• 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.

Matt weighs in on PS5 I/O, PS5 vs XSX and what it means for PC.

Journey

Banned
18% difference or a 2TF is not a lot when we're comparing a 10TF GPU to a 12TF GPU.

RTX 2070 Super 9TF
RTX 2080 Super 11.1TF



PS4 and XBOX GPUs are about 40% apart. This is not the case with the XSX and PS5.

Anyone who has been gaming on PC knows that reaching that target frame-rate will only result in a small reduction of resolution.

You can watch reviews when it comes to the difference between both GPUs and they'll tell you that the difference is not big, it's small.

XSX should have better RT due to more CUs, but we still don't know how much of a better performance the PS5 is going to receive due to a higher clock.

What's the tefaflop difference between the 2080 S and TI?


Never mind, I looked it up

13.45 for the 2080 TI
11.15 for the 2080 Super

The difference between those 2 is quite staggering with certain games.

Final edit

I can see why, one has 496GB/s Bandwidth (Super) vs 616GB/s Bandwidth (TI)

448GB/s (PS5) vs 560GB/s (XSX) of memory bandwidth for VRAM is something that's often overlooked, it's not just a 2TF difference (At best case scenario) but a difference of 112GB/s for the portion of XSX dedicated to VRAM, that can have a significant impact.
 
Last edited:

DForce

Member
What's the tefaflop difference between the 2080 S and TI?


Never mind, I looked it up

13.45 for the 2080 TI
11.15 for the 2080 Super

The difference between those 2 is quite staggering with certain games.

Final edit

I can see why, one has 496GB/s Bandwidth (Super) vs 616GB/s Bandwidth (TI)

448GB/s vs 560GB/s of memory bandwidth is something that's often overlooked, that can have a significant impact.

the 2TF count is not significant, no matter what you say.

It's strange that people are ignoring devs, tech experts etc. whenever they say the TF count is not significant. But people on here think there's some huge gap even though we have proof that a 2TF difference is not a huge gap, it's small.

If it's smaller than XB1\PS4 and XB1X and PS4 Pro, then it's not big.
 

Nikana

Go Go Neo Rangers!
What's the tefaflop difference between the 2080 S and TI?


Never mind, I looked it up

13.45 for the 2080 TI
11.15 for the 2080 Super

The difference between those 2 is quite staggering with certain games.

Final edit

I can see why, one has 496GB/s Bandwidth (Super) vs 616GB/s Bandwidth (TI)

448GB/s (PS5) vs 560GB/s (XSX) of memory bandwidth for VRAM is something that's often overlooked, it's not just a 2TF difference (At best case scenario) but a difference of 112GB/s for the portion of XSX dedicated to VRAM, that can have a significant impact.

Nah bro it doesn't make a difference. It's smaller than before so it doesn't matter. Games will be the same. SSD.

Summed up every response you will get.
 

MCplayer

Member
the 2TF count is not significant, no matter what you say.

It's strange that people are ignoring devs, tech experts etc. whenever they say the TF count is not significant. But people on here think there's some huge gap even though we have proof that a 2TF difference is not a huge gap, it's small.

If it's smaller than XB1\PS4 and XB1X and PS4 Pro, then it's not big.
The CU count definitly matter in RT and RAM on Xbox also has a bigger Bus Width
 
Last edited:

Journey

Banned
the 2TF count is not significant, no matter what you say.

It's strange that people are ignoring devs, tech experts etc. whenever they say the TF count is not significant. But people on here think there's some huge gap even though we have proof that a 2TF difference is not a huge gap, it's small.

If it's smaller than XB1\PS4 and XB1X and PS4 Pro, then it's not big.

I'm not ignoring what Matt just said, but maybe you are? Did you ignore some of his points? here let me help you:

I have a ton of good to say about the SX. It’s an fantastic box that I’m incredibly excited for, and it has a power advantage over the PS5 in many ways. It is also capable of doing things the PS5 can’t or can’t do as well.

They are both great machines and can both be appreciated.
DirectStorage is great for what it is, but yeah, I expect brute force speed to be the biggest factor here.

The SX more conventional than the PS5, and therefore less novel to talk about. It’s still a stronger system in many ways.


