. What doesn't make sense is their claim of sustained clocks above 2.0/3.0 using less power than 2.0/3.0 fixed.

Groundhog Day! I guess we just have to wait for me details from Sony.

 

[Kaotik]

I don’t get the excitement over the CPUs, to be honest I’m disappointed- last gen CPUs were criticised for being too weak so 5x isn’t much of an upgrade.

Zen 2 is pretty much the most powerful x86-architecture/implementation out there at the moment, I don't know how anyone could have really asked for more on these consoles. 16 cores?

 

[PSman1700]

Zen 2 is pretty much the most powerful x86-architecture/implementation out there at the moment, I don't know how anyone could have really asked for more on these consoles. 16 cores?

Probably Zen 3 I think we should be really happy a 8 core 16 thread CPU running at close to 4ghz in a console. Powerfull GPU's and and SSDs, i think games designed with that in mind will be amazing on every level.

 

[Rurouni]

They could have still throttled under 2.0/3.0 for the outliers, I'm not referring to the benefits of a power cap, I get that. What doesn't make sense is their claim of sustained clocks above 2.0/3.0 using less power than 2.0/3.0 fixed.

Do they claim that?
If they do, it might be about the possibility of one game to run at higher clock while using less power vs another game that run at lower clock but using more power. For an extreme example, God of War probably will use more power at 2.0/3.0 vs Minecraft at 2.23/3.5. With a fixed clock you're basically leaving extra performance on the table because not all of the games will be able to work the CPU and GPU as hard as God of War.

Of course you can design a system with a fixed clock that will not overheat even under extreme load, like fixing it to a lower clock or using a more potent cooling. You'll get a system that potentially can perform better but running slower to account for the peak or a system that rarely use its cooling system at max capacity or something in between that.

What I'm curious about is why MS doesn't use variable clockspeed since arguably they are a more PC centric company and since PC games doesn't mind variable clockspeed, why they chose a fixed speed? It probably does make it easier for a dev since they have less thing to worry about, but then they need to make (at least for a year?) their game run on XsX and XO. Also considering how much Xbox exclusives are actually only console exclusive (available on PC even for MS own games), if you ask me which company will embrace variable clockspeed first, I will answer MS.

 

[dobwal]

I get what you're saying but it still doesn't make sense to me, unless you're referring to absolute worst corner case. Say clocks of 2.0/3.0 were consuming 200W under typical load, how is 98% of 2.23/3.5 going to consume less? They are claiming to be able to sustain higher clocks than the fixed version.

The clocks are basically fixed except with certain caveats. I think they just used horrible language when they used the term “boost”.

Imagine buying a PS4 Pro with a chip that barely passed QA/QC and is prone to power leakage. Every time you play a gpu intensive game it whines like a jet engine.

The PS5 solution for such a chip is not to let it wind up the fans to such high levels. The console will back down the frequency of its cpu/gpu to compensate and lower the fps of the game to compensate.

It’s not a given that all PS5s will need to downclock as chips across a wafer will have different power characteristics. Some will handle 2.23 and 3.5 GHz more readily than others as they won’t need as much power to sustain those clocks.

 

[KeanuReeves]

They could have still throttled under 2.0/3.0 for the outliers, I'm not referring to the benefits of a power cap, I get that. What doesn't make sense is their claim of sustained clocks above 2.0/3.0 using less power than 2.0/3.0 fixed.

We have been through this one a few times now Fixed 2.0/3.0 is just a hypothetical. If they keep it there fixed then the outlier scenarios would just need a slightly bigger fan and powersupply to manage it (at the cost of fan noise). But with the variable frequency they can have higher clocks that drop when power hungry instructions are used too much. That means they don't need to have a bigger power supply and they don't need to have a bigger fan. They just drop the frequency by a few points.

 

[Rockster]

So many issues being conflated. Hopefully, some of you find this helpful.

1) There are other constraints that limit the maximum frequency a part can run at besides heat and power. Ultimately you'd be limited by the speed of the transistors and wiring delays. So the chip design and physical signalling components impose hard limits. Please don't think console manufacturers plan clocks around power supply and cooling solution budgets. Everything they do in that regard is within a predetermined window.

2) The PS5 implementation of variable clocks essentially ignores temperature. Sony has set a console standard, or "model", as they refer to it, where typical PS5's are placed in what I'd assume to be a sort of worst-case cooling environment and tested. They then measure the max power draw at which their cooling solution meets their reliability and acoustic thresholds. The result of this testing established a fixed maximum power rating for their SoC which is to be applied to all PS5's regardless of the actual temperature each might run at. So if person A is running their PS5 in a freezer and person B is in the desert, both will have the same power limit imposed.

