• 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.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.
The reason i believe for the 25TF is that normal calculations are 32bit in it's 12TF mode, for RayTracing the calculations can be done in 16bit and thus 2 calculations can be done at once thus the 25TF number appears, it's called Rapid Packed Math i think and the PS Pro had it aswell. So it's not like it can operate in this mode for everything but only for select things, which in this case is RayTracing.
No, the original quote (I think it was from Digital Foundry) was that purely running RT on unified shaders would require 13TF of compute, however that has been offloaded on to dedicated hardware. So the Xbox has 12TF of traditional shader compute and additional 13TF of ray tracing compute, which added together is "25TF".

All of this is absolute bollocks, and whomever decided to make this bullshit up is preying on most console fans having little understanding of the technology. Its misleading bollocks.
 

pasterpl

Member
nah i'm still 9.2 TF in heart
#TeamGitHub

giphy.gif
 
A 25TF xbox... yeah and what's next? A 1080p Switch for next summer? A new PS4 pro that double the power of the 1st one? A 8TF PS5 at 199 with a 33TF Pro for 399?





Those months are gonna be long! 😱
 

Shio

Member
With all this talk of 25TF from the xbox side it seems more and more likely that the PS5 could be more than 12.1TF and thus they have to advertise 25TF to get across that even though their 12.1 is less than the competition that it is still more (25TF?), sounds alot like PS4 PRo and Xbox One X where people were going on about the PS4 Pros 8.4TF in Rapid Pack Math mode. I can hear the arguments now, PS5 is faster than Xbox Series X, no but Xbox is faster as its reallly 25TF!!...lols
 

Pug

Member
With all this talk of 25TF from the xbox side it seems more and more likely that the PS5 could be more than 12.1TF and thus they have to advertise 25TF to get across that even though their 12.1 is less than the competition that it is still more (25TF?), sounds alot like PS4 PRo and Xbox One X where people were going on about the PS4 Pros 8.4TF in Rapid Pack Math mode. I can hear the arguments now, PS5 is faster than Xbox Series X, no but Xbox is faster as its reallly 25TF!!...lols
MS is not pushing the 25 TF bollocks, it's bloggers who have no idea what they are talking about. MS have pushed 12TF since they mentioned it at the reveal at the game awards.
 
Last edited:

DrDamn

Member
I think MS are playing the information game quite well currently. Information is regular, significant enough each time and doing a good job of building momentum and hype. Whilst the Minecraft and Gears 5 example can be a little underwhelming it's important information about how XSX can enhance your current experiences. Load time demos also help reinforce this. They can start transitioning to games specifically enhanced and then games developed with XSX in mind to show how all the new features build on this and what you will experience on release.

Sony will release more info when they are ready, they don't need to respond to MS, just stick to plan and have a similarly well thought through build up of information.
 

PaintTinJr

Member
So tommy said ps5 gpu is stronger than xsx but xsx cpu is stronger than ps5 .

Funny enough last month Phil Spencer said resolution (gpu ) is not that important anymore and FPS (CPU) is more important :



Would they ever say this if they had gpu advantage over their competitor? I doubt knowing MS

🤔🤔🤔🤔
They could be right, and it would certainly explain why die-space has been used for 8 general purpose cores, when 2 or 3 core at 3.6GHz should be more than ample to hit 120fps in 4K gaming - even if focused on that tiny, tiny, tiny niche console gamer.

Personally, I loved the hardware reveal - top notch design and build quality, not to mention performance - but I could only see myself buying it as a PC running Windows 10 Pro for video encoding, graphics programming and web surfing... short of them having the first and second party games like Nintendo and Sony that demand I buy their consoles every generation. But never the less, it looks great and for me is the first real iconic console class piece of hardware they’ve produce at the very start of a generation that brings Xbox as a console brand inline with their competition on hardware statement.
 

DrDamn

Member
MS is not pushing the 25 TF bollocks, it's bloggers who have no idea what they are talking about. MS have pushed 12TF since they mentioned it at the reveal at the game awards.

It's being taken out of context and overblown (like the FP16 PS4 Pro comments), but it is someone from MS making the claim ...


"Without hardware acceleration, this work could have been done in the shaders, but would have consumed over 13 TFLOPs alone," says Andrew Goossen. "For the Series X, this work is offloaded onto dedicated hardware and the shader can continue to run in parallel with full performance. In other words, Series X can effectively tap the equivalent of well over 25 TFLOPs of performance while ray tracing."
 

Evilms

Banned
In my opinion the 25 TFlops are misinterpreted.