Sounds like you're ignoring some points 🤔
 

DForce

Member
I'm not ignoring what Matt just said, but maybe you are? Did you ignore some of his points? here let me help you:






Sounds like you're ignoring some points 🤔
I wasn't just referring to Matt, but devs commenting on the difference between TFs.

You assumed he was talking about TF.
As I have said before, I expect the difference in third party titles to be modest, as they can’t be designed around a faster solution. Maybe the PS5’s IO advantages will be as noticeable as the SX’s TF advantages in those titles.

So are you saying Matt is wrong now after reading this quote?
 
Last edited:

DForce

Member
One is a maybe, the other is a certainty 🤷‍♂️
He's always said the TF wasn't significant. That means he's referencing the I/O about how it may be as small as the TF count on third party titles. That means he believes is possible that the O/I can show noticeable differences.

Fact still remains that it's not a huge gap. NX Gamer, RedGamingTech, Digital Foundry, devs and others.
 

Kenpachii

Member
What's the tefaflop difference between the 2080 S and TI?


Never mind, I looked it up

13.45 for the 2080 TI
11.15 for the 2080 Super

The difference between those 2 is quite staggering with certain games.

Final edit

I can see why, one has 496GB/s Bandwidth (Super) vs 616GB/s Bandwidth (TI)

448GB/s (PS5) vs 560GB/s (XSX) of memory bandwidth for VRAM is something that's often overlooked, it's not just a 2TF difference (At best case scenario) but a difference of 112GB/s for the portion of XSX dedicated to VRAM, that can have a significant impact.

2080ti at 1700mhz is some serious slow performing 2080ti.
 
Last edited:
18% difference or a 2TF is not a lot when we're comparing a 10TF GPU to a 12TF GPU.

RTX 2070 Super 9TF
RTX 2080 Super 11.1TF



PS4 and XBOX GPUs are about 40% apart. This is not the case with the XSX and PS5.

Anyone who has been gaming on PC knows that reaching that target frame-rate will only result in a small reduction of resolution.

You can watch reviews when it comes to the difference between both GPUs and they'll tell you that the difference is not big, it's small.

XSX should have better RT due to more CUs, but we still don't know how much of a better performance the PS5 is going to receive due to a higher clock.

Yes we do.

DF did a comparison of normal and overclocked 5700 cards. The overclocked lower cu cards almost never beat the slower clocked higher cards.

Madz gaming also did a comparison review yesterday showing low cu high clock vs high cu lower clocked cards... the trend holds for framerate and resolution almost universally.

We pretty much know how this race begins and ends...but at the end of the day so what? Neither resolution nor framerate are definers of quality.

PS5 will have spectacular games that probably couldn't be replicated on any other device currently or planned. And thats great too.
 

DForce

Member
Yes we do.

DF did a comparison of normal and overclocked 5700 cards. The overclocked lower cu cards almost never beat the slower clocked higher cards.

Madz gaming also did a comparison review yesterday showing low cu high clock vs high cu lower clocked cards... the trend holds for framerate and resolution almost universally.

We pretty much know how this race begins and ends...but at the end of the day so what? Neither resolution nor framerate are definers of quality.

PS5 will have spectacular games that probably couldn't be replicated on any other device currently or planned. And thats great too.

You mean RDNA 1 cards that doesn't scale as good as RDNA 2 cards?

This was stated in DF video.


Why are you guys saying things differently than what these guys are saying? You guys are trying hard to push this narrative.


2TF difference is not significant.

Tell me why you guys are saying things totally different than what they're saying?
 
You mean RDNA 1 cards that doesn't scale as good as RDNA 2 cards?

This was stated in DF video.


Why are you guys saying things differently than what these guys are saying? You guys are trying hard to push this narrative.


2TF difference is not significant.

Tell me why you guys are saying things totally different than what they're saying?

So you would rather use an imaginary in your mind yardstick of performance rather than using benchmark performance to estimate?

You think that RDNA2 is gonna allow 36cu to scale the gap of 52 cus?Because RDNA2 is more magical than Rdna 1?

Ok
 

DForce

Member
So you would rather use an imaginary in your mind yardstick of performance rather than using benchmark performance to estimate?

You think that RDNA2 is gonna allow 36cu to scale the gap of 52 cus?Because RDNA2 is more magical than Rdna 1?