3) Power draw changes with frequency and load. I think this is causing the most confusion. Processors (CPU and GPU), even when running at their maximum configured clock speed, will draw relatively little power if the majority of its execution units are idle. So even at fixed clock speeds, sitting in a menu doing nothing will draw less power and in turn, run cooler then while in a busy game scenario. Processors are massively parallel units, meaning they can execute lots of operations concurrently. For example, every CU in the GPU is technically able to carry out 128 concurrent operations each clock cycle. Applications, such as games, vary on how effectively they can utilize all the available compute components within a processor simultaneously. As programmers optimize their code and data so the processor can utilize greater and greater amounts of its available compute concurrently, the processing load, and in turn power draw, increases. At some computational utilization threshold (load), the application would reach the defined power limit. In order to be able to concurrently utilize all its available compute and still be under Sony's previously defined power ceiling, the PS5 reduces clock speeds to compensate for that load.

4) Sony has said, when the CPU and GPU aren't being fully utilized, they can run up to 3.5GHz and 2.223 GHz respectively. And that at higher loads, they will run at lower (as of yet undisclosed) clock-speeds. The only sense we have for clock-speeds in a hypothetical 100% load scenario across the entire SOC, are Cerny's comments that 3GHz on the CPU "was causing headaches" and 2GHz on the GPU "was looking like an unreachable target". As 100% utilization across the entire SoC in a practical sense is not possible just due to the inherent inefficiencies of real-world code. data sets, and compute requirements; the expectation Sony has is that they will run at or near the max clocks much of the time.

 

[temesgen]

CPU requirements tend to rise also with higher framerates, atleast on pc that's the case.

I think this is a bigger hint than many are realizing.

We may care about frame rates north of 30, but the average gamer doesn't. Sony may very well have designed the PS5 to be able to generate 4k with a little additional headroom and focused the overall design around actually providing the developers with tools to make game design without all the legacy requirements mechanical drives and legacy sound applications limits their work to.

Immersion has more to do less waiting around and more vivid sound once the graphics get to a certain point.

VR can add to that immersion too but they may see it as a niche market until there is significant improvement with the controls, dual game development (vr and conventional) and cost reductions. If that's the case, they don't need a GPU capable of 120 fps.


RTRT remains an open question but the Sony developer comment on Twitter may hint at Ray tracing this generation being a fundamentally limited application in a practical sense and if it adds significant cost why bother?

A cheaper machine pumping out beautiful 4k images with great texture and sound and essentially no loading could be the right answer. Of course if the cooling solution adds significant cost or the dynamic clocks adds BC issues migrating to PS6, it will be a mistake.

 

[Deleted member 11852]

Is there a transcript available for the Road to PS5 video?

There's no actual transcribing application you can just readily use on youtube.

YouTube does transcribing for most videos. Click the three dots (". . .") under the video then Open Transcript. It's not formatted but it's all time encoded but you can hide the time codes for copying and pasting. The start of the transcript (not Cerny) looks like this:

Transcript

00:00 hi unfortunately we had to cancel the
00:03 the talk that we had planned for GDC but
00:08 we do have some super exciting news
00:10 about ps5 and who better to bring that
00:15 to you than the one and only mark Cerny
00:17 without further ado over to you mark
00:18 thank you Jim there will be lots of
00:24 chances later on this year to look at
00:25 the PlayStation 5 games today I want to
00:28 talk a bit about our goals for the
00:30 PlayStation 5 hardware and how they
00:32 influenced the development of the​

Useful feature hidden away in a submenu!

 

[Globalisateur]

I get what you're saying but it still doesn't make sense to me, unless you're referring to absolute worst corner case. Say clocks of 2.0/3.0 were consuming 200W under typical load, how is 98% of 2.23/3.5 going to consume less? They are claiming to be able to sustain higher clocks than the fixed version.

At the same clocks a console can consume up to 2x more power, we can see this on Pro when you compare dashboard or indies games vs max power consumption. Power consumption depends a lot of actual load. And In many cases on PS4 the thing consumes the most when it's not even in actual gameplay: during custcenes (Cerny talked about it actually), in the dashboard while the game runs in the background or in a menu: in both latter cases it's usually because the framerate is uncapped and runs as high as it can.

What Cerny has done on PS5 is actually smart and is the way forward in console gaming. Like the share button before I wouldn't surprised if Microsoft copy Sony again and do the same in their next console. The whole package is really innovative. But it's going to depend of the cooling, it needs to be as innovative to cool that much heat. We'll see but I think the only way is to cool both sides of the APU.

Anecdotally when the Navi cards where first benchmarked I saw a video of a guy who used a big fan in order to cool the back side of the GPU board. Thanks to this he could immediately overclock the card by about 150-200 mhz (don't remember the exact values here). Too bad I can't find the video again.

 

[Metal_Spirit]

Power consumption is dependent on workload and chip "quality", which is why same code doesn't produce same power consumption on every console. Leakage affects both how high you can clock the soc and how much voltage each soc requires for said clocks.

That's irrelevant... ALL consoles are affected by chip quality, and ALL consoles are affected by power consumption. So ALL consoles might have problems with clocks or thermals.
Thats why ALL consoles are set withing a safety margin to account for those variations.

Sorry for the ALL in caps, but in the context of the conversation I have to make clear that what you mention is not a specific PS5 thing.

