Ok what is a teraflop?

[Deleted member 35204]

teraflop
/ˈtɛrəflɒp/

noun
COMPUTING a unit of computing speed equal to one million million (1012) floating-point operations per second.

At the risk of losing my gamer cred, I can't quite figure this out. It seems to mostly relate to the processor, but then why not use the processor speed. I would understand that and that's worked every other generation.

I know what a motherboard is and how that effects thing's.

I know what ram is and I can understand a graphics card.

Basically I want to know what 1 teraflop is then I can quantity it?

Click to shrink...

In a continuous search for simplification of complex matters TFLOP/s is what has been used as unit of measure for performance in computing.
Depending on your age you may remember its predecessors such a number of bits (e.g. PS1 is 32bits and N64 is 64bit), number of polygons per second (PS2 pushes 75 millions of polygons per second while GameCube pushes 20 millions) or the frequency of the processor (this Pentium 4 is at 1.7Ghz while this G4 is at 867Mhz).
Like the predecessors it has been used to compare to different pieces of hardware to determine which one is better/more performant and powerful... the more time goes on and the more we change these colloquial units of power the more they become reliable (e.g counting FLOP/s is more accurate than counting number of polygons and the latter is more accurate than counting the number of bits) but in the end you always arrive at a point where the equation that "bigger number means better" doesn't work anymore.
Once again we have arrived in an impasse where counting this unit of performance is not as reliable as we generally thought, it has always been this way as history tell us where in recent years the number of "FLOP/s" on an Nvidia card gave more performance than an equivalent from AMD/ATI because the architectures were substantially different.


Let's arrive at the actuality of things: You probably are asking this question because you have seen the PS5 having a considerable lower amount of TFLOP/s and some people are saying/might say this difference is not as marked as it seems because other stuff is faster while other say that there is a huge difference or maybe you are asking this question simply because you want to quantify this difference;
This number of TERAFLOP/s is given from a specific part of the graphic processor called ALU (arithmetic logic unit) and equals to one billion (TERA) floating point (FL) operations (OP) per second (/s), the graphic processor though is made by multiple parts some more useful than others depending on the context such has amount of cache, number of TMU (texture mapper units) or number of ROP (Raster Operation Pipeline), all these things are fundamental in producing good looking images and you can say that no one is more important than the other.
As what a TFLOP/s can do i cannot show you because it can't be showed in a simple way, sometimes it can be useful just like it can be useless as a unit, i could show you a machine that produces a terrible picture and one that gives a beautiful one while both of them truthfully have the same "teraflop/s power".
In the given context though it can give you a general and rough idea on how the two consoles compare to each other given the fact that they pretty much share the same exact architectures and are used for the same things although always going back to the fact that it's a simplification of complex matter it cannot be used as the definitive meter of paragon to say "that one is x amount better than the other".

With time it will be surely found a new unit (or units) of measure that will be more reliable and relatable with the times.

 

[azertydu91]

Imagine a 3D game.
The world in the game is stored in the computer as a mesh of triangles.
Any triangle has 3 corners right?
Each corner is described with a coordinate. Something like X = 2.4, Y = 1.9, Z = 5.2
But your screen is 2D, right?
So your screen sees those triangles from a certain viewpoint.
To calculate what the final image on your screen will look like, the computer must angle those triangles correctly, then scale them etc. to be in the right position on the screen. I.e. are you seeing the triangle from the side, or above; how is it rotated, etc.
That calculation is done every frame for every coordinate of every triangle.
A game world in a AAA can easily consist of many many millions of triangles.
So a hell of a lot of calculations needs to be done every frame, right?

A coordinate of a corner in a triangle consists of 3 values, like I described above. X, Y, Z.
The computer-haxxor term for such a number is a float, or a Floating Point value.
A computer-haxxor word for calculation is a an operation.
So a ton of Floating Point Operations needs to be performed every frame.
How many Floating Point Operations can a computer/console do?
Luckily there's a term for that. FLOPS. Floating Point Operations per Second.
More FLOPS, more calculations per second, more triangles per frame.
Better looking games.

Click to shrink...

Alright that I can understand but what I don't understand is why GCNA and RDNA flops aren't the same, because it seems like and objective measure.
That's also why I don't understand Nvidia card with less flops that used to perform better than AMD cards.
What makes a flop less powerful on a device than on another?