Ok
There's no imaginary yard stick. You're complete ignoring what other devs are saying. They literally said they have to test RDNA 2 cards and how data is not complete.

You're dodging these factors. Now tell me again how they're wrong about the TF count now being big.
 
Last edited:

Psykodad

Banned
This sounds reasonable. They just fail to also say lower framerate and lower resolution.
There are advantages and drawbacks to each console.
Who claimed equal or better resolution and resolution on top of the SSD benefits?
 
Last edited:

Psykodad

Banned
the 2TF count is not significant, no matter what you say.

It's strange that people are ignoring devs, tech experts etc. whenever they say the TF count is not significant. But people on here think there's some huge gap even though we have proof that a 2TF difference is not a huge gap, it's small.

If it's smaller than XB1\PS4 and XB1X and PS4 Pro, then it's not big.
I remember a lot of gaffers claiming the killzone and infamous reveal visuals wouldn't ever be possible on PS4 hardware, back in 2013.

Most of those with technical knowledge are just as clueless as the rest of us.
 
I doubt PS5 can sustain 10TF all the time. When the PS5 drops into the 9's, we're looking at a 3TF difference, then there's the CPU also having a slight advantage, more so with earlier games not needing 16 threads, up to 300Mhz difference. We'll see, Matt says overall XSX is more powerful and can either do things better or things the PS5 can't do at all. Interesting and fun times ahead.

These TF numbers are theoretical for only one aspect of the system (GPU-compute) and do not represent real workloads, which typically do not come close to full utilization. Performance profilers are used to help developers achieve higher utilization. It is my understanding that even the best games only achieve around 70-75% on the PS4 due to bottlenecks associated with CPU contention. A more efficient system can achieve higher utilization, and real world performance could even show a weaker system out performing a stronger one. Games designed around the PS5’s I/O system should demonstrate real noticeable benefits.

I expect typical cross-platform games to be within 10% of each other for next gen consoles, with the differences very hard to discern. With a fixed frame rate and variable resolution, the stronger GPU would probably spend slightly more time at the maximum resolution.
 

Naddy

Banned
I remember a lot of gaffers claiming the killzone and infamous reveal visuals wouldn't ever be possible on PS4 hardware, back in 2013.

Most of those with technical knowledge are just as clueless as the rest of us.

I think killzone is a really bad example, I mean remember what they did with millions 2?

Hopefully they don’t do this shit again.
 
There's no imaginary yard stick. You're complete ignoring what other devs are saying. They literally said they have to test RDNA 2 cards and how data is not complete.

You're dodging these factors. Now tell me again how they're wrong about the TF count now being big.

Im not ignoring anything. Across every modern architecture, that's not ampere or RDNA2, we see this same pattern.

What makes you think that RDNA2, especially PS5s RDNA implementation, will the one circumstance where it isn't?

The architectural gains will apply to both chips in both systems.
 
Last edited:

DForce

Member
Im not ignoring anything. Across every modern architecture, that's not ampere or RDNA2, we see this sane pattern.

What makes you think that RDNA2 especially PS5s RDNA implementation will the the one circumstance where it isn't?

The architectural gains will apply to both chips in both systems.
You are. You still haven't addressed the devs saying 2TF gap is not big.

Are they being dishonest or not?

Digital Foundry said in the test video that they need to test a card with MORE than 40CUs. The RDNA 2 is much wider in comparison to what's based on the RDNA 1 Navi 10 chip.

So does XSX also took a similar approach for going with a higher clock with a few less CUs. Still this part is irreverent.

The fact still remains that 2TF Gap is not huge when it's 10TF vs 12TF.
 

Naddy

Banned
You are. You still haven't addressed the devs saying 2TF gap is not big.

Are they being dishonest or not?

Digital Foundry said in the test video that they need to test a card with MORE than 40CUs. The RDNA 2 is much wider in comparison to what's based on the RDNA 1 Navi 10 chip.

So does XSX also took a similar approach for going with a higher clock with a few less CUs. Still this part is irreverent.

The fact still remains that 2TF Gap is not huge when it's 10TF vs 12TF.


