Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
Yeah and that's just the case for multiplatform, we still don't know how big of a graphics hindrance Lockhart would bring to Anaconda for their first party titles, that's probably an area where PS5 could overtake the lead when nothing is holding it back.

Also Osiris strikes again!

Floodgate opening any day now.
Shit reads like the MisterxMedia of second comings.
 
Basically we have two rumors: 300mm2 and 2GHz for PS5 APU.

If the chip is in 7nm+ the APU should have 56CUs@2GHz then it would have 14.336TFflops.
even it is 52CUs active it also has 13.312TFflops. Both numbers are super high.

However if APU is only in 7nm then 44CUs@2GHz it only has 11.264TFflops. This is probably
higher than stadia but the yield is a big question since all the CUs must be active.
 
I have my reservation regarding this for two reasons :

For PS5 and Anaconda, I expect it to be 4K, or in certain cases middle between 2K and 4K with good up-scaling. Everyone on this forum is aware that by far biggest performance hit comes with increase of resolution, therefore a console which is targeting 4K, and provides 10TF GPU, will reserve alot of rendering budget for increase in resolution, vs one that targets 1080p/1440p and has 4-5TF.

So from that stand point, I don't see what console with 10TF Navi, 8 core Zen2 and 16GB of usable RAM can do, that console with 4-5TF Navi, same 8 core Zen2 and 12GB of usable RAM can't at 1/3rd of resolution.

.

Here's an example!

Microsoft has Lockhart at 1440, and Anaconda at 4K.

With lockhart beeing a 4Tflops GPU is just and improved PS4 Pro that aims for the same resolution (1440p).

And if the lower console is basicaly a PS4 Pro with better CPU and better bandwidth and memory, the fact remains that if the bottleneck is GPU power, lockhart games would be not better that current gen ones. And not even RDNA would make a real diference here.
And if the lower console is forcing games to be this way (mostly current gen), upper console will suffer.

With Lockhart Microsoft is locked. They cannot use the Anaconda power freely. They have to use it on resolution, otherwise there is no scalling back for Lockhart.

But not Sony! So a dev, if it wishes can create 4K reconstructed game, going as low as 1080p, using the ful power of the GPU. But unlike Microsoft that power would not be used for extra pixels processing, but for realistic effects and graphics!

This game would absolutely visually destroy all Anaconda could do. And if Microsoft wished to do something like that, it would have to sacrifice lockhart. If an Anaconda game dropped to 1080p, at what resolution would you work on Lockhart? 720p? In 2026??

But why would Sony do that? First because 1080p is still an awesome resolution, and although lower than 4K it is still awesome, and good enough for reconstruction methods, unlike resolution bellow it where detail starts to lack. Now, imagine that they launch a PS5 Pro that could run that same game at 4K! Sony could drop resolutions on the launch PS5, creating it´s own version of Lockhart, and smashing Microsoft.

Besides, here is a question! As you know, some great games are on Switch! The Witcher, Doom, etc! And Switch has not the same memory, or GPU or CPU power as PS4 or Xbox One. Yet the power available sufices to run the games!
What I mean is, not all next gen games will use the CPU, memory, and bandwidth resources to the full! And this means, that some of those games could very well run on the 6 Tflops Xbox One X!

Now here's the question! Will they run?

- If not, then why? Xbox One X having enough CPU, memory, bandwidth and above all GPU, could run it!
- But if they do run... With 6 Tflops, gfx could very well use a superior resolution than the one used by Lockhart. On an old gen console!

Will the use of RTX make a diference? Not really! Even with better lightning, Xbox One X could very well have the superior version!

There are a lot of questions here, and due to them I think Lockhart is a terrible idea!
 
Basically we have two rumors: 300mm2 and 2GHz for PS5 APU.

If the chip is in 7nm+ the APU should have 56CUs@2GHz then it would have 14.336TFflops.
even it is 52CUs active it also has 13.312TFflops. Both numbers are super high.

