• Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.

Simuly

Alt-Account
Banned
Jul 8, 2019
1,281
True. To the mass market, all these spec stuff wouldn't matter and it always boils down to the games. And you are preaching to the choir cause I am buying a PS5 regardless. Already too entrenched into the ecosystem to back out now.

Just speaking from technology and design perspective. Of how weird it would be if sony made a machine that is weaker in every way except the SSD but still ends up costing the same with the XSX. That to me seems like what is a major design gaffe. I would feel better about my purchase if I felt I was by one hell of a $399 console, as opposed to me buying an overpriced "premium" $499 console.

At that point, it becomes like an apple and PC argument.

"But I have better specs at the same price, more flexibility and support to more advanced features and hardware"

"yh, but I am apple".

Well I think the sensationalism over tflops and specs is only in the echo chamber of Era. Sony's idea is that 10tf is plenty and to truly lift the gaming experience to 'next gen', whilst respecting a console budget, more money should be put into the SSD.

I would actually trust them on this one. As we'll get more or less the same experience on a 10 and 12tf machine, even if judging them on this silly metric is misleading. Let's see what devs can do first.

Also, I think theyll be priced the same...
 

Isayas

Banned
Jun 10, 2018
2,729
tbh, thats the series x. MS has done a fantastic job packing in 2080 ti performance in a $500 box (hopefully) with an ssd that should be 50x faster than last gen, a super fast 8 core 16 thread cpu, fast ram with some really intelligent ways to use ssd as virtual ram, and even a dedicated audio chip to make sure nothing in that system is a bottleneck.

the 10 gb of vram is the only bottleneck i can see, but the virtual ram stuff they have in there should work much like cerny's even if it's not as mind blowing.

also, if the 12 tflops gpu is offering 17 turing tflops performance than the 10 tflops gpu* should be around a 2080 super. it will be interesting to see just how efficient these rdna 2.0 tflops are.

* if it does stay at 10 tflops consistently. im not sold just yet.

The Series X has a 2080Ti like performance? HOW?
 

AegonSnake

Banned
Oct 25, 2017
9,566
The Series X has a 2080Ti like performance? HOW?
they are running gears 5 at native 4k 60 fps locked with even more effects that are not in the pc version which runs at native 4k 61 fps on the rtx 2080 ti. The rtx 2080 runs the game at native 4k 45 fps on ultra.

Minecraft path tracing demo is also running at similar franerates as the rtx 2080 ti. If anything it's probably even more powerful.
 

Isayas

Banned
Jun 10, 2018
2,729
they are running gears 5 at native 4k 60 fps locked with even more effects that are not in the pc version which runs at native 4k 61 fps on the rtx 2080 ti. The rtx 2080 runs the game at native 4k 45 fps on ultra.

Minecraft path tracing demo is also running at similar franerates as the rtx 2080 ti. If anything it's probably even more powerful.

When I have the time, I will watch that video. That is insane to me but these are consoles in a closed system.

It's more around a 2080 Super

A 2080 super can't do that.
 

BreakAtmo

Member
Nov 12, 2017
12,824
Australia
That right there would be the joke of this gen to me. And why I am almost certain sony sells the PS5 for less than the XSX even if its just to save face.

How messed up is that if they end up making a similarly priced console that ends up with a weaker CPU, weaker GPU, lower bandwidth and smaller SSD. And all they have to say for themselves is our SSD is 2x as fast??

Well, the advantage in SSD percentage-wise would probably be as big as all the XSX advantages put together, but yeah, that's not going to work on the casual customer and may not be as visible.

Honestly, I think the smaller SSD is the hardest proof that the PS5 is going to be cheaper than XSX, and probably $399. The percentage differences in CPU, GPU, bandwidth could stay a niche thing, but "825GB vs 1TB" is going to be right there on the boxes for all to see. They really should go $399 - if the PS5 is $100 cheaper, as well as being considerably smaller should their "efficient cooling" also be space-efficient, that's all they would need. Suddenly the response to XSX having that selection of technical advantages would go from "Wow, I guess that's the better console" to "Well, I'd fucking hope so". It's also the best way to counter the sandwich strategy - be dead in the center of a $299 Series S and a $499 Series X, but be much closer to the Series X in power.
 

