Support NeoGAF

[mrBarrelNut]

Sony fanbois are the saddest bunch of people.

RDNA 3 is coming out in 2022 (or very late 2021).

The first PS5 SOC engineering samples were ready in very late 2018, before the RDNA1 even came out.

Do you really believe that AMD had anything related to RDNA 3 available in late 2018 / early 2019 for inclusion in PS5? Years before RDNA 3 is coming out? Do you really?

 

[onQ123]

What is this 'Gfx L2 cache'?
Takes up a fifth of the gpu size!

It's the 5MB of L2 cache that we already know about

 

[Vaztu]

Sony fanbois are the saddest bunch of people.

RDNA 3 is coming out in 2022 (or very late 2021).

The first PS5 SOC engineering samples were ready in very late 2018, before the RDNA1 even came out.

Do you really believe that AMD had anything related to RDNA 3 available in late 2018 / early 2019 for inclusion in PS5? Years before RDNA 3 is coming out? Do you really?

People are taking it the wrong way.

Lets talk about something everyone knows, Cache Scrubbers. In a hypothetical scenario:

Its custom built for the PS5 not used in RDNA2.

Now imagine, if down the line AMD decide to implement it in PC GPUs. Lets say, they put it in RDNA 4 GPUs.
So, its not that PS5 has RDNA 4 features, but RDNA 4 has PS5 features.

Similarly, this AMD-Sony collaboration could have other custom PS5 features put in RDNA3 by AMD.

This is what I'm getting out of all this RDNA3 debacle.

 

[Lethal01]

Sony fanbois are the saddest bunch of people.

RDNA 3 is coming out in 2022 (or very late 2021).

The first PS5 SOC engineering samples were ready in very late 2018, before the RDNA1 even came out.

Do you really believe that AMD had anything related to RDNA 3 available in late 2018 / early 2019 for inclusion in PS5? Years before RDNA 3 is coming out? Do you really?

Do you really think that Amd is a giant company that plans theses things many years advance.

It's very reasonable to think that you could see feature from a later architecture.. Do you seriously think this would be the first time this happens? It's pretty common and a very strange thing to try to claim would be out of the ordinary?

 

[Bo_Hazem]

Well, they helped them to make RDNA anyway:

Sources: AMD Created Navi For Sony's PlayStation 5, Vega Suffered

Sources reveal not only that the PS5 uses Navi graphics, but nearly 2/3 of AMD's engineering resources were devoted to the task.

www.forbes.com

 

[mejin]

If MS doesn't create anything, it's impossible for Sony. The famous "Sony too".

Anyway, if Sony solution will work it is a subject for later discussion, but they already have the new SSD tech to show right now they are thinking beyond.

 

[geordiemp]

Infinity cache 128mb takes up 100+mm^2 on RX 6000 series.

The PS5 SOC is ~310mm^2.

Do the math.

Infinity cache is not 128 MB only, its a shared L1 cache that gives an IPC upift supported by L2, and L2 larger to offset bandwidth/

Now that you understand that, you could have a shared L1 cache, smaller L2 and get say 80 % of the IPC benefit.

Remember shared L1 / L2 matrix and large L2 bandwidth are complimetary and Malleable.

For 40 CU, sony could still keep 256 bit bus as they needed 16 MB, so a shared L1 plus even 8 MB would be sufficient (If it has).

WE dont know, you dont know. But sharing caches is an architecture, as is fine gated frequency control of WGP.

 

[geordiemp]

Scale down 128 MB Infinity Cache to dual SE with 40 CU design. It would be stupid to recycle RX 5700/5700 XT's memory bandwidth while loading it up with a heavy memory bandwidth-consuming BVH raytracing search engine. PS5 GPU should have Infinity Cache greater than 32 MB.

The technical paper on IPC benefits is actually all shared L1 cache, which gives 22 % uplift just from sharing L1, that would be 4 mm2 fo rps5 from AMD tech presentation and paper.

The large L2 assists even more in this, and benefits to boost bandwidth as well, so dual purpose, and indeed that IPC will be > 22 % for big L2 cache and for 40 CU AMD can get away with 192 lane bus.

Infinity cache is a marketing term.

So, Ps5 could technically have most of the benefits with a shared L1 cache and even 8 MB L2, as it already has a 256 bus which is needed for 16 MB GDDR6. Will it be as good as 128 MB, hell no.