However if APU is only in 7nm then 44CUs@2GHz it only has 11.264TFflops. This is probably
higher than stadia but the yield is a big question since all the CUs must be active.
We have one rumour from May vis Reddit from someone (from Asia)w specified PCB "leak" for PS5 dev kit.

I remember everyone saying its BS because no way are devkits with actual APUs sent that early, but by July/August we got actual PS5 dev kit case and there is very little chance these sported PC parts, and not actual APU.

That leak gave parts of PCB as well as dimesions of a chip, so might be leaked by someone in manufacturing. Chip size was 316mm² which would fit with PS4 > PS4 Pro reduction (348mm² > 325mm²).

It also gave us snippets of RAM arrangement which pointed at 18Gbps chips on 256 bit bus from Samsung (which was kida confirmed by Flute leak - which was 100% legit).

I think people are failling to see what is happening here. Sony and MS provided target specs and back when they did (2018 by Sony), only target you could give to devs was one in GCN flops.

This is why 12-13TF on 256bit bus make 0 sense. This is why anything beyond 10TF Navi with RT makes 0 sense. Its just HEAVILY TDP constrained.

I am sure we have PS5. Flute, Gonzalo, Prospero and Ariel are all PS5 related. TS score of 20K points to 9TF system, no way its 13TF Navi. Flute BW also points at system with less then 10TF because 530GB/s is not enough for 13TF console.
Lets wait and see, but I am 100% we have PS5 and target specs released in 2018 where on point (if we look at actual performance of console).

8 core Zen 2 with 1/4th of cache
~9TF Navi with RT and VRS
16GB GDDR6 - 256bit bus - 512-576GB/s
Custom SSD

For MS its harder to predict since we have no codenames and leaked benchmarks, but that one already has 320bit bus so they may have left some room for more, although I am sure it cant be 12TF.
 
Not 4GB lower (Not for many games at least.). From 13GB to 9GB is a lot.

Well, it won't run that at 1080p. Not with 4Tflops. 720p RT games on Lockhart incoming.

They have claimed the HDD is mapped as memory so the game may go from

50+13 to 50+9
This seems a lot less.

Is the memory difference because it needs the space or because they want the bandwidth?

With a fast HDD do they need as much loaded into memory all the time?

The new hardware seems so different it's hard to understand the tangible difference to experiences that are possible as that is what the reduction in baseline would affect.
 
I'm still disappointed that we have very little idea of how powerful they'll be at ray tracing. It makes TF rumours even more meaningless (but I love the speculation regardless)
 
https://gizmodo.com/a-story-about-two-tiny-fuzzy-photos-of-a-playstation-5-1837594238

The illustrations are rumored to be mockups of a pre-production PS5, and they circulated widely after their discovery in August. Our tipster sent us the photo in June. (In the interest of protecting our tipster, we won’t post the photos here.)

The timing is crucial here because, conceivably, some jackass could have found the design illustrations of the PS5, and by some sorcery (3D printing, Photoshop) created a model to trick us. Our photo predates the circulation of the registration images.

On May 25, 2019, the law firm Kasznar Leonardos registered an industrial design in Brazil (login required) in the interest of its client, Sony Interactive. Published on August 13, that registration referenced a Japanese industrial design registration filed back on November 30, 2018. The Japanese registration is sealed with a note on the documentation claiming it is a secret design and images cannot be displayed.


From May 20th. Alleged first dev kit PCB info.

https://lordsofgaming.net/2019/07/p...hmark-leaked-ps5-based-on-amd-flute-platform/

Flute benchmark from July. Removed the next day.

AMD-flute-gonzalo.jpg
 
instead of releasing a pro model, what's preventing them to add the option to link two PS5s with some special fast port to get double power ?
They showed old demos at the time of gran turismo running on several PS3s.
 
instead of releasing a pro model, what's preventing them to add the option to link two PS5s with some special fast port to get double power ?