In comes a new tweet from a former PlayStation developer, who says to have heard from several developers that Microsoft’s console is the superior console and that the power difference between the Xbox Series X and PS5 is quite shocking.


also, its not only About TFLOPS, PS5 has a variable frequency for CPU and GPU.
XSX is sustained for both. Also, XSX has a higher memory bandwidth.
and much more CUs for RT
Also, the CPU clock is higher
 
Last edited:

DForce

Member

also, its not only About TFLOPS, PS5 has a variable frequency for CPU and GPU.
XSX is sustained for both. Also, XSX has a higher memory bandwidth.
and much more CUs for RT
Also, the CPU clock is higher
Same dev who said PS5 will struggle with open world games.

You have him vs many others.

He also said only one console can only do real time ray tracing. How is that even possible when PS5 is confirmed to do Ray Tracing?


You guys are running in circles trying to make it appear like there's some huge gap. I can quote many developers and tech analysis stating otherwise.

You guys? You guys want to continue to disagree with them.
 

Shmunter

Gold Member
Same dev who said PS5 will struggle with open world games.

You have him vs many others.

He also said only one console can only do real time ray tracing. How is that even possible when PS5 is confirmed to do Ray Tracing?


You guys are running in circles trying to make it appear like there's some huge gap. I can quote many developers and tech analysis stating otherwise.

You guys? You guys want to continue to disagree with them.
Bro, bro., Let them dream (tm)

Look, XsX is a solid console, and exactly what pretty much anyone would have expected from a new gen. It is just that Sony went nuclear with their philosophy of what a system should be like. It’s not hard to feel excited by the prospect of real change, but we need to be sympathetic to the xbros who may feel like there’s a party and they weren’t invited. It’s not a time to point and laugh, it is a time to spread the joy.
 

ToadMan

Member
factually, wrong. Epic usually shows a tech demo when a new playstation console launches, they did the same with PS4, look here:


I think they have a marketing contract.

The first time UE4 was shown was in 2012. Running on a PC.

Same for UE3 - first showing was PC.

Not same for UE5 though, because that’s where it looks best right now.
 
Last edited:
factually, wrong. Epic usually shows a tech demo when a new playstation console launches, they did the same with PS4, look here:


I think they have a marketing contract.
One year later
You gotta be shitting me. Pray tell how a SSD processes pixels on the screen?
You can process heavy assets.
I doubt PS5 can sustain 10TF all the time. When the PS5 drops into the 9's, we're looking at a 3TF difference, then there's the CPU also having a slight advantage, more so with earlier games not needing 16 threads, up to 300Mhz difference. We'll see, Matt says overall XSX is more powerful and can either do things better or things the PS5 can't do at all. Interesting and fun times ahead.
So bases in your doubt we have to ignore what Cerny and other devs are already said, great then I can think XSX cannot sustain 12 TF .... that doesn't sounds logic, right ?
 

ToadMan

Member
There is more to a System than just the SSD...
There is a reason why cerny never mentioned Raytracing or said nonsense like TF dont matter

You should watch the video. He “mentioned” (in fact talked at length about RT) both of those things.

Of course he didn’t use the exact wording you’ve written here - “TF dont matter” - he has a better command of language than that.
 

ToadMan

Member
Tons of people said one X couldn't hit 4k with a 1.8 TF difference and they did many times. Comparing percents isn't always fair.

RDR2

Xb1 1.3 tflops vs PS4 1.8tflops.

0.5 tflops difference/40% relative.

= 1 step of resolution, same FPS. Xb1 900p, PS4 1080.

Xb1x 6tflops ps4pro 4.2tflops

1.8tflops/40% relative.


= 1 step resolution difference. 1920x2160 ps4p 4k xb1x

FPS largely similar.


A jump of one resolution “step” is about 40% more pixels. That’s why 0.5 tflops and 1.8tflops both just get a 1 step improvement in resolution and frames stay the same.


Xsex has 1.8tflops more - same as xb1x vs ps4p.

But that is only 15% less tflops to PS5 this time or 18% more to Xsex if you wish to calculate that way. Not enough for a full resolution bump and maintain FPS on Xsex.

Xsex will see same res and in ideal scenario 3 or 4 FPS more if the port process doesn’t lose them (or they try to do more RT on Xsex - then it will run lower res and frames than PS5).
 

ToadMan

Member
You do know that XSX also had a SSD and great IO etc.? It's not like XSX is using an old HDD.