Edit: Nevermind the answer above me began to answer my question even if I still don't understand everything I do realize I probably don't know enough to understand those subtle diferences.

 

[Weltall Zero]

is the computer-haxxor term for angel Floating Point People, because that's what you guys are for dropping this knowledge in genuinely accessible ways

Click to shrink...

You're making me blush, hahah. I just enjoy explaining things I know to other people, and I'm used to do it often, so it's easy for me to explain them. :)

 

[Aether]

A lo tof computing in an CPU or GPU are specialized tasks.

Depending on the hardware design hardware can bebetter in some tasks, and better on others.
Terraflops is as close as you can get to "general computing power", since floating pointoperations are everywhere, especialy in computer graphics.
At the same time it dies not say much about the true potential of an architecture.
For example, an architecture could be extremly powerfull for some tasks because it has more TF, but the other has some dedicated transistors that are engineered in a way that it computes tasks that are regularly used in games with easier with way less.


In other words: for real world performance it is essentially useless when the numbers are in a simular ballpark.
(10 and 12? not really that relevant. 6 and 12? yeah, highly unlikely that they are equaly good)

 

[Deleted member 12790]

Alright that I can understand but what I don't understand is why GCNA and RDNA flops aren't the same, because it seems like and objective measure.
That's also why I don't understand Nvidia card with less flops that used to perform better than AMD cards.
What makes a flop less powerful on a device than on another?

Click to shrink...

a flop is nothing more than a single mathematical operation, how many mathematical operations it can do on non-whole numbers per second. Processors have more operations than just rote mathematics they can do on non-whole numbers, usually combined into complex operations. These complex operations can use more or less floating point operations to accomplish a similar task depending on architecture, compiler optimization, and task at hand. Its much the same way one can rewrite a single math problem in multiple ways. Like how:

(5*2) + (5*2) + (5*2) = 30 is 5 operations, while (5*2)*3 = 30 is the exact same math, but only 3 operations. Say this was part of a larger calculation on two processors, and processor A does things the former way, and Processor B does things the latter way. And while Processor A can perform more FLOPs, the complex operations it is doing takes more FLOPS to accomplish for the same task.

Really, FLOPS are a very, very poor metric for performance. I've compared it before, but it's like comparing the "power" of two authors in terms of number of pen strokes per second they can put on paper. Author A can put 10 times more pen strokes per second than Author B, does that mean Author A can write better than author B? What does "better" even mean in this instance? In this example, we don't even account for language being written, say Author B writes in Hanji/Kanji and thus every single character takes like 8-9 strokes each, while author B writes in english in cursive, and thus each individual word only takes 1 pen stroke each. Who can write more in that instance, despite author A having more pen-strokes per second?

Things to keep in mind:

A) Floating point math isn't only done in the GPU, CPUs crunch floating point math just as often

B) Ultimately, it's the program written that is using the floating point math. There are ways, for example, to avoid floating point math all together in circumstances.

C) there is also non floating point math which can have different cycles per operations.

Other things to keep in mind: the metrics put forward are FLOPS under the most ideal of circumstance, i.e. every register filled to maximize SIMD performance. There are modes in many processors, for example, which will reserve registers not always available for different tasks. There are always trade offs. There might be instances, like to avoid a cache-miss, where intentionally using every register available isn't very wise.

 

[Aether]

Jesus, how is Sony going to fit that many in the PS5...

Click to shrink...

The GC shrinking factories are getting better and better every year.
Mostly, because with the shrunken gamecubes they can develop new gamecube shrinking technologies. Somewimes they also increase the density of gamecubes by aranging them differently.

 

[azertydu91]

In a continuous search for simplification of complex matters TFLOP/s is what has been used as unit of measure for performance in computing.
Depending on your age you may remember its predecessors such a number of bits (e.g. PS1 is 32bits and N64 is 64bit), number of polygons per second (PS2 pushes 75 millions of polygons per second while GameCube pushes 20 millions) or the frequency of the processor (this Pentium 4 is at 1.7Ghz while this G4 is at 867Mhz).
Like the predecessors it has been used to compare to different pieces of hardware to determine which one is better/more performant and powerful... the more time goes on and the more we change these colloquial units of power the more they become reliable (e.g counting FLOP/s is more accurate than counting number of polygons and the latter is more accurate than counting the number of bits) but in the end you always arrive at a point where the equation that "bigger number means better" doesn't work anymore.
Once again we have arrived in an impasse where counting this unit of performance is not as reliable as we generally thought, it has always been this way as history tell us where in recent years the number of "FLOP/s" on an Nvidia card gave more performance than an equivalent from AMD/ATI because the architectures were substantially different.