The console will be able to handle 4k at 60fps with Ray Tracing enabled using dedicated hardware exclusively for the RT.

Without this dedicated hardware the console would need a power of 25 TFlops or more to get the same performance.

BUT that doesn't mean that the console is a 25 TFlops machine, basically the console will run a game without RT at 12 Tflops in FP32 and with RT it will still be 12 Tflops with the dedicated hardware.

If Microsoft says that their console is a 12 TF machine it's not for nothing, otherwise they would have shouted from the rooftops that it's a 25 TF beast which of course doesn't make sense.
 
Last edited:

B_Boss

Member
That's the point of a concept. Get attention by creating wacky shit knowing it'd never work in the real world.

It's like concept cars.

I don't even mind odd sleek space age curves. Aesthetics are subjective.

But then you got cars which have tight wheel wells where only about two inches of tire actually touches the road, and those same cars also are so low all it takes is a speedbump or a pot hole and the whole car would smash. Looks nice on a smooth ceramic floor at an auto show. Put it on the road outside and it would be disaster driving it out of the parking lot.

Many of those concept cars aren't even operable anyway. It's a shell where you can't even check the insides or underneath. Some team got paid millions to create a funky shell, dashboard and seats. The rest is probably hollow.

There's some cars that are even permanent convertibles. So I guess you can only drive it in hot dry conditions. lol

Sure (and I definitely get your perspective 🍻) but even concepts use real world data sometimes to give an idea so to speak. The brainstorm phase is usually when things can go with less “boundaries” I’d imagine lol. Knowing the console will be pretty powerful, knowing also what we’ve seen from the vents, I’d have all the reason to create concepts based on those factors.
 
Last edited:

Gavin Stevens

Formerly 'o'dium'
Edit: I keep click this thread thinking it’s the other one, lol.

People are totally misunderstanding what the hardware is doing, so let me make this really simple for you.

You have two cups of water. One cup is your game, essentially everything you would render on a console right now. Another cup, is your ray traced stuff, this would be for anything that would require rays, be it ray traced contact shadows, Gi, reflections etc.

The general idea is that cup 1 is 12 tf and cup 2 is 13tf.

You render your game using the original 12tf, and can reach 12tf of performance this way, using everything EXCEPT DXR. This means you would have essentially 12tf of pure performance. THEN, on top of that, you have another 13tf of power devoted “solely” to DXR. This 13tf of power will NOT allow you to do more stuff the other could, so you won’t be rendering more models or textures or whatever. But it WILL allow you to render a theoretical 13tfs worth of DXR “stuff”.

So in short, the console is still 12tf, but it’s actual output is far greater if you’re using DXR, but ONLY for DXR related stuff. It’s not a 25tf console.

This is an annoying thing that should be cleared up.

The other thing I keep seeing already is that SSD game loading is soooooo slow, compared to Spidermans instant loading. Spiderman was a custom built game for that demo to show what could be done, state of decay 2 was essentially a “pop the disk in and see” video to show you the general speed increase that will happen to media UNPATCHED. It’s about as fair a comparison as two kettles, one of which is made from chocolate.
 
Last edited:

Panajev2001a

GAF's Pleasant Genius
It's being taken out of context and overblown (like the FP16 PS4 Pro comments), but it is someone from MS making the claim ...


I have no problem MS talking about it this way and to remind people about AMD’s GCN and RDNA enhancements for compute and machine learning (double rate FP16, quad rate 8 but ops, octal rate 4 bit ops, etc...). They are neither lying nor misleading with these two sets of numbers.

Just find it funny all those posters that shat on Sony and Cerny’s FP16 Rapid Packed Math comments and kept on dragging them through the mud every chance they have. They know who they are and why they are staying quiet about this (because it comes from their plastic box of choice’s manufacturer).
 

Pug

Member
I think MS are playing the information game quite well currently. Information is regular, significant enough each time and doing a good job of building momentum and hype. Whilst the Minecraft and Gears 5 example can be a little underwhelming it's important information about how XSX can enhance your current experiences. Load time demos also help reinforce this. They can start transitioning to games specifically enhanced and then games developed with XSX in mind to show how all the new features build on this and what you will experience on release.

Sony will release more info when they are ready, they don't need to respond to MS, just stick to plan and have a similarly well thought through build up of information.