Sure Xsex has a next gen SSD which is awesome by “normal” standards and of course people were greatly excited by it. MS are to be applauded for producing such a powerful storage solution - it’s that which has pushed Sony to even greater heights.

PS5 has a next next gen solution. It’s an order of magnitude better performance than Xsex offers.

That is something special for consumers - competition pushes both sides to achieve and Sony have come out on top. Without MS, we wouldn’t being seeing Sony excel like this. I hope that MS don’t lose heart from the losses they’re taking in gaming - it’s important they keep throwing their money into the market.

That Sony are offering their SSD tech while also maintaining or exceeding parity of processing power and a line up of quality pS5 exclusive games is really incredible considering console gaming is supposedly “dying“.
 
Last edited:

Naddy

Banned
RDR2

Xb1 1.3 tflops vs PS4 1.8tflops.

0.5 tflops difference/40% relative.

= 1 step of resolution, same FPS. Xb1 900p, PS4 1080.

Xb1x 6tflops ps4pro 4.2tflops

1.8tflops/40% relative.


= 1 step resolution difference. 1920x2160 ps4p 4k xb1x

FPS largely similar.


A jump of one resolution “step” is about 40% more pixels. That’s why 0.5 tflops and 1.8tflops both just get a 1 step improvement in resolution and frames stay the same.


Xsex has 1.8tflops more - same as xb1x vs ps4p.

But that is only 15% less tflops to PS5 this time or 18% more to Xsex if you wish to calculate that way. Not enough for a full resolution bump and maintain FPS on Xsex.

Xsex will see same res and in ideal scenario 3 or 4 FPS more if the port process doesn’t lose them (or they try to do more RT on Xsex - then it will run lower res and frames than PS5).

You can’t compare GCN TFLOPS to RDNA 2.0 tflops.
Thing is with current is that the TFLOPS a) couldn't be really used due to weak CPU and b) those are GCN Tflops, that's why it wasn't such a huge deal this gen.

But this completely changes nextgen, because:


1. We do not have a CPU Bottleneck anymore, so we can make much more use of the TFLOPs.

2. RDNA 2.0 is much more efficient than GCN (current gen), so basically, 1 Tflop of RDNA 2.0 is as powerful as 2 TFLops of GCN. So, you can still do a lot more work with 2TF's RDNA2 than you can with 500GF's of GCN.

Also, a difference the whole resolution and FPS is much more noticeable than having a little better textures.
 

II_JumPeR_I

Member
I have a dejavu with that "bEtTeR AuDiO" nonsense...
Some of you hopefully know that Xbox has dedicated audio Hardware too? 3D audio/Athmos etc isnt new for xbox either
 
You can’t compare GCN TFLOPS to RDNA 2.0 tflops.
Thing is with current is that the TFLOPS a) couldn't be really used due to weak CPU and b) those are GCN Tflops, that's why it wasn't such a huge deal this gen.

But this completely changes nextgen, because:


1. We do not have a CPU Bottleneck anymore, so we can make much more use of the TFLOPs.

2. RDNA 2.0 is much more efficient than GCN (current gen), so basically, 1 Tflop of RDNA 2.0 is as powerful as 2 TFLops of GCN. So, you can still do a lot more work with 2TF's RDNA2 than you can with 500GF's of GCN.

Also, a difference the whole resolution and FPS is much more noticeable than having a little better textures.
You know than also PS5 is RDNA 2 and use Zen 2 that why he compare systems in similar scale as if you compare TF GCN vs TF GCN
is a correct analogy for now.

So RDNA2 TF vs RDNA2 TF only will establish the peak theoretical performance of both (one which is never reached).

Do you really think you will have a jump in resolution bigger than the delta in TF ? I mean common this is not work in that way.

If you want to compare TF of two similar architectures then you have to do it using a porcentaje difference not just think that
18% delta in TF will become much more because now the TF in XSX can do more compare to PS5.... do you never touch a computer
for program something or what?
 
I have a dejavu with that "bEtTeR AuDiO" nonsense...
Some of you hopefully know that Xbox has dedicated audio Hardware too? 3D audio/Athmos etc isnt new for xbox either
Yes because this how works you only said has a audio chip and then you are equals in power to all the audio chips in the market, we already heard that
Tempest engine is greath chip from some devs more than those which mention the chip inside XSX.