 

[Coulomb_Barrier]

Looking forward to hearing more about the Geometry Engine in any case...if they disclose more details.

 

[Md Ray]

The technical paper on IPC benefits is actually all shared L1 cache, which gives 22 % uplift just from sharing L1, that would be 4 mm2 fo rps5 from AMD tech presentation and paper.

The large L2 assists even more in this, and benefits to boost bandwidth as well, so dual purpose, and indeed that IPC will be > 22 % for big L2 cache and for 40 CU AMD can get away with 192 lane bus.

Infinity cache is a marketing term.

So, Ps5 could technically have most of the benefits with a shared L1 cache and even 8 MB L2, as it already has a 256 bus which is needed for 16 MB GDDR6. Will it be as good as 128 MB, hell no.

Hi, when do you think AMD will release white paper on RDNA 2 and do architecture deep-dive?

 

[geordiemp]

Hi, when do you think AMD will release white paper on RDNA 2 and its architecture deep-dive?

No Idea on timescales, but they always do shortly after. If you wanted to do detective work, you could look at the date of RDNA1 reveal and date of RDNA1 white paper ?

I cant be arsed . Its a bit different this time, as the white paper will show what certain consoles are missing so their are partners to consider.

 

[thelastword]

Well, they helped them to make RDNA anyway:

Sources: AMD Created Navi For Sony's PlayStation 5, Vega Suffered

Sources reveal not only that the PS5 uses Navi graphics, but nearly 2/3 of AMD's engineering resources were devoted to the task.

www.forbes.com

I think people are underestimating how far this collaboration went. I mean MS had a collaboration with AMD too, but not as deep. We heard about Sony and AMD's deep collaboration way before MS and certainly there was more work done and a longer process between Sony and AMD based on the more extensive customization features on PS5..........The same happened with the two companies in the past if you look at the work they did on PRO. Now just imagine, they continued from there on PS5 and the engineers on both sides were at it for much longer than they were on PS4 PRO........


https://www.tweaktown.com/articles/8643/ps5-powered-navi-2020-amd-making-sony-input/index.html

 

[THE:MILKMAN]

Sony fanbois are the saddest bunch of people.

RDNA 3 is coming out in 2022 (or very late 2021).

The first PS5 SOC engineering samples were ready in very late 2018, before the RDNA1 even came out.

Do you really believe that AMD had anything related to RDNA 3 available in late 2018 / early 2019 for inclusion in PS5? Years before RDNA 3 is coming out? Do you really?

Vaztu has the right idea here I think but just to the point about AMD having future features ready well in advance the answer is yes.

PS4 Pro's SoC had Vega features and was final spec in a dev kit in December 2015 (6 months before Polaris dGPUs launched) as it was tested and approved by MIC Japan so I assume ES chips were sometime before that and that would be 2+ years before Vega launched.

 

[Bo_Hazem]

I think people are underestimating how far this collaboration went. I mean MS had a collaboration with AMD too, but not as deep. We heard about Sony and AMD's deep collaboration way before MS and certainly there was more work done and a longer process between Sony and AMD based on the more extensive customization features on PS5..........The same happened with the two companies in the past if you look at the work they did on PRO. Now just imagine, they continued from there on PS5 and the engineers on both sides were at it for much longer than they were on PS4 PRO........


https://www.tweaktown.com/articles/8643/ps5-powered-navi-2020-amd-making-sony-input/index.html

MS was in collaboration with them for DirectX, that's all. They are a software company.

 

[djkimothy]

What do people think of the comments made by Moore's Law is Dead recently in his QA?

 

[rnlval]

The technical paper on IPC benefits is actually all shared L1 cache, which gives 22 % uplift just from sharing L1, that would be 4 mm2 fo rps5 from AMD tech presentation and paper.

The large L2 assists even more in this, and benefits to boost bandwidth as well, so dual purpose, and indeed that IPC will be > 22 % for big L2 cache and for 40 CU AMD can get away with 192 lane bus.

Infinity cache is a marketing term.

So, Ps5 could technically have most of the benefits with a shared L1 cache and even 8 MB L2, as it already has a 256 bus which is needed for 16 MB GDDR6. Will it be as good as 128 MB, hell no.