Having worked in marketing for longer than I care to mention, I'd love to see Xbox marketing mapping plans from start to finish for series X. For sure there will be usual buzz words that need to be mentioned etc (walls covered in post it notes!), but I'm more interested on what the goals were/are from a content management angle. Having a guess, now they have the bulk of HW story out there although I suspect they will be more with regards to cloud gaming etc to come. But after that as we head towards launch (if that happens!) they can focus on software, the one area where many gamers agree MS failed to convince this generation until way too late. If this is the case it does give MS the ability to really drive the software as we head towards launch and that holds true if the PS5 is more powerful or not. Then again they maybe making it up as they go along!
 

DrDamn

Member
The general idea is that cup 1 is 12 tf and cup 2 is 13tf.

I think the other point worth making, unless I am misunderstanding, is that the RT cup does something which would take 13TF of cup 1 to achieve. But it is a task cup 1 is not specifically designed or optimised to do, so cup 2 is not doing 13TF of work as such, just the work which non-dedicated hardware would need 13TF to do.
 

Shmunter

Member
I read that. It doesn't specify the resolution or the graphics features enabled to run 100 FPS. It reads like 2 different paragraphs (and I'm sure it is). The other question is why wouldn't the optimizations to the engine not be available for the Nvidia/AMD GPUs? UE4.0 has a LOT of frame-pacing issues. Some of the UE games with very little features are still bottlenecked. We don't see this with every graphics engine.
It’s like the console landed on earth from the depths of space, and it’s your first contact with a closed purpose built system.
 

geordiemp

Member
Edit: I keep click this thread thinking it’s the other one, lol.

People are totally misunderstanding what the hardware is doing, so let me make this really simple for you.

You have two cups of water. One cup is your game, essentially everything you would render on a console right now. Another cup, is your ray traced stuff, this would be for anything that would require rays, be it ray traced contact shadows, Gi, reflections etc.

The general idea is that cup 1 is 12 tf and cup 2 is 13tf.

You render your game using the original 12tf, and can reach 12tf of performance this way, using everything EXCEPT DXR. This means you would have essentially 12tf of pure performance. THEN, on top of that, you have another 13tf of power devoted “solely” to DXR. This 13tf of power will NOT allow you to do more stuff the other could, so you won’t be rendering more models or textures or whatever. But it WILL allow you to render a theoretical 13tfs worth of DXR “stuff”.

So in short, the console is still 12tf, but it’s actual output is far greater if you’re using DXR, but ONLY for DXR related stuff. It’s not a 25tf console.

This is an annoying thing that should be cleared up.

The other thing I keep seeing already is that SSD game loading is soooooo slow, compared to Spidermans instant loading. Spiderman was a custom built game for that demo to show what could be done, state of decay 2 was essentially a “pop the disk in and see” video to show you the general speed increase that will happen to media UNPATCHED. It’s about as fair a comparison as two kettles, one of which is made from chocolate.

Did MS say they did no work on it ? Are we sure ?

Personally dont yhink it makes much difference anyway, a few seconds here nor there, I am not that impatient, but I do believe a large corporation like MS will have prepared and rehearsed every finite detail and nothing was left to chance. They would not just pop something in and see what happens for sure, but playing devils advocate they may not ahve optimised the code.
 
Last edited:

TLZ

Banned
Edit: I keep click this thread thinking it’s the other one, lol.

People are totally misunderstanding what the hardware is doing, so let me make this really simple for you.

You have two cups of water. One cup is your game, essentially everything you would render on a console right now. Another cup, is your ray traced stuff, this would be for anything that would require rays, be it ray traced contact shadows, Gi, reflections etc.

The general idea is that cup 1 is 12 tf and cup 2 is 13tf.

You render your game using the original 12tf, and can reach 12tf of performance this way, using everything EXCEPT DXR. This means you would have essentially 12tf of pure performance. THEN, on top of that, you have another 13tf of power devoted “solely” to DXR. This 13tf of power will NOT allow you to do more stuff the other could, so you won’t be rendering more models or textures or whatever. But it WILL allow you to render a theoretical 13tfs worth of DXR “stuff”.

So in short, the console is still 12tf, but it’s actual output is far greater if you’re using DXR, but ONLY for DXR related stuff. It’s not a 25tf console.

This is an annoying thing that should be cleared up.

The other thing I keep seeing already is that SSD game loading is soooooo slow, compared to Spidermans instant loading. Spiderman was a custom built game for that demo to show what could be done, state of decay 2 was essentially a “pop the disk in and see” video to show you the general speed increase that will happen to media UNPATCHED. It’s about as fair a comparison as two kettles, one of which is made from chocolate.
Yes that's what I thought. Thanks for confirming. So basically using RT doesn't tax the 12tf gpu. That is great to know.
 