disco_potato

Member
Nov 16, 2017
3,145
I dont think cooling will be as expensive for the PS5 as some here are making it out to be. Keep in mind the XSX is clocked a lot higher than what most here thought it be at, 1825mhz on a 52 CU GPU, PS5 is clocked 400mhz higher on a much smaller chip. Depending on how efficient RDNA2 is the cooling on PS5 wont necessarily need to be as expensive as whats on XSX, or maybe on par if RDNA2 isn't designed to be clocked that high. If the idea for PS5 was always be clocked that high, then I dont see the cooling being an issue.
The idea that PS5's cooling has to be $20 cheaper than xsx when xsx's cooling itself doesn't cost half that is a bit odd.
We don't know what the rdna2 efficiency curve looks like and until we do, it's probably a good idea to not make claims as to what is or isn't within norm. History does tell us however that consoles use low bin, high yield silicon.
 

nib95

Contains No Misinformation on Philly Cheesesteaks
Banned
Oct 28, 2017
18,498
In the real world the performance difference might be a bit higher. We know RDNA2 have some power efficiency increase, but we don't know how will be the efficiency of a GPU with 36CU running at 2.23GHz would be. At those frequencies the power required will be high and is the GPU is "power starved" the high clock speeds wont make much difference.

Whilst what you're saying is entirely possible, it also isn't particularly logical. Why would Sony pursue a much higher clock frequency that would itself require many additional considerations (eg higher thermal dynamics, more demanding cooling design and costs, extra power usage, riskier longevity, yields etc), if not for the fact that in real world testing they found it to offer tangible benefits? And why would they not give it the relevant voltages or power considerations to unlock that added potential performance?

We're not talking about some teenager overclocking their GPU for the first time (hence forgetting to adjust voltages to account for the extra power load), we're talking about a multi billion dollar tech company that actually helped make the GPU in the first place.

Sony didn't "simply incorporate" an AMD RDNA 2 GPU into the PS5, it helped build it

Sometimes it feels like some of you are for whatever reason alluding to or implying potential worst cases with the PS5, almost to a level of naivete, as if Sony were floundering with the hardware side of things and had flagrantly, not carefully decided some of these hardware choices.
 
Last edited:

xem

Member
Oct 31, 2017
2,043
their ssd requires lots of extra little things on the apu die as well. there is a good chance that it might be around the same size as the xbox series x.

H7gqWPY.png


supposedly, MS also as SRAM on their apu, but sony's solution seems to be far more complex and im afraid they used precious transistors for the ssd instead of the graphics card.
and I would much rather have a twice as fast SSD than 17% more GPU TF
 
Jan 19, 2020
636
Bellingham
and I would much rather have a twice as fast SSD than 17% more GPU TF
Same here. I feel like there are many more game design possibilities with a super fast SSD than there is with a bump in GPU power. This isn't to say I think one is clearly better than the other, just that, in my own personal preference of game design/philosophy over native 4k and locked 60 fps, the SSD represents an opportunity to change what we see as standard game design.
 

ThatNerdGUI

Prophet of Truth
Member
Mar 19, 2020
4,550
Whilst what you're saying is entirely possible, it also isn't particularly logical. Why would Sony pursue a much higher clock frequency that would itself require many additional considerations (eg higher thermal dynamics, more demanding cooling design and costs, extra power usage, riskier longevity, yields etc), if not for the fact that in real world testing they found it to offer tangible benefits? And why would they not give it the relevant voltages or power considerations to unlock that added potential performance?

We're not talking about some teenager overclocking their GPU for the first time (hence forgetting to adjust voltages to account for the extra power load), we're talking about a multi billion dollar tech company that actually helped make the GPU in the first place.

Sony didn't "simply incorporate" an AMD RDNA 2 GPU into the PS5, it helped build it

Sometimes it feels like some of you are for whatever reason alluding to or implying potential worst cases with the PS5, almost to a level of naivete, as if Sony were floundering with the hardware side of things and had flagrantly, not carefully decided some of these hardware choices.

This is why I mentioned we have to wait and see RDNA2 power improvements first. Sony could have trillions, but the can't change the laws of physics. There's lots of "ifs" going around, mostly outside the realms of possibilities. But I guess this is not a topic a VLSI design engineer will know anything about.
 