Let's arrive at the actuality of things: You probably are asking this question because you have seen the PS5 having a considerable lower amount of TFLOP/s and some people are saying/might say this difference is not as marked as it seems because other stuff is faster while other say that there is a huge difference or maybe you are asking this question simply because you want to quantify this difference;
This number of TERAFLOP/s is given from a specific part of the graphic processor called ALU (arithmetic logic unit) and equals to one billion (TERA) floating point (FL) operations (OP) per second (/s), the graphic processor though is made by multiple parts some more useful than others depending on the context such has amount of cache, number of TMU (texture mapper units) or number of ROP (Raster Operation Pipeline), all these things are fundamental in producing good looking images and you can say that no one is more important than the other.
As what a TFLOP/s can do i cannot show you because it can't be showed in a simple way, sometimes it can be useful just like it can be useless as a unit, i could show you a machine that produces a terrible picture and one that gives a beautiful one while both of them truthfully have the same "teraflop/s power".
In the given context though it can give you a general and rough idea on how the two consoles compare to each other given the fact that they pretty much share the same exact architectures and are used for the same things although always going back to the fact that it's a simplification of complex matter it cannot be used as the definitive meter of paragon to say "that one is x amount better than the other".

With time it will be surely found a new unit (or units) of measure that will be more reliable and relatable with the times.

Click to shrink...

If I understood you (which I probably didn't) wouldn't a new measure consisting of an agglomerate of FLOPs,ROPs and TMUs be a better representation of a GPU.I mean in the PS5/XseX some people claim that a higher clock might mitigate those diferences or are those units simply most suited to indicate diferents graphical aspects (like particles,draw distances,lod etc.)?

 

[azertydu91]

a flop is nothing more than a single mathematical operation, how many mathematical operations it can do on non-whole numbers per second. Processors have more operations than just rote mathematics they can do on non-whole numbers, usually combined into complex operations. These complex operations can use more or less floating point operations to accomplish a similar task depending on architecture, compiler optimization, and task at hand. Its much the same way one can rewrite a single math problem in multiple ways. Like how:

(5*2) + (5*2) + (5*2) = 30 is 5 operations, while (5*2)*3 = 30 is the exact same math, but only 3 operations. Say this was part of a larger calculation on two processors, and processor A does things the former way, and Processor B does things the latter way. And while Processor A can perform more FLOPs, the complex operations it is doing takes more FLOPS to accomplish for the same task.

Really, FLOPS are a very, very poor metric for performance. I've compared it before, but it's like comparing the "power" of two authors in terms of number of pen strokes per second they can put on paper. Author A can put 10 times more pen strokes per second than Author B, does that mean Author A can write better than author B? What does "better" even mean in this instance? In this example, we don't even account for language being written, say Author B writes in Hanji/Kanji and thus every single character takes like 8-9 strokes each, while author B writes in english in cursive, and thus each individual word only takes 1 pen stroke each. Who can write more in that instance, despite author A having more pen-strokes per second?

Things to keep in mind:

A) Floating point math isn't only done in the GPU, CPUs crunch floating point math just as often

B) Ultimately, it's the program written that is using the floating point math. There are ways, for example, to avoid floating point math all together in circumstances.

C) there is also non floating point math which can have different cycles per operations.

Click to shrink...

Thanks that is really interesting and up until now I thought Flops would be an accurate measure.Hence why I never understood the differences between AMD and Nvidia cards.

 

[Aether]

If I understood you (which I probably didn't) wouldn't a new measure consisting of an agglomerate of FLOPs,ROPs and TMUs be a better representation of a GPU.I mean in the PS5/XseX some people claim that a higher clock might mitigate those diferences or are those units simply most suited to indicate diferents graphical aspects (like particles,draw distances,lod etc.)?

Click to shrink...

Not really. These are all hoghly complex things, and change between every hardware generation. In computer hardware, you can only really compare hardware with same architectures to each other, and even that often with an *.
It just is not easy to boil hardware architecture and design where highly specialized people worked on for years to raw numbersand say "this has more x, this is better".It almost comes down to "what is the concrete usecase it was designed for".
But you cant show people a 5-10 page paper as an explanation what hardware is better for their needs (or which console is better), so the marketing team always tries to boil it down to ... something.