Pug

Member
It's being taken out of context and overblown (like the FP16 PS4 Pro comments), but it is someone from MS making the claim ...

I'm sure someone at MS did make the comparison. What I was trying to say is that 25TF will not be a tick box marketing phrase you will see MS use, 12TF for sure. Its been taken out of context by reporters as you said. Now if you say to me MS marketing guys would know this would be the case, good on their marketing teams!
 

Gavin Stevens

Formerly 'o'dium'
geordiemp geordiemp its in the DF video even, among other places. No changes were made to the game, unlike the gears 5 demo.

TLZ TLZ exactly. It’s important to note that if you were to turn off the DXR enhancements, you wouldn’t get performance back. If the DXR shit is taxing the system, that’s down to the implementation, and not the base GPU grunt. It’s a totally dedicated hardware solution so that the two can work together in conjunction, but not impact one another. Having worked with RTX and ray tracing for a while now, that’s the absolute best way to do it, because I can’t stress this shit enough... RT is expensive from a performance level, and you absolutely do NOT want your base performance impacted by this.

And with that... I’ll float away again, and stop mistaking this thread for the other 🤣
 
Last edited:

Sejanus

Member
I believe that both ms and Sony will use the same Gpu (custom rx5800).
Which would probably be the specs:
Rx5800 xrt
Rdna2.0 ~ 320mm / 275w
56cu (2ghz-max) ~ 14.3tf (rtx2080ti)
12gb 384bit / 672gb / s
Price: $ 499

Rx5800 pro rt
Rdna2.0 ~ 320mm / 225w
52cu (1,9ghz-max) ~ 12.6tf (rtx2080s)
10gb 320bit / 560gb / s
Price: $ 399

(new q4-2020 prices for rx5700xt: 299 $ -rx5700pro: $ 199)

The one that came closest to me is the coretechs for me

5: 36min
From 07/19/19
 

DrDamn

Member
I'm sure someone at MS did make the comparison. What I was trying to say is that 25TF will not be a tick box marketing phrase you will see MS use, 12TF for sure. Its been taken out of context by reporters as you said. Now if you say to me MS marketing guys would know this would be the case, good on their marketing teams!

I expect it was just someone at MS excited and proud of stuff they'd worked on. Quite sad that it gets used in a way it was never meant. It's an impressive technical achievement by both MS and AMD.
 
Really impressed with the XsX. I hope what Sony brings has a very similar performance.

I think the rumored 2GHz clock is very believable now for PS5 as I didn't expect a 1.8 Ghz clock for XsX.
I am also very impressed with the Series X as well.
I would much rather have the Xbox then to put to much time energy and money into a new PC in a few years from now. I will definitely by a new GPU once AMD, Nvidia show their new hotness and pricing.

I am really excited for the future of gaming. The Xbox is going to be very convenient with high end graphics an a small form factor case. I just want all that power with that convenience. I have a serious spinal cord injury that requires emergency surgery next week and I would like to remain optimistic on my recovery.

I want to be able to lift things again like a PC. Especially performing my cleaning maintenance, and to tinker with minor components.

The X could hopefully fill all those gaming needs and I won't have to fiddle with a big heavy ATX tower. Can't wait to get all that power in a small box. Wowa its sleepy eye time
 

ZywyPL

Banned
MS is not pushing the 25 TF bollocks, it's bloggers who have no idea what they are talking about. MS have pushed 12TF since they mentioned it at the reveal at the game awards.

Indeed, clueless kids making drama out of something they clearly don't understand, or are too ignorant to even want to understand. MS/DF clearly said that if they wanted to achieve what RT engine is doing on the standard CUs/rasterization they would need additional 13TF. What it means is basically this:



Despite how powerful 1080Ti is, it's falling on it's knees once it has to run such precise shadows, reflections etc.
 

PaintTinJr

Member
Edit: I keep click this thread thinking it’s the other one, lol.

People are totally misunderstanding what the hardware is doing, so let me make this really simple for you.

You have two cups of water. One cup is your game, essentially everything you would render on a console right now. Another cup, is your ray traced stuff, this would be for anything that would require rays, be it ray traced contact shadows, Gi, reflections etc.

The general idea is that cup 1 is 12 tf and cup 2 is 13tf.

You render your game using the original 12tf, and can reach 12tf of performance this way, using everything EXCEPT DXR. This means you would have essentially 12tf of pure performance. THEN, on top of that, you have another 13tf of power devoted “solely” to DXR. This 13tf of power will NOT allow you to do more stuff the other could, so you won’t be rendering more models or textures or whatever. But it WILL allow you to render a theoretical 13tfs worth of DXR “stuff”.