It would probably achieve double power consumption, but it wouldn't reach twice the performance.
 
This is why 12-13TF on 256bit bus make 0 sense. This is why anything beyond 10TF Navi with RT makes 0 sense. Its just HEAVILY TDP constrained..

One thing no one knows about Navi is why it's so power hungry. When looking at performance per watt, I would say it's worse than NVIDIA who is a node behind. To me that says that AMD has a lot of potential to still improve power efficiency.

Are there micro-architectural design issues with Navi that simply couldn't be addressed since they had to meet schedule? This happens all the time in the semiconductor industry. Schedules and time to market demands drive products and designs ship not when they're completely and perfectly done, but when they're good enough. Every chip out there has issues and bugs that the design team wasn't able to address. I'm not saying I know that this is the case with Navi, but I could see that being the case. They prioritized performance over power and for the PC space, that's absolutely the right call.

However given that the console is shipping 15 months later than the initial PC product, I can easily envision all the additional work going into the console APU to be focused on improving power efficiency: fixing bugs, optimizing timing paths, using custom physical libraries, doing custom layouts, etc. My personal expectation, is that the consoles will look a lot closer to RDNA1 (with improved area and power efficiency) than shipping RDNA2 PC GPU's.
 
One thing no one knows about Navi is why it's so power hungry. When looking at performance per watt, I would say it's worse than NVIDIA who is a node behind. To me that says that AMD has a lot of potential to still improve power efficiency.
I think I remember some statements from nvidia claiming their 12FFN has very close performance characteristics to TSMC's early 7nm, which is why they'd be using gigantic GPUs this gen while waiting for EUV processes to come up.

Regardless, it's true that there should be some headroom for optimization on AMD's side even on plain 7nm. We know that at least the reference Navi 10 cards are factory overvolted towards ridiculous levels, and we've seen core voltage tweaks on those cards increasing their efficiency by a lot.
 
Basically we have two rumors: 300mm2 and 2GHz for PS5 APU.

If the chip is in 7nm+ the APU should have 56CUs@2GHz then it would have 14.336TFflops.
even it is 52CUs active it also has 13.312TFflops. Both numbers are super high.

However if APU is only in 7nm then 44CUs@2GHz it only has 11.264TFflops. This is probably
higher than stadia but the yield is a big question since all the CUs must be active.
Probably higher than stadia ? 7.5tf navi is higher performance than stadia gpu.
 
The Cell had a theoretical maximum of 25.6 GFLOPs * (1 PPE + 8 SPE) = 230.4 GFLOPs FP32.
The $1000 Core i9 10980XE with 18 cores at sustained 3.8GHz theoretically does 18 cores * 4 instructions-per-cycle * 2 FMAD * 3.8 GHz = 547.2 GFLOPs,
He was comparing the 9900K to Cell, the 9900K is the fastest CPU for gaming right now (faster than even the 10980XE), so in that respect the 9900K is only capable of 160 GFLOPS, which is significantly less than Cell.
 
OK let's break it down a bit using insiders Klee, Matt and Jason as basis. Klee said both XB2 and PS5 are above Stadia (10.7 TF) in raw teraflops at the time before this new leak that contains updated hardware. So the older XB2 and PS5 devkits would be at least north of 11 TF and in RDNA flops with PS5 slightly faster in the realm of 10% so ~12 TF. Now with the new XB2 devkit it would seem they have either reached or getting close to 12 TF thus matching PS5 devkit more or less.
Matt later added even with the improved XB2 devkit the difference in power is still negligible so maybe they're just much closer than ever before now. Jason pretty much shot down that pastebin post of a 10.24 TF PS5. If they're all to be believed nextgen consoles are at the cusp of becoming 12 TF monsters that would eat 2080 for breakfast.

Way to conservative, PS5 will beat 2080Ti, coming close to 2x 2080Ti in raw power, that's without optimisations in mind.