If the chip inside the XSX was more powerful they would have said that, they love to use metrics like flops as marketing but just look them quiet.
 

FranXico

Member
I doubt PS5 can sustain 10TF all the time. When the PS5 drops into the 9's, we're looking at a 3TF difference, then there's the CPU also having a slight advantage, more so with earlier games not needing 16 threads, up to 300Mhz difference. We'll see, Matt says overall XSX is more powerful and can either do things better or things the PS5 can't do at all. Interesting and fun times ahead.
There is no such thing as "sustained TF". This kind of replies is just ludicrous.
 

longdi

Banned
There is no such thing as "sustained TF". This kind of replies is just ludicrous.

Yes there is.

While TF is not a full picture, and even less use in PC space. But it is still a simple metric to identify and visualize the power difference.
You can even use Mark Sony 'float all boat' idea.

The sustained here means the ability to deliver 12TF of floating boats constantly without throttling.
For PS5 to be able to deliver 10TF of floating boats, it needs to run the GPU at 2.23ghz constantly.
 

longdi

Banned
Yes because this how works you only said has a audio chip and then you are equals in power to all the audio chips in the market, we already heard that
Tempest engine is greath chip from some devs more than those which mention the chip inside XSX.

If the chip inside the XSX was more powerful they would have said that, they love to use metrics like flops as marketing but just look them quiet.

The tempest engine will also draw from PS5 448gbs memory bandwidth. Mark actually warned that if not carefully used, it will leech too much.
 

longdi

Banned
RDR2

Xb1 1.3 tflops vs PS4 1.8tflops.

0.5 tflops difference/40% relative.

= 1 step of resolution, same FPS. Xb1 900p, PS4 1080.

Xb1x 6tflops ps4pro 4.2tflops

1.8tflops/40% relative.


= 1 step resolution difference. 1920x2160 ps4p 4k xb1x

FPS largely similar.


A jump of one resolution “step” is about 40% more pixels. That’s why 0.5 tflops and 1.8tflops both just get a 1 step improvement in resolution and frames stay the same.


Xsex has 1.8tflops more - same as xb1x vs ps4p.

But that is only 15% less tflops to PS5 this time or 18% more to Xsex if you wish to calculate that way. Not enough for a full resolution bump and maintain FPS on Xsex.

Xsex will see same res and in ideal scenario 3 or 4 FPS more if the port process doesn’t lose them (or they try to do more RT on Xsex - then it will run lower res and frames than PS5).

This post will not age well. :pie_smirking: 🤙

But keep your 'micro-center' movement going
 
Last edited:

sinnergy

Banned
Yes because this how works you only said has a audio chip and then you are equals in power to all the audio chips in the market, we already heard that
Tempest engine is greath chip from some devs more than those which mention the chip inside XSX.

If the chip inside the XSX was more powerful they would have said that, they love to use metrics like flops as marketing but just look them quiet.
Xbox had DSPs for ages in their consoles , hardware blocks , so it’s nothing new really .
 
You can’t compare GCN TFLOPS to RDNA 2.0 tflops.
Thing is with current is that the TFLOPS a) couldn't be really used due to weak CPU and b) those are GCN Tflops, that's why it wasn't such a huge deal this gen.

But this completely changes nextgen, because:


1. We do not have a CPU Bottleneck anymore, so we can make much more use of the TFLOPs.

2. RDNA 2.0 is much more efficient than GCN (current gen), so basically, 1 Tflop of RDNA 2.0 is as powerful as 2 TFLops of GCN. So, you can still do a lot more work with 2TF's RDNA2 than you can with 500GF's of GCN.

Also, a difference the whole resolution and FPS is much more noticeable than having a little better textures.

Let's use Cernys math to indicate the true difference since the 18% difference has reached mythological status.

Since everyone wants to compare the difference between prior generations of GCN to this generation lets just do THAT. Remember "Cerny math."

Cerny said that 36 RDNA CUs is equivalent to 58 GCN CUs. That is because RDNA does 1.6 times as much work.

So that means that when comparing the diffeence between xb1 and ps4 and xb1x and ps4pro and THIS GEN you have to use that multiplier.

Cerny is a damn genius.