 

[Dekuman]

When comparing between X series and PS5 it's equivalent because both are on AMD architecture. I think the comparisons becomes much less clear to nvidia's products, which tend to require less teraflop count for equivalent or superior performance. Just compare the same nvidia and AMD card in the same performance category, nvidia cards are always less TF

 

[azertydu91]

Not really. These are all hoghly complex things, and change between every hardware generation. In computer hardware, you can only really compare hardware with same architectures to each other, and even that often with an *.
It just is not easy to boil hardware architecture and design where highly specialized people worked on for years to raw numbersand say "this has more x, this is better".It almost comes down to "what is the concrete usecase it was designed for".
But you cant show people a 5-10 page paper as an explanation what hardware is better for their needs (or which console is better), so the marketing team always tries to boil it down to ... something.

Click to shrink...

Okay that is a really interesting answer So basically unless we're looking at device with the same architectures we can't really compare numbers and even with same architecture it may still not be representative of the final performance (I don't say games because I do understand that optimisation, time and money play a big role in it too).

 

[Era of not Yakuza]

Its ammo to be used in the console war

Click to shrink...

yep
MAUs
Sales
Metacritic score
Flops
It only matters when a company can use it to claim some kind of victory

 

[Aether]

Okay that is a really interesting answer So basically unless we're looking at device with the same architectures we can't really compare numbers and even with same architecture it may still not be representative of the final performance (I don't say games because I do understand that optimisation, time and money play a big role in it too).

Click to shrink...

Essentially.
A Simple example:
A Really strong standalone CPU struggels with a task (new compression algorythm, or just a new video codec), and you have a PC that is constantly running close to 100%. The company realizes that that is a task that comes up often, and ads dedicated feature to the cpu that can encode/decode the algorythm/codec.
Your sister buys a cheap laptop a year after you bought your beefy gaming pc. With tehe same task (lets say, playing the video with that codec) her laptop cpu sits at around 5%, while youre is constantly hitting 100%.

Is her laptop Stronger than your high end cpu? no. it just has a dedicated space on the chip for hardware that is made specificaly for that workload.

At the same time: either the chip needs to get bigger, or other stuff needs to be removed (-> worse in other areas). Its always a balancing act.
For the same task there are countless ways to do something. And usually you try to design stuff in such a way, that is multy purpose. One will decide, that this is more importaint, and we will implement the solution in a way that case a is faster than case b, and the other developer decides, no, case b is more likely to be used/ocure, lets develop it so that case b is faster.

Krejloocs post was good as an comparison.

a flop is nothing more than a single mathematical operation, how many mathematical operations it can do on non-whole numbers per second. Processors have more operations than just rote mathematics they can do on non-whole numbers, usually combined into complex operations. These complex operations can use more or less floating point operations to accomplish a similar task depending on architecture, compiler optimization, and task at hand. Its much the same way one can rewrite a single math problem in multiple ways. Like how:

(5*2) + (5*2) + (5*2) = 30 is 5 operations, while (5*2)*3 = 30 is the exact same math, but only 3 operations. Say this was part of a larger calculation on two processors, and processor A does things the former way, and Processor B does things the latter way. And while Processor A can perform more FLOPs, the complex operations it is doing takes more FLOPS to accomplish for the same task.

Really, FLOPS are a very, very poor metric for performance. I've compared it before, but it's like comparing the "power" of two authors in terms of number of pen strokes per second they can put on paper. Author A can put 10 times more pen strokes per second than Author B, does that mean Author A can write better than author B? What does "better" even mean in this instance? In this example, we don't even account for language being written, say Author B writes in Hanji/Kanji and thus every single character takes like 8-9 strokes each, while author B writes in english in cursive, and thus each individual word only takes 1 pen stroke each. Who can write more in that instance, despite author A having more pen-strokes per second?

Things to keep in mind:

A) Floating point math isn't only done in the GPU, CPUs crunch floating point math just as often

B) Ultimately, it's the program written that is using the floating point math. There are ways, for example, to avoid floating point math all together in circumstances.

C) there is also non floating point math which can have different cycles per operations.