 

[RobertR1]

Consoles generally don't get to bin like that, hence why many speculated it might be more difficult to push the silicon so much further than had been demonstrated previously. The traditionally fixed performance behavior was considered another obstacle, since fewer chips are likely to hit an aggressive performance target and power budget on a sustained basis, particularly with all the safety margins that entails.
At the time, I asked the question rhetorically as to whether the high clock was compatible with the typical fixed-clock console target, and we now have the answer for Sony.

In some ways, variable clocks could give some breathing room for yields versus a chip that must sustain a clock target at all times. In that regard, the PS5 might do better than a comparable fixed-clock chip, depending on how close to the edge the silicon is pushed, or how much Sony allows the clock to deviate.

Sony needs to bin for peak boost clocks being advertised. From a binning perspective that doesn’t change anything. The yields will depend on voltage tolerance range for the bin, assuming the chip can hit the peak frequencies in the first place.

Since they’re throttling based on power draw, the greater the delta in voltage, the greater the performance delta will be between units or they will have to build their base profiles to the lowest performance envelope (highest chip voltage allowed) to ensure consistency in performance across unit.

 

[Xbat]

What Cerny has done on PS5 is actually smart and is the way forward in console gaming. Like the share button before I wouldn't surprised if Microsoft copy Sony again and do the same in their next console

Yup sounds like a no brainier but and it's a big but is the proof is in the pudding and we will see through the life of the PS5 if it's a sound decision or if there might be hidden issues that they didn't predict.

 

[Metal_Spirit]

If fixed clocks of 2.0 GPU and 3.0 CPU were hitting a power wall, how is it possible for the variable approach to constantly keep both near 2.23 and 3.5? Doesn't add up.

Trying to explain this and keeping it simple:

First you need to understand this:

Imagine your water supply pipeline.
Did you knew that if all the persons on your street, opened all fawcets at the same time, there would be no water coming out?
Why? Because pipes are not calculated by multiplying the number of fawcets by the desired water comming out of each.
There is a zero possibility of every person opening all fawcets at the same time.
So estimated consumption is made by average person consumption, making an estimation on regulation plans and soil occupation. And even then, only a percentage of that is used, because not all persons will use the watter at the same exact moment.

The same is made with power usage. Consoles could estimate power for a full 100% usage of all components, lock the speeds, and then make the cooling system.
You would end with a huge case, and a high cost cooling solution.
And for what? For nothing! That scenario will never happen. No game will ever use the GPU that way.

So, workloads are estimated using an average. They check what games are doing and using statistic methods make a determination on what power usage will happen, creating a cooling system for that.

This will happen on all consoles, PS5 or XsX.

Note that thermals are caused by power usage, and power usage comes as a result of workloads. And workloads also depend on clock frequency since the system will be making more calculation every second.

So a better way to try and keep power usage within control is to control also some variables. The most used one is the clock speed. You lock it, so that workloads can be more constant. Your output will no longer be dependable of clock speeds, but only on GPU components usage.

Even so, GPU clocks are locked bellow the maximum possible. And why? Because of thermal problems!

Since GPU components are not locked, an increase on their usage above the one expected when creating the cooling solution will increase power usage, so if the GPU usage goes above the estimated power usage, it will overheat.

But overheating cannot be deterministic. I mean your system cannot simply lock because the system went above the expected heat. So the chip has to have a margin. It can heat above the expected, and that's why fans are prepared do speed up.
Problem is, the cooling system is not efficient in that way. They were not optimized for those temperatures. And as such, if thermals keep above the ideal, the system will eventually overheat out of control, crash or reboot. Chips have then a safety precaution, where they will turn off if a certain temperature is achieved.

PS5 goes the other way around.

First the PS5 will not lock the speed variable or the workload variable. It will be deterministic and just lock the maximum power usage, also locking maximum temperature, and making a coolling system optimized for that temperature.
This is way more efficient on getting thermals under control.

This means PS5 is designed to never go above that temperature. But if by any chance, if thermals were badly calculated and temperatures do go up, even then it will not overheat. It will downclock a bit, reducing power usage and keeping thermals in place.

This means the Chip doesn´t need safety margins to handle increasing in temperatures, and as such you can squeeze the maximum speed out of it, that keeps it within your cooling system temperature control range.

Hope I made it clear.

 

[Kaotik]

That's irrelevant... ALL consoles are affected by chip quality, and ALL consoles are affected by power consumption. So ALL consoles might have problems with clocks or thermals.
Thats why ALL consoles are set withing a safety margin to account for those variations.

Sorry for the ALL in caps, but in the context of the conversation I have to make clear that what you mention is not a specific PS5 thing.

Of course it is not just a specific PS5 thing, but it's different for PS5 compared to XSX because the latter has locked clocks.
For XSX you just need your chips to reach set clocks, they're locked there and total consumption will vary between consoles.
PS5 instead uses variable clocks and according to Cerny it's power determined, so each individual console needs to be calibrated relative to it's specific power consumption instead of absolute power or each console would clock different on different loads