So 36 PS5 RDNA2 CUis = 58 GCN CUs
Thus 52 XSX RDNA CUs = 84 GCN CUs

44% rears its head again as ratios dont change. But what about teraflops which is the basis of this 18% chatter?

Remember we are looking at RDNA2 including its efficiency multiplier of 1.6 and calculating against GCN. CERNY MATH.

36 RDNA2 CU *2*64* 2.23 = 10.275 TFlop
58 GcN CU *2*64*2.23 = 16,555 Tflops

The PS5GPU is the equivalent of a 16.5 TFLop GCN card. NICE.

52 RDNA2 CU *2*64*1.875= 12.147 TF
84 GCN CU *2 * 64*1.875 = 20.625 TF

The difference is closer to 25% GCN but 16 and 20 aren't really close in terms of operational power.

The .23 GHz boost accounted for clawing back 15% of the difference that would have existed between the two cards.
 

Panajev2001a

GAF's Pleasant Genius
Yes there is.

No there is not else no developer would ever optimise any code whatsoever... else hyper threading as a concept would not exist to improve utilisation... etc... once you realise 100% utilisation is often a pipe dream on both CPU and GPU other power consumption strategies start to make even more sense than they were already making.
 

longdi

Banned
No there is not else no developer would ever optimise any code whatsoever... else hyper threading as a concept would not exist to improve utilisation... etc... once you realise 100% utilisation is often a pipe dream on both CPU and GPU other power consumption strategies start to make even more sense than they were already making.

efficiency goes both ways, especially if cut from the same rDNA2 cloth. 🤷‍♀️

Im was just trying to say the basics, the float all boat if you will.
If you ride on Series X wave, you can float 12 boats all the time
If you ride on PS5 wave, you can float 10 bloats ? of the time, choppy variable waters?

The same boat floats are seen in 1X and PS4P, and One and PS4
 
The tempest engine will also draw from PS5 448gbs memory bandwidth. Mark actually warned that if not carefully used, it will leech too much.
You asspull harder then legend of korra.

 
Last edited:

geordiemp

Member
What's the tefaflop difference between the 2080 S and TI?


Never mind, I looked it up

13.45 for the 2080 TI
11.15 for the 2080 Super

The difference between those 2 is quite staggering with certain games.

Final edit

I can see why, one has 496GB/s Bandwidth (Super) vs 616GB/s Bandwidth (TI)

448GB/s (PS5) vs 560GB/s (XSX) of memory bandwidth for VRAM is something that's often overlooked, it's not just a 2TF difference (At best case scenario) but a difference of 112GB/s for the portion of XSX dedicated to VRAM, that can have a significant impact.

Oh dear, we are back on not understanding anything again. WRONG.


I doubt PS5 can sustain 10TF all the time.

Of course you do, so does timdog and colbert, Luckily we ignore trolling FUD from the not so clever (sustained). Do you think posters believe such fanboy dribble over Cerny, really ?

Posters who worry and doubt about sustained have no clue how semiconductor chips work, it really is sad. and should go educate themselves
 
Last edited:

Panajev2001a

GAF's Pleasant Genius
efficiency goes both ways, especially if cut from the same rDNA2 cloth. 🤷‍♀️

Im was just trying to say the basics, the float all boat if you will.
If you ride on Series X wave, you can float 12 boats all the time
If you ride on PS5 wave, you can float 10 bloats ? of the time, choppy variable waters?

The same boat floats are seen in 1X and PS4P, and One and PS4

I do not think you got Cerny’s metaphor there mate ;).
 

longdi

Banned
You asspull harder then legend of korra.

🤷‍♀️

"GPUs process hundreds or even thousands of wavefronts; the Tempest engine supports two," explains Mark Cerny. "One wavefront is for the 3D audio and other system functionality, and one is for the game. Bandwidth-wise, the Tempest engine can use over 20GB/s, but we have to be a little careful because we don't want the audio to take a notch out of the graphics processing. If the audio processing uses too much bandwidth, that can have a deleterious effect if the graphics processing happens to want to saturate the system bandwidth at the same time."

 
🤷‍♀️



Why does your avatar look like the guy just let loose an incredibly wet fart in the elevator?

And 20gbs for audio? Nobody is going to use that. Most people use tv speakers. Unless they gpt that 10.1 atmos beauty then most companies ain't gonna focus too much on audio.
 
Top Bottom