Other things to keep in mind: the metrics put forward are FLOPS under the most ideal of circumstance, i.e. every register filled to maximize SIMD performance. There are modes in many processors, for example, which will reserve registers not always available for different tasks. There are always trade offs. There might be instances, like to avoid a cache-miss, where intentionally using every register available isn't very wise.

Click to shrink...

 

[azertydu91]

Essentially.
A Simple example:
A Really strong standalone CPU struggels with a task (new compression algorythm, or just a new video codec), and you have a PC that is constantly running close to 100%. The company realizes that that is a task that comes up often, and ads dedicated feature to the cpu that can encode/decode the algorythm/codec.
Your sister buys a cheap laptop a year after you bought your beefy gaming pc. With tehe same task (lets say, playing the video with that codec) her laptop cpu sits at around 5%, while youre is constantly hitting 100%.

Is her laptop Stronger than your high end cpu? no. it just has a dedicated space on the chip for hardware that is made specificaly for that workload.

At the same time: either the chip needs to get bigger, or other stuff needs to be removed (-> worse in other areas). Its always a balancing act.
For the same task there are countless ways to do something. And usually you try to design stuff in such a way, that is multy purpose. One will decide, that this is more importaint, and we will implement the solution in a way that case a is faster than case b, and the other developer decides, no, case b is more likely to be used/ocure, lets develop it so that case b is faster.

Krejloocs post was good as an comparison.

Click to shrink...

Alright so if I do understand (which is hard for me since I don't have all the technical knowledge and english isn't my native language) you mean that feature and specifities of a device are more important than raw numbers and a simple feature may trivialize a previously complex or power consuming operation.

Which would mean that Flops can give a global idea of a device's power but fine tuning and features (like architecture or dedicated feature) are more important to the use of a product.

 

[Khrno]

Teraflop is an expression used to define the success (or lack of) of videogames.

Take for example the classic ET from Atari, this game was such a flop, that it is categorised at the highest level, therefore, it is a Teraflop.

 

[Grimminski]

Just because you don't understand how something works or it's value doesn't make it "pointless horseshit".

Click to shrink...

Quite the fucking assumption to make there.

And that "hoax" shit? Kiss my ass with your Trump tier bullshit.

 

[Aether]

Alright so if I do understand (which is hard for me since I don't have all the technical knowledge and english isn't my native language) you mean that feature and specifities of a device are more important than raw numbers and a simple feature may trivialize a previously complex or power consuming operation.

Which would mean that Flops can give a global idea of a device's power but fine tuning and features (like architecture or dedicated feature) are more important to the use of a product.

Click to shrink...

Essentially.
If you look at the Touring Family of NVidia Graphics Cards:

Turing (microarchitecture) - Wikipedia

en.wikipedia.org

you can see, besides the usual cuda cores that are used for traditional rasterized shaders, and floating point operations,
they have aditional RT cores specialized on ray tracing, and Tensor cores for AI/machine learning.
Booth of those tasks (Ray Tracing and Machine Learning) could be done with the usual cuda cores (Compute Units) (and as far as i remember nvidia added the option to older cards to do RayTracing? I think there where performance comparisons when they released it), but it is way slower since it is kinda brute forced.
If the new cards would have said: no Machine learning or Ray Tracing, and use the space for the traditional compute units they used, then performance in games made with the prior cards in mind would have been even higher, since it is not using the new cores to their full potential.

 

[Deleted member 35204]

If I understood you (which I probably didn't) wouldn't a new measure consisting of an agglomerate of FLOPs,ROPs and TMUs be a better representation of a GPU.I mean in the PS5/XseX some people claim that a higher clock might mitigate those diferences or are those units simply most suited to indicate diferents graphical aspects (like particles,draw distances,lod etc.)?

Click to shrink...

You understood perfectly despite my bad writing, an hypothetical new measure that counts all those things together would indeed give a better representation but it's not feasible because just as you said all those parameters indicate different traits of a 3d scene and are all pretty independent from each other.
In the case of PS5 and XSeX they share the architecture so it's kinda ok making a comparison to have a general idea of performance difference but there might be for examplesome games or even single scenes in a game that take more benefits from a faster fewer TMU than more slower ones while in others the complete opposite is true.

 

[Raiden]

Nobody knows, but you have to have the most to win.

 

[TheGhost]

It's something everyone wanted high numbers of until Cerny told them it's a dangerous line of thinking.