FYI, Infinity Cache is just 3rd level cache (aka "Game Cache" on Zen 2) based on Zen 2's 3rd level cache design.

AMD Radeon RX 6800 XT Specs

AMD Navi 21, 2250 MHz, 4608 Cores, 288 TMUs, 128 ROPs, 16384 MB GDDR6, 2000 MHz, 256 bit

www.techpowerup.com

L0 Cache: 32 KB per WGP
L1 Cache: 128 KB per Array
L2 Cache: 4 MB
L3 Cache: 128 MB

RDNA has a proper cache hierarchy with delta color compression everywhere e.g.

 

[geordiemp]

FYI, Infinity Cache is just 3rd level cache (aka "Game Cache" on Zen 2) based on Zen 2's 3rd level cache design.

AMD Radeon RX 6800 XT Specs

AMD Navi 21, 2250 MHz, 4608 Cores, 288 TMUs, 128 ROPs, 16384 MB GDDR6, 2000 MHz, 256 bit

www.techpowerup.com

L0 Cache: 32 KB per WGP
L1 Cache: 128 KB per Array
L2 Cache: 4 MB
L3 Cache: 128 MB

Yes i understood, being lazy and calling big L2 is what L3 effectively is. The important point is how the L1 is arranged and if it has shared betweeen arrays.

 

[rnlval]

Vaztu has the right idea here I think but just to the point about AMD having future features ready well in advance the answer is yes.

PS4 Pro's SoC had Vega features and was final spec in a dev kit in December 2015 (6 months before Polaris dGPUs launched) as it was tested and approved by MIC Japan so I assume ES chips were sometime before that and that would be 2+ years before Vega launched.

PS4 Pro's Vega Rapid Pack Math, but missing ROPS being connected to the multi-MB L2 cache.

Xbox One X's ROPS was attached to 2MB render cache which is missing on baseline Polaris 10.

Baseline Polaris 10 has a 2MB L2 cache connected to TMU.

Vega 56/64 has 4MB L2 cache connected to TMU and ROPS.

 

[TheDreadLord]

I think people are underestimating how far this collaboration went. I mean MS had a collaboration with AMD too, but not as deep. We heard about Sony and AMD's deep collaboration way before MS and certainly there was more work done and a longer process between Sony and AMD based on the more extensive customization features on PS5..........The same happened with the two companies in the past if you look at the work they did on PRO. Now just imagine, they continued from there on PS5 and the engineers on both sides were at it for much longer than they were on PS4 PRO........


https://www.tweaktown.com/articles/8643/ps5-powered-navi-2020-amd-making-sony-input/index.html

This cooperation doesn't seem to have worked very well for the Pro given the One X existence.

 

[FranXico]

MS was in collaboration with them for DirectX, that's all. They are a software company.

DirectX support and better feature parity with NVidia is a big deal for AMD. And important for their success in the PC space.

But people are believing loud marketing too easily. Often the same people who pride themselves of being skeptical of anything Sony marketing claims.

 

[oldergamer]

A lot of arguing over a single source claiming something that cant be proven. As far as im concerned red gaming isnt credible enough to make any such rdna3 hardware claims.

 

[Explosive Zombie]

A lot of arguing over a single source claiming something that cant be proven. As far as im concerned red gaming isnt credible enough to make any such rdna3 hardware claims.

Well we should stop the discussion then, a random person with a letter avatar on NeoGAF told us what to think.

 

[THE:MILKMAN]

PS4 Pro's Vega Rapid Pack Math, but missing ROPS being connected to the multi-MB L2 cache.

Xbox One X's ROPS was attached to 2MB render cache which is missing on baseline Polaris 10.

Baseline Polaris 10 has a 2MB L2 cache connected to TMU.

Vega 56/64 has 4MB L2 cache connected to TMU and ROPS.

I don't doubt this but was just making the point that AMD GPU roadmap features can and do make it in to the consoles well before they have in AMD dGPUs e.g. Vega features in PS4 Pro before PC Polaris GPUs were out.

And from what Mark Cerny went out of his way to describe in R2PS5 it could be interpreted as the collaboration this time was even deeper with Mark/Sony putting more of their ideas in to the design that will then be incorporated in to RDNA3?

 

[Knitted Knight]

This cooperation doesn't seem to have worked very well for the Pro given the One X existence.