ShapeGSX

Member
Nov 13, 2017
5,210
We don't have Minecraft with DXR yet so I can't tell about that, but I can do better than Gears 5 on my 2080 Super.

They were running Gears 5 at 100fps. No mention of resolution though.

" There were also some noticeable improvements in a few other areas as well. Load times were extremely fast, and the team was able to turn on some features that, while previously implemented, had to be turned off for the Xbox One X version. This included contact shadows (providing extra depth to objects) and self-shadow lighting on plants and grass, making every scene feel more realistic. Rayner also shared that the game is already running over 100 FPS and that the team is investigating implementing 120 FPS gameplay for multiplayer modes, giving players an experience never before seen on consoles. Most impressive of all? The fact that the team was able to get all of this up and running in a matter of weeks."

news.xbox.com

Xbox Series X: A Closer Look at the Technology Powering the Next Generation - Xbox Wire

A few months ago, we revealed Xbox Series X, our fastest, most powerful console ever, designed for a console generation that has you, the player, at its center. When it is released this holiday season, Xbox Series X will set a new bar for performance, speed and compatibility, all while allowing...
 

nib95

Contains No Misinformation on Philly Cheesesteaks
Banned
Oct 28, 2017
18,498
This is why I mentioned we have to wait and see RDNA2 power improvements first. Sony could have trillions, but the can't change the laws of physics. There's lots of "ifs" going around, mostly outside the realms of possibilities. But I guess this is not a topic a VLSI design engineer will know anything about.

Ask yourself, do you honestly think Sony would pursue a much higher clockspeed with all the potentially negative and/or costly ramifications that come with it, if in the end it really "won't make much difference" to real world performance, and then also "power starve" the GPU so it makes even less of a performance advantage on top?

Lest we ignore the fact that Cerny actually specifically spoke about some of the advantages the higher clockspeed offers in his presentation, and hints at why they decided on it.
 
Last edited:

ThatNerdGUI

Prophet of Truth
Member
Mar 19, 2020
4,550
They were running Gears 5 at 100fps. No mention of resolution though.

" There were also some noticeable improvements in a few other areas as well. Load times were extremely fast, and the team was able to turn on some features that, while previously implemented, had to be turned off for the Xbox One X version. This included contact shadows (providing extra depth to objects) and self-shadow lighting on plants and grass, making every scene feel more realistic. Rayner also shared that the game is already running over 100 FPS and that the team is investigating implementing 120 FPS gameplay for multiplayer modes, giving players an experience never before seen on consoles. Most impressive of all? The fact that the team was able to get all of this up and running in a matter of weeks."

news.xbox.com

Xbox Series X: A Closer Look at the Technology Powering the Next Generation - Xbox Wire

A few months ago, we revealed Xbox Series X, our fastest, most powerful console ever, designed for a console generation that has you, the player, at its center. When it is released this holiday season, Xbox Series X will set a new bar for performance, speed and compatibility, all while allowing...

As you said, no mention of resolution or settings. Also, paying through the game is much less intensive than the in-game benchmark, the multiplayer is even less than that.
 

ThatNerdGUI

Prophet of Truth
Member
Mar 19, 2020
4,550
Ask yourself, do you honestly think Sony would pursue a much higher clockspeed with all the potentially negative and/or costly ramifications that come with it, and then also "power starve" the GPU so it makes even less of a performance advantage on top, if in the end it really "didn't make much difference" to real world performance?

If they have the power budget to reach the frequency at a power lower than the optimal power required at that frequency but still get a couple of percent extra, yes they would. This is why you can't just compare the performance of each console based on TF numbers. Those numbers are just being used for marketing.
 

BradGrenz

Banned
Oct 27, 2017
1,507
their ssd requires lots of extra little things on the apu die as well. there is a good chance that it might be around the same size as the xbox series x.

H7gqWPY.png


supposedly, MS also as SRAM on their apu, but sony's solution seems to be far more complex and im afraid they used precious transistors for the ssd instead of the graphics card.

That image isn't to scale. The I/O complex in PS5 is going to be very small compared to the die space required for the much larger GPU in the XSX.
 

nib95

Contains No Misinformation on Philly Cheesesteaks
Banned
Oct 28, 2017
18,498
If they have the power budget to reach the frequency at a power lower than the optimal power required at that frequency but still get a couple of percent extra, yes they would. This is why you can't just compare the performance of each console based on TF numbers. Those numbers are just being used for marketing.