But I'm curious if he would still say that if they had the same amount as Microsoft

 

[azertydu91]

Essentially.
If you look at the Touring Family of NVidia Graphics Cards:

Turing (microarchitecture) - Wikipedia

en.wikipedia.org

you can see, besides the usual cuda cores that are used for traditional rasterized shaders, and floating point operations,
they have aditional RT cores specialized on ray tracing, and Tensor cores for AI/machine learning.
Booth of those tasks (Ray Tracing and Machine Learning) could be done with the usual cuda cores (Compute Units) (and as far as i remember nvidia added the option to older cards to do RayTracing? I think there where performance comparisons when they released it), but it is way slower since it is kinda brute forced.
If the new cards would have said: no Machine learning or Ray Tracing, and use the space for the traditional compute units they used, then performance in games made with the prior cards in mind would have been even higher, since it is not using the new cores to their full potential.

Click to shrink...

Okay RT and Tensor cores are perfect examples thank you very much for the time you took to explain to me especially during these times when a lot of people are really tensed.It feels good to dive a bit in the more technical aspect of games and on what they are played.
Far from console warring and everything sometimes we"re too quick to forget we are gamers the kind of people that saves after a autosave and that must not forget that any progress not saved will be lost.

 

[Conmex]

Its "meaningless" for now. Once everyone realizes its not meaningless we will be about 1 - 2 years into the new gen and it wont hurt as much anymore.

 

[azertydu91]

You understood perfectly despite my bad writing, an hypothetical new measure that counts all those things together would indeed give a better representation but it's not feasible because just as you said all those parameters indicate different traits of a 3d scene and are all pretty independent from each other.
In the case of PS5 and XSeX they share the architecture so it's kinda ok making a comparison to have a general idea of performance difference but there might be for examplesome games or even single scenes in a game that take more benefits from a faster fewer TMU than more slower ones while in others the complete opposite is true.

Click to shrink...

And thank you too sorry for not quoting you and maybe my lack of a proper teaching in english suited perfectly how you talked but you made it crystal clear for me.

 

[Aether]

Okay RT and Tensor cores are perfect examples thank you very much for the time you took to explain to me especially during these times when a lot of people are really tensed.It feels good to dive a bit in the more technical aspect of games and on what they are played.
Far from console warring and everything sometimes we"re too quick to forget we are gamers the kind of people that saves after a autosave and that must not forget that any progress not saved will be lost.

Click to shrink...

Its a hard topic to wrap your head around if you dont have a technical background, and people like to jump to conclusions here, so im happy that you were interested enough to learn something =)

A bit off topic, but a video where the SNES and GBA are compared, and how simular effects are realized through different methods:

How Graphics worked on the Nintendo Game Boy Advance | MVG

In 2001 Nintendo released the Game Boy Advance which had some familiar but also advanced graphical features that helped define the system into such a fantast...

www.youtube.com

since this kinda remined me of it.

 

[Deleted member 35204]

And thank you too sorry for not quoting you and maybe my lack of a proper teaching in english suited perfectly how you talked but you made it crystal clear for me.

Click to shrink...

Dude don't worry, English is not my first language either and in general i am also a terrible writer so i'm glad someone understood what i said ahahahahah.

 

[headspawn]

It's a word that apparently had an expiration date of 3/18/2020 @ 9:15am PST.

 

[Aether]

And thank you too sorry for not quoting you and maybe my lack of a proper teaching in english suited perfectly how you talked but you made it crystal clear for me.

Click to shrink...

Dude don't worry, English is not my first language either and in general i am also a terrible writer so i'm glad someone understood what i said ahahahahah.

Click to shrink...

There was a time where i would copy every post i wrote into word and rewrite it, so that it is clear and proper english.
Then i realized, im just posting on a forum. a response should not take minutes to write. It is good to try to better your own language skills, but don't put yourself down because of it, as long as its in an informal setting and understandable (and you both are perfectly understandable) it is okay. =)

 

[azertydu91]

There was a time where i would copy every post i wrote into word and rewrite it, so that it is clear and proper english.
Then i realized, im just posting on a forum. a response should not take minutes to write. It is good to try to better your own language skills, but don't put yourself down because of it, as long as its in an informal setting and understandable (and you both are perfectly understandable) it is okay. =)

Click to shrink...