I would have had zero issue with them using the 25TF statement had this been a recognised convention used by Nvidia when they started with RTX cards, but the reason it isn’t is because dedicated fast paths, be it T&L or GLSL pipelines don’t get compared to the previous gen in performance because it is a strawman argument.

I also suspect the BVH algorithms could equally be accelerated on just one or two CELL processors because of their compute versatility, or on four Xeon chips...which when combined by FLOPs aren’t even 2TF, but would be thrashing memory bandwidth instead. The 13TF of GPU compute claim is likely because CUs are the wrong tool for the job, hence the fast path hardware needed.

If Sony have stuck twice the amount of RT cores in their APU, by their own logic will Microsoft acknowledge that the PS5 with RT cores enabled has a compute advantage in the 13TF ballpark? I doubt they will.
 

xool

Member
47bckV5.jpg

Some explaining needed for this pic (I know it's been enhanced etc)

Firstly the die size does seem to be 490mm2 [typo 390mm2] (GDDR6 is 14x12mm) - but there are black bars on either side . effective calculated size is 460mm2 [typo 360mm2]- checks out.

Now look at the Zen2 cores - my calculated size = 28mm2 each (wtf?) .. even if that is just the cores and not cache that's small Zen2 CCX size is supposed to be 74-80mm2 .. and even if those blank area between and below are cache they seem a little small - l2+l3 should be about the same size as the core . these look about 2/3 the size.

  1. Cores are smaller becuase 7nm+
  2. Probably cut down cache too.
 
Last edited:

psorcerer

Banned
Why does it have to be one of those?.

Wouldn't a 256bit bus with 16gbps GDDR6 be 512GB/s?

According to rdna2 arch on PC you have a 64bit controller per block of 10 CU.
If it's 256bit, then you have 4 blocks. Which is 40 CU. To reach 56 CU you need to have 14 CU per block. That's 40% increase. I would expect them to be bandwidth starving.
 

pawel86ck

Banned
With all this talk of 25TF from the xbox side it seems more and more likely that the PS5 could be more than 12.1TF and thus they have to advertise 25TF to get across that even though their 12.1 is less than the competition that it is still more (25TF?), sounds alot like PS4 PRo and Xbox One X where people were going on about the PS4 Pros 8.4TF in Rapid Pack Math mode. I can hear the arguments now, PS5 is faster than Xbox Series X, no but Xbox is faster as its reallly 25TF!!...lols
To be fair 16 bit 8.4TF on PS4P wasnt really usable in real games, while RT will be probably used in many games, so all power dedicated to RT will be used in XSX.
 

John254

Banned
Sony are more powerful, that's why they've been quiet, they didn't want to give Microsoft time to respond.

13.3tf's RDNA2, more CU's at higher clocks and more RT Cores, 24gb of RAM (20 useable for games), CPU at 3.6hz 8c16t zen 2.
Please stop beeing delusional. Even if PS5 was more powerful, revealing that changes nothing. You can't just produce new, more powerful chip 6 months before console release.
And stuff that you can change, like clock speed can be change in summer and at that time Sony will reveal their console.
 
With all this talk of 25TF from the xbox side it seems more and more likely that the PS5 could be more than 12.1TF and thus they have to advertise 25TF to get across that even though their 12.1 is less than the competition that it is still more (25TF?), sounds alot like PS4 PRo and Xbox One X where people were going on about the PS4 Pros 8.4TF in Rapid Pack Math mode. I can hear the arguments now, PS5 is faster than Xbox Series X, no but Xbox is faster as its reallly 25TF!!...lols
Yes. Tables likely turned. Sony was labelling PS4 Pro 8.4TF.
 
S
Please stop beeing delusional. Even if PS5 was more powerful, revealing that changes nothing. You can't just produce new, more powerful chip 6 months before console release.
And stuff that you can change, like clock speed can be change in summer and at that time Sony will reveal their console.
Sony were aiminh for 14.2TF from the beggining. I think they got close.
MS was targetting 12TF also and didnt account for rdna2 efficiencies.
 

DrDamn

Member
To be fair 16 bit 8.4TF on PS4P wasnt really usable in real games, while RT will be probably used in many games, so all power dedicated to RT will be used in XSX.

Functionality was usable, but in a targeted rather than general way. You couldn't apply it to everything but you could make good use of it to make some performance gains.
 
Status
Not open for further replies.
Top Bottom