Does not compute? Pro, released in 2016 @ $399.99. Sold more compared to the Xbox One X. Third Party games released for the most part had no discernible difference between both - some were even optimized better for the Pro (a lower spec console). Pro stopped potential hardcore PS gamer flight to PC, and uplifted Sony exclusives as graphical benchmarks for the industry vs. the non-existent competition, including vs. PC. Xbox One X, released in 2017 @ $499.99, discontinued 2 years after, didn't sell very well. Did not turn the gen around for MS despite having some third party multiplats looking slightly better than the Pro.

The Pro did what it was designed to do from a strategic standpoint. It however showed to be a niche offering.

 

[rnlval]

This is something fishy but nice question.
Renoir apu only has 12mb cache.
Why does Series X soc get 76mb?
My full desktop 16c/32t 3950x has only 72mb cache?

I wonder what you get pairing 352b memory bandwidth with 64mb soc cache in Series x? lots more power?

can someone with twitter account ping/dm MS guys? Let's get an official answer from official sources.

FYI, XSX's 8 core Zen 2 is the mobile variant e.g. L2+L3 = 16MB.

PC's Zen 2 APU has a lower latency when compared to the non-APU Ryzen 3000 series.

 

[Explosive Zombie]

Does not compute? Pro, released in 2016 @ $399.99. Sold more compared to the Xbox One X. Third Party games released for the most part had no discernible difference between both - some were even optimized better for the Pro (a lower spec console). Pro stopped potential hardcore PS gamer flight to PC, and uplifted Sony exclusives as graphical benchmarks for the industry vs. the non-existent competition, including vs. PC. Xbox One X, released in 2017 @ $499.99, discontinued 2 years after, didn't sell very well. Did not turn the gen around for MS despite having some third party multiplats looking slightly better than the Pro.

The Pro did what it was designed to do from a strategic standpoint. It however showed to be a niche offering.

Pro wasn't enough for me. I did get one but I got a 1070 first. And then I found the Pro... kinda lukewarm. I had dreams for it, honestly, was a big Destiny fan at the time and somehow assumed the Pro could get the game to run at 60FPS but... nope... and certainly not for D2. Chasing 4K was just so silly, tbh, they're barely going to be able to do it with PS5 but the Pro was trying with its jaguar CPU? LOL. You're generally right that the Pro sold better than X but I feel like neither put as big a dent in the sales charts as Sony/Microsoft hoped for.

 

[rnlval]

I don't doubt this but was just making the point that AMD GPU roadmap features can and do make it in to the consoles well before they have in AMD dGPUs e.g. Vega features in PS4 Pro before PC Polaris GPUs were out.

And from what Mark Cerny went out of his way to describe in R2PS5 it could be interpreted as the collaboration this time was even deeper with Mark/Sony putting more of their ideas in to the design that will then be incorporated in to RDNA3?

FYI, Microsoft has its own customization which focused on rasterization areas e.g. high-speed 2MB render cache for the ROPS and early variable-rate shading like a solution.

From https://www.eurogamer.net/articles/digitalfoundry-2017-the-scorpio-engine-in-depth

However, Andrew Goossen tells us that the GPU supports extensions that allow depth and ID buffers to be efficiently rendered at full native resolution, while colour buffers can be rendered at half resolution with full pixel shader efficiency.

Xbox One X already has a type VRS like feature since 2017.

XBO has additional customization to support FP10 from the XBox 360.

Improve pixel shader/rasterization path = better results with NVIDIA's Gameworks titles and namely, Unreal Engine 4.

Sony was focusing on GPU's compute path e.g. compute path's Vega like rapid pack math on PS4 Pro. PS4 Pro runs into memory bandwidth issues worst than X1X. Programming on PS4 Pro needs to be compute shader/TMU path since it's connected to 2MB of L2 cache and it should be software tiled render methods.

RX 6800's 96 ROPS connected to a very fast 128 MB L3 cache yields a powerful rasterization path. RX 6800 XT has 128 ROPS connected to a very fast 128 MB L3 cache.

TMUs can be used ROPS like usage, but without MSAA hardware support.

 

[THE:MILKMAN]

FYI, Microsoft has its own customization which focused on rasterization areas e.g. high-speed 2MB render cache for the ROPS and early variable-rate shading like a solution.