Thank you I probably learned more in this thread with you and LeleSosho than on any other thread I previously visited.And you are right we are enthusiasts and forum members and it's good to exchange ideas and knowledge no matter our mistakes and background.

Except if you're one of those people that takes their shoes off in public transport ... then we are de facto enemies.

 

[Mercador]

The definition you put is correct.

However, here's an analogy that might help you. Even if you have the best F1 motor, if not everything is in sync (synced? synched?) in your formula 1 car, it won't help. So you can put 1000hp on your car, if the car or the roads around it don't allow it, you won't help your maximum speed.

 

[Prolepro]

I could tell you

but then Id have to [REDACTED]

 

[Aether]

Thank you I probably learned more in this thread with you and LeleSosho than on any other thread I previously visited.And you are right we are enthusiasts and forum members and it's good to exchange ideas and knowledge no matter our mistakes and background.

Except if you're one of those people that takes their shoes off in public transport ... then we are de facto enemies.

Click to shrink...

Add to to tat using a bluetooth speaker in public transport to go on evrybodies nerves... so yeah, on the same page reagds public transport

 

[azertydu91]

Add to to tat using a bluetooth speaker in public transport to go on evrybodies nerves... so yeah, on the same page reagds public transport

Click to shrink...

Great ...*stops squinting eyes and slowly put the knife back in the kitchen without breaking eye contact nor blinking*...

 

[Malkier]

It's a unit of power measured at times of war to declare a victor.

 

[Megatron]

Has anyone said 'a miserable pile of secrets.' Yet?

 

[gozu]

FLOPs are just a better indication of power than Ghz in the context of 3D graphics, because that's what you want the processor to do; you don't care how many cycles it takes.

Click to shrink...

Imagine a 3D game.
More FLOPS, more calculations per second, more triangles per frame.
Better looking games.

Click to shrink...

Good stuff. 👍🏼

teraflops don't matter if all you do is 16.16 fixed-point math

Click to shrink...

wb krej.

could it not results in greater LOD onthe ps5 compared to xsx in open world games. Less pop lip, higher textures etc?

Click to shrink...

Unlikely. I was hoping for the same.

 

[SolVanderlyn]

Blast processing for a new generation.

 

[henhowc]

A dinosaur

 

[Jashobeam]

All you need to know is the XBOX has more so by default it wins the console warz

 

[Username1198]

Let's see what urban dictionary has to say:


teraflop

This occurs when a male dinosaur takes off his pants

"As the dinosaur took off his pants, his teraflop caused a great commotion."

 

[Shoichi]

It's information that are used by platform enthusiasts to see who has the better purchase. Seems we are seeing the same with SSD speeds now.

Every generation they need something. We have new terms for this one especially after MS and Sony pushed the term

 

[zoodoo]

1TF = 1000 blast processor

 

[hateradio]

teraflop
/ˈtɛrəflɒp/

noun
PORTMANTEAU of terrible and flop. Ex: "Hey, did you play <popular game>? It was a real teraflop!"

Click to shrink...

I like this.

 

[More Butter]

Who cares? TFs don't matter anymore.

 

[Kerotan]

A Terraflop is something Sony were very quick to flaunt last gen when they had more of them but next generation they are meaningless.

Click to shrink...

Don't forget the second half of last gen the Xbox X was the most powerful console and Sony had no problem continuing to dominate sales. The first xbox was stronger then the ps2 and 3rd party games were better on the 360. Power don't decide shit.

 

[j7vikes]

A trillion math operation, simple.

Click to shrink...

Simple eh? *looks up trillion, math, and operation...*

Oh yeah I totally got it all figured out 😀

 

[Fredrik]

For the saying that TF do not matter. In this context (PS5 vs XSX) is not meaningless, because both are using the same architecture.

It matters in this context, and the XSX has the advantage.

Click to shrink...

Yeah it most definitely matters, especially for multiplat developers since they usually won't build a game from the ground up to take advantage of unique system features, higher raw system power will absolutely matter and makes it easier to optimize the game to good performance.

 

[mxbison]

Everything you ever imagined, and more

 

[Knight613]

Nobody knows, but you have to have the most to win.

Click to shrink...

The Switch only needs 1 to win though :P

 

[Suicide King]

The iPhone XS can reach up to 5 teraflops, which is more than double that of the base PS4. That means it's a better gaming device, and can actually use the same controller.

 

[Shambala]

unlimitedpower.gif