But it isn't just about power budget, but also thermal dynamics and cooling, hardware longevity and more. That stuff doesn't come free.

These much higher clocks require far more power (by Cerny's own words, a 2% drop in max frequency lends to 10% less power usage) and most probably also much better cooling, hence they have actual manufacturing, design and efficiency costs associated with them. As just one example, if a lower frequency clock allows you better yields and to include a less exotic cooling solution, over the long term you've already saved millions in manufacturing costs just there. Ultimately, you don't just willy nilly decide to push much higher frequencies unless there's actual tangible benefits to doing so.
 
Oct 27, 2017
7,135
Somewhere South
I don't think they're thermally constrained at top frequencies for your average gameplay load (a very important detail people seem to miss, iso freqs not all loads draw the same power, gameplay stuff is generally less efficient than, say, a synthetic bench) else Cerny wouldn't point out they actually had to cap the freqs because it could go further up, but would break the chip logic (i.e. logic gates not switching fast enough, at least not without a meaningful increase in voltages).
 

nib95

Contains No Misinformation on Philly Cheesesteaks
Banned
Oct 28, 2017
18,498
I don't think they're thermally constrained at top frequencies for your average gameplay load (a very important detail people seem to miss, iso freqs not all loads draw the same power, gameplay stuff is generally less efficient than, say, a synthetic bench) else Cerny wouldn't point out they actually had to cap the freqs because it could go further up, but would break the chip logic (i.e. logic gates not switching fast enough).

I do appreciate this, but then these types of frequency clocks are almost historically unheard of in GPU's, and whilst it's true that RDNA2 might just be far more efficient in that regard, it does then beg the question of why Microsoft wasn't simply able to push a higher clock with their system, especially given their extremely impressive cooling solution. A 22% frequency clock speed difference in the PS5's favour isn't exactly insignificant.
 

Crayon

Member
Oct 26, 2017
15,580
The 2% clock only equals 10% power when the workload saturates it, right? If the workload is partial, then that pulling that 2% would save 10%? Im not a computer scientist yet. Need halp.
 
Oct 27, 2017
7,135
Somewhere South
it does then beg the question of why Microsoft wasn't simply able to push a higher clock with their system, despite their extremely impressive cooling solution. A 22% frequency clock difference isn't exactly small.

Exactly because they went with the traditional way of doing console GPUs: "locked" frequencies with floating power load. Going further up probably means they'd have to vastly overprovision their already oversized cooling system to deal with corner cases where the GPU is going through a very efficient load and shooting power draw up through the roof.

If the PS5 sees one of these loads (it does power balancing based on loads) it just says "nope, not doing that" and throttles the GPU. Also, quite likely that whatever cooling Sony is doing there that allows it to cruise at those freqs is just bonkers.

The 2% clock only equals 10% power when the workload saturates it, right? If the workload is partial, then that pulling that 2% would save 10%? Im not a computer scientist yet. Need halp.

It wouldn't, no. But it's likely that, if it isn't a saturating load, it just won't need to drop at all.
 

Dokkaebi G0SU

Member
Nov 2, 2017
5,922
I'm still waiting for more info on their variable clocks, power consumption, worst case scenario detailed blog.
Or maybe a developer to advise how this benefits them when making games.

Are there any leaks of more Sony PS5 info coming out soon?
 

ThatNerdGUI

Prophet of Truth
Member
Mar 19, 2020
4,550
But it isn't just about power budget, but also thermal dynamics and cooling, hardware longevity and more. That stuff doesn't come free.

These much higher clocks require far more power (by Cerny's own words, a 2% drop in max frequency lends to 10% less power usage) and most probably also much better cooling, hence they have actual manufacturing, design and efficiency costs associated with them. As just one example, if a lower frequency clock allows you better yields and to include a less exotic cooling solution, over the long term you've already saved millions in manufacturing costs just there. Ultimately, you don't just willy nilly decide to push much higher frequencies unless there's actual tangible benefits to doing so.