From https://www.eurogamer.net/articles/digitalfoundry-2017-the-scorpio-engine-in-depth

However, Andrew Goossen tells us that the GPU supports extensions that allow depth and ID buffers to be efficiently rendered at full native resolution, while colour buffers can be rendered at half resolution with full pixel shader efficiency.

Xbox One X already has a type VRS like feature since 2017.

XBO has additional customization to support FP10 from the XBox 360.

Improve pixel shader/rasterization path = better results with NVIDIA's Gameworks titles and namely, Unreal Engine 4.

Sony was focusing on GPU's compute path e.g. compute path's Vega like rapid pack math on PS4 Pro. PS4 Pro runs into memory bandwidth issues worst than X1X. Programming on PS4 Pro needs to be compute shader/TMU path since it's connected to 2MB of L2 cache and it should be software tiled render methods.

Again, I don't question any of this but this is a thread about PS5 possibly having RDNA 3 feature(s) or even features that make in to future AMD GPUs.

 

[rnlval]

Well, they helped them to make RDNA anyway:

Sources: AMD Created Navi For Sony's PlayStation 5, Vega Suffered

Sources reveal not only that the PS5 uses Navi graphics, but nearly 2/3 of AMD's engineering resources were devoted to the task.

www.forbes.com

Xbox and Playstation game consoles have up to 2/3 of RTG's human resource with some Zen engineers helping.
MS and Sony have their request changes.

PC has 1/3 of RTG's human resources. Raja "Mr TFLOPS" Koduri shouldn't complain.

Vega 56/64 design has server compute bias and certain Bollywood use cases (e.g. high bandwidth cache, useless for raster games, high latency) and it didn't have raster gaming as its primary focus.

Vega 56/64 design has a higher latency pipeline when compared RDNA v1 and both projects can run GCN instructions.

 

[thelastword]

This cooperation doesn't seem to have worked very well for the Pro given the One X existence.

The One X came much later. It had more vram and bandwidth, yet still had less rops. If Sony had as much Vram as One X with PRO's superior customized architecture, meaning better fp16 support, more rops, CB hardware, ID buffer etc.....It would have outperformed XBONEX on even more titles than it did at similar resolutions to XBONEX....

As it stands, PRO has had so many wins over ONE X in the framerate department. It's exclsuive games are really the best looking ones all gen long, that is when first party devs really used it's customizations. So there's more to it than TF, which will be seen again this gen on PS5, even moreso, because this is a new gen with much more customizations this time around..

 

[rnlval]

Again, I don't question any of this but this is a thread about PS5 possibly having RDNA 3 feature(s) or even features that make in to future AMD GPUs.

The majority of PC gaming workloads are raster path bias. I rather see rasterization and lower graphics pipeline latency improvements.

Sony has a CELL SPU programming direction and may not suit NVIDIA's Gameworks. Sony's CELL programming mentality is evident in PS4 Pro's compute shader path bias design direction.

 

[rnlval]

The One X came much later. It had more vram and bandwidth, yet still had less rops. If Sony had as much Vram as One X with PRO's superior customized architecture, meaning better fp16 support, more rops, CB hardware, ID buffer etc.....It would have outperformed XBONEX on even more titles than it did at similar resolutions to XBONEX....

As it stands, PRO has had so many wins over ONE X in the framerate department. It's exclsuive games are really the best looking ones all gen long, that is when first party devs really used it's customizations. So there's more to it than TF, which will be seen again this gen on PS5, even moreso, because this is a new gen with much more customizations this time around..

X1X's 32 ROPS is coupled with 2MB render cache and higher memory bandwidth.

Do the math on 32 ROPS vs memory bandwidth.

For X1X's 32 ROPS

RGBA32F: 1172 Mhz x 32 x 16 bytes = 600 GB/s (X1X 32 ROPS has 2MB L2 render cache).

RGBA16F: 1172 Mhz x 32 x 8 bytes = 300 GB/s

RGBA8 (integer): 1172 Mhz x 32 x 4 bytes = 150 GB/s. (use TMU to shift bottleneck towards memory bandwdith).

PS4 Pro's 64 ROPS is nearly pointless i.e. runs into memory bandwidth problems.

 

[thelastword]

X1X's 32 ROPS is coupled with 2MB render cache and higher memory bandwidth.

Do the math on 32 ROPS vs memory bandwidth.

For X1X's 32 ROPS

RGBA32F: 1172 Mhz x 32 x 16 bytes = 600 GB/s (X1X 32 ROPS has 2MB L2 render cache).

RGBA16F: 1172 Mhz x 32 x 8 bytes = 300 GB/s

RGBA8 (integer): 1172 Mhz x 32 x 4 bytes = 150 GB/s. (use TMU to shift bottleneck towards memory bandwdith).

PS4 Pro's 64 ROPS is nearly pointless i.e. runs into memory bandwidth problems.

That is what I'm saying, there wasn't enough memory/bandwidth to fully utilize the 64 rops. If they had the bandwidth, then PRO would have had much more 4k native games or higher resolutions. All the games where PRO is performing better than XBONEX at lower resolutions, they would have been able to do it at the same resolutions as XBONEX or even higher....

There is a reason why PS4 was a 1080p machine, it was due to the number of rops that was well fed.....Same could have been on PRO, but Sony had to maintain a $400 price point. If they went with $500 for PRO in 2016 just like MS did a whole year later for XBONEX, they would have had the superior machine for resolutions too...

 

[Lysandros]

That is what I'm saying, there wasn't enough memory/bandwidth to fully utilize the 64 rops. If they had the bandwidth, then PRO would have had much more 4k native games or higher resolutions. All the games where PRO is performing better than XBONEX at lower resolutions, they would have been able to do it at the same resolutions as XBONEX or even higher....

There is a reason why PS4 was a 1080p machine, it was due to the number of rops that was well fed.....Same could have been on PRO, but Sony had to maintain a $400 price point. If they went with $500 for PRO in 2016 just like MS did a whole year later for XBONEX, they would have had the superior machine for resolutions too...

If only PS4 Pro had 256 GB/s of bandwidth instead of 218 GB/s the situation would be much better coupled with 64 ROPS i think. Sony went a bit too far trying to be cost effective in that area.

 

[rnlval]

If only PS4 Pro had 256 GB/s of bandwidth instead of 218 GB/s the situation would be much better coupled with 64 ROPS i think. Sony went a bit too far trying to be cost effective in that area.

ROPS needs to be feed by CU and memory bandwidth.

 

[Lysandros]

ROPS needs to be feed by CU and memory bandwidth.

Yes and?..

 

[MasterCornholio]

I never said there will be an ABYSSAL difference but one will have a clear advantage ... but the real problem here is that consoles cost the same price. I've explained 100 times why I think it's a last year (or around there) rework. and I'll repeat it again if you want.The console didn't changed "design" they added the variable clock part and upclocked it ...the console was designed to have lower (not by a lot) performance and the same cu's number. The target was probably those 9.something TFs that were rumored to be Oberon (github leak). Then we start to see the first real images of the ps5 and it's monstrously big ... the biggest console in history. Then comes the official teardown and we find out that on the official ps5 SOC we have the same AMD Oberon code. All this is even stranger if we compare the two consoles .... less powerful,not contemplated as full rdna2, much larger, consumes more watts with less silicon and cost the same.. a design gap that is not often seen.
We heard ms talking about waiting amd for the new full rdna2 architecture and this point out again to oberon and how it was possible for Sony to have the soc already in his hands. We have millions of rumors about how ps5 devkits were already in the hands of developers when no Xbox devkits were out yet...etc etc etc
the liquid metal the enormous and expensive heatsink to mitigate the clock problem. Sony could not present the ps5 with 9 tf. we all know how long the enthusiast would see it as a shit console and the echo chamber generated would only create problems. If this is little for you, I don't know what else to say

 

[pratyush]

Sony fanbois are the saddest bunch of people.

RDNA 3 is coming out in 2022 (or very late 2021).

The first PS5 SOC engineering samples were ready in very late 2018, before the RDNA1 even came out.

Do you really believe that AMD had anything related to RDNA 3 available in late 2018 / early 2019 for inclusion in PS5? Years before RDNA 3 is coming out? Do you really?

Design for RDNA3 must already be in its last stages. From now on there will be incremental release of RTL for backend design. Most of the product that you see getting released are designed 2-3 years before. It's called a roadmap. Also that is not what people are mentioning here. Features are used in future products all the time. It's nothing new.

 

[Jon Neu]

Does not compute? Pro, released in 2016 @ $399.99. Sold more compared to the Xbox One X. Third Party games released for the most part had no discernible difference between both

 

[Duchess]

I think we should go back to the way we did it in the 90s and compare specs based on how many flat shaded polygons per second a system can push

 

[duhmetree]

Sony fanbois are the saddest bunch of people.

RDNA 3 is coming out in 2022 (or very late 2021).

The first PS5 SOC engineering samples were ready in very late 2018, before the RDNA1 even came out.

Do you really believe that AMD had anything related to RDNA 3 available in late 2018 / early 2019 for inclusion in PS5? Years before RDNA 3 is coming out? Do you really?

Well technically its Sony's feature set made specifically for the PS5. Since they work closely with AMD it may inspire a RDNA3 feature set.... you know, since its allegedly more advanced... keep up.

It's funny because I see the exact opposite. I dont have a horse in the race. I game predominantly on PC. Microsoft warriors gobble up the PR... They start war chanting the PR buzzwords like zealots. Punching their chests FULL RDNA2... Backword Compatibility... These are the things championed... and I couldn't care less. Sadder, all the 'negative' Sony threads, from what I've seen, are started by Microsoft warriors... suspicious to say the least. Objectivity has gone out the window

 

[longdi]

It's the 5MB of L2 cache that we already know about

Interesting, but where are the 64mb cache situated than?

 

[Clear]

Newsflash: RDNAx is not a thing in itself. Its just a branding that incorporates a number of architectural innovations and advances (not all of which will be directly interrelated) that AMD has decided to incorporate into a single product tier

 

[Concern]

Well technically its Sony's feature set made specifically for the PS5. Since they work closely with AMD it may inspire a RDNA3 feature set.... you know, since its allegedly more advanced... keep up.

It's funny because I see the exact opposite. I dont have a horse in the race. I game predominantly on PC. Microsoft warriors gobble up the PR... They start war chanting the PR buzzwords like zealots. Punching their chests FULL RDNA2... Backword Compatibility... These are the things championed... and I couldn't care less. Sadder, all the 'negative' Sony threads, from what I've seen, are started by Microsoft warriors... suspicious to say the least. Objectivity has gone out the window

Post history determined that was a lie.


We have two threads spawned (this one included) over the amd event saying xsx is the only full RDNA2 next-gen console. Based off nothing but fanboy rumors. This guy said both consoles would have RDNA3 features too lol.

But when so much crow is served, you must grasp for whatever straws you can.

 

[duhmetree]

Post history determined that was a lie.


We have two threads spawned (this one included) over the amd event saying xsx is the only full RDNA2 next-gen console. Based off nothing but fanboy rumors. This guy said both consoles would have RDNA3 features too lol.

But when so much crow is served, you must grasp for whatever straws you can.

what crow? ACKTHUALLY, I've gone through your post history...

Consoles - Sony is the better product. Microsoft offer me nothing of interest. IF THEY DID, I would not fight against it. I dont care about RDNA2 feature sets.. I am getting a 3080. I dont care about NATIVE 4k. because I dont even game at 4K on PC. I dont care about BC because I dont play old games. I haven't used a Console in 4ish years? (edit- thats a lie I played GoW4) Instead of pandering to their zealots, how about Microsoft produce SOME ACTUAL FUCKING GAMES.

Guess when I play on a Sony PlayStation.... take a guess... WHEN THEIR 'CANT MISS 1ST PARTY GAMES' RELEASE. Is that so hard to comprehend? Are you that emotionally contorted? and I haven't even played TLoU2, SpiderMan or GoT. I would say I'm in for a treat. If only I had FULL RDNA2 It would be that much sweeter

 

[El Pistolero]

Sony fanbois are the saddest bunch of people.

RDNA 3 is coming out in 2022 (or very late 2021).

The first PS5 SOC engineering samples were ready in very late 2018, before the RDNA1 even came out.

Do you really believe that AMD had anything related to RDNA 3 available in late 2018 / early 2019 for inclusion in PS5? Years before RDNA 3 is coming out? Do you really?

Yes, they freaking do. Do you think they will toss the whole RDNA/RDNA 2 basis and start anew? Do you believe the architecture refinements have not been worked on for quite sometime now? What?

 

[KungFucius]

No way I am reading this. Why would