This is why anything beyond 10TF Navi with RT makes 0 sense.

You might want to reconsider that, both will be coming in at over 15TF navi flops, even closing in on the 20TF level of performance.

8 core Zen 2 with 1/4th of cache

Wrong, it will be Zen 3, 8 cores clocked at atleat 3.8Ghz, boost to 4/5Ghz range.

16GB GDDR6 - 256bit bus - 512-576GB/s

Rumors point to 'very powerfull', atleast 24GB, expect 32. HBM2.

I'm still disappointed that we have very little idea of how powerful they'll be at ray tracing.

They will put RTX3080Ti to shame in both raw rasterisation performance, and RT performance aswell.

My personal expectation, is that the consoles will look a lot closer to RDNA1

They will be sporting RNDA2, with features from RNDA3 GPUs.

the 9900K is the fastest CPU for gaming right now (faster than even the 10980XE)

True but the next consoles will sport much faster CPU's then that.
 
In the history of console gaming, have rumours ever not pointed to consoles being monster machines? Without leaks, we consistently hear (or so it seems) that a new console will blow socks off, and invariable the final thing never lives up to those impossbile expectations and socks remain firmly on, at best a little loosened.
 
We have one rumour from May vis Reddit from someone (from Asia)w specified PCB "leak" for PS5 dev kit.

I remember everyone saying its BS because no way are devkits with actual APUs sent that early, but by July/August we got actual PS5 dev kit case and there is very little chance these sported PC parts, and not actual APU.

That leak gave parts of PCB as well as dimesions of a chip, so might be leaked by someone in manufacturing. Chip size was 316mm² which would fit with PS4 > PS4 Pro reduction (348mm² > 325mm²).

It also gave us snippets of RAM arrangement which pointed at 18Gbps chips on 256 bit bus from Samsung (which was kida confirmed by Flute leak - which was 100% legit).

I think people are failling to see what is happening here. Sony and MS provided target specs and back when they did (2018 by Sony), only target you could give to devs was one in GCN flops.

This is why 12-13TF on 256bit bus make 0 sense. This is why anything beyond 10TF Navi with RT makes 0 sense. Its just HEAVILY TDP constrained.

I am sure we have PS5. Flute, Gonzalo, Prospero and Ariel are all PS5 related. TS score of 20K points to 9TF system, no way its 13TF Navi. Flute BW also points at system with less then 10TF because 530GB/s is not enough for 13TF console.
Lets wait and see, but I am 100% we have PS5 and target specs released in 2018 where on point (if we look at actual performance of console).

8 core Zen 2 with 1/4th of cache
~9TF Navi with RT and VRS
16GB GDDR6 - 256bit bus - 512-576GB/s
Custom SSD

For MS its harder to predict since we have no codenames and leaked benchmarks, but that one already has 320bit bus so they may have left some room for more, although I am sure it cant be 12TF.

If this the real Endproduct /Retail then good Night. It is the same Bottleneckarchitecture like on the Ps4. What People forget with an UMA Architecture is , you cannot have the full 256 Bit Bandwidth for the GPU because you must share it with the CPU , and a Zen2 CPU at 3 or 3,5 Ghz , 16 x Multithreading , need more Bandwidth then the oldJaguar CPU at 1,6 Ghz. It is not only about Gbytes/s . And Gddr6 is a high Latency Memory not really good for CPU Operations. So in real you have only 192 Bit for the GPU and the Rest goes to the CPU , not enough for native 4K gaming and 60 Fps , and not enough for the Zen2 CPU to utilize her full strength. In such Bottleneckarchitecture a Zen2 CPU is far more limited as in a PC Architecture/Motherboard.

So yes 16 Gbyte Ram and a 256 Bit Bus is good for the Buisness , but bad for the Games, bad for the Progress, and bad for my Eyes. So what have they done in the 7 Years at the R&D Headquarter?
 
Status
Not open for further replies.
Back
Top