Yes, I say this assuming they can keep a reasonable temperature which I assume they do because that's the first thing you have to do before yo can push clocks that high. Like I said there can be benefits, but are they optimal? 2% in frequency being 10% power drop is a lot. What are the required load conditions, etc. My point is they could do it, but the performance gained might not be as great as people want to make it seem or believe. We would have to wait until the consoles either come out or we see the power improvements when RDNA2 GPU's come out. Until then, people need to keep their expectations in check.
 

nib95

Contains No Misinformation on Philly Cheesesteaks
Banned
Oct 28, 2017
18,498
Exactly because they went with the traditional way of doing console GPUs: "locked" frequencies with floating power load. Going further up probably means they'd have to vastly overprovision their already oversized cooling system to deal with corner cases where the GPU is going through a very efficient load and shooting power draw up through the roof.

If the PS5 sees one of these loads (it does power balancing based on loads) it just says "nope, not doing that" and throttles the GPU.

Fair point. I guess it falls on us to just wait and see how this increase in frequency clock actually impacts performance, and also how this new boost "paradigm" that the PS5 had based on a set power load instead of thermal dynamics, actually benefits or impacts performance in real world terms. I just can't imagine they'd pursue such a unique design and set up unless there were genuine tangible benefits.
 

vivftp

Member
Oct 29, 2017
19,747
All this talk about the PS5's cooling, are most people assuming they're using a design similar to that patent from last year with the heatsinks on both sides, or something even more exotic?
 

tusharngf

Member
Oct 29, 2017
2,288
Lordran
they are running gears 5 at native 4k 60 fps locked with even more effects that are not in the pc version which runs at native 4k 61 fps on the rtx 2080 ti. The rtx 2080 runs the game at native 4k 45 fps on ultra.

Minecraft path tracing demo is also running at similar franerates as the rtx 2080 ti. If anything it's probably even more powerful.

thats insane.
 

D BATCH

Member
Nov 15, 2017
148
Have you ever used one? They said the in-game benchmark the Series X have an average of 46fps with maxed settings at 4k. I get ~54fps avera at 4k max settings. If that makes it a special edition then sure.
Yes that is accurate. Here is a chart below of 4k benchmarks. Still XBSX was running the game at 2080ti 4K FPS with higher than PC settings and 50% more particles is impressive. Just imagine how good the PC card will be if the APU is capable of even coming close or in this case beating a 2080ti
pmcCXPA.png
 

ThatNerdGUI

Prophet of Truth
Member
Mar 19, 2020
4,550
Yes that is accurate. Here is a chart below of 4k benchmarks. Still XBSX was running the game at 2080ti 4K FPS with higher than PC settings and 50% more particles is impressive. Just imagine how good the PC card will be if the APU is capable of even coming close or in this case beating a 2080ti
pmcCXPA.png

In the DF video they clearly say the in-game benchmark had an average of ~46fps which is below the 2080ti in the graph you just showed. That doesn't stop it from being impressive, but people really need to keep the expectations in check. The whole "Series X shows 2080ti performance" its just not accurate.
 

D BATCH

Member
Nov 15, 2017
148
In the DF video they clearly say the in-game benchmark had an average of ~46fps which is below the 2080ti in the graph you just showed. That doesn't stop it from being impressive, but people really need to keep the expectations in check. The whole "Series X shows 2080ti performance" its just not accurate.
I think that was after 2 weeks . currently is running at 60fps and they have it at 100fps in multi-player
 

Key222

Member
Dec 11, 2017
148
Ya probably closer to 2070 super so 2070-2070 super levels

Isn't a 2080ti roughly 14 Tflops and a 2070 Super 9 Tflops? If so, your comparison still doesn't make any sense if you're equating the XSX GPU to a 2080 ti or better. You're saying a 10TF AMD GPU equates to a 9TF or less nVidia GPU but a 12TF AMD GPU equates to a 14TF or better nVidia GPU. Basically, that a 2TF difference for AMD GPU's equals to a 5TF or more difference for nVidia GPUs.

The same benchmark that showed nearly a 30% performance difference between the 2080 Ti (~14TF) and the 2070 Super (~9TF) for Gears 5 at 4K Ultra, only shows a difference of less than 10% for the 5700 (~7.9TF) vs the 5700XT (~9.7TF). I imagine ~1.9TF difference between the two AMD GPUs would be far more comparable to the power difference between the PS5 and XSX GPU (percentage-wise) than the ~5TF difference between the two nVidia GPUs.
 
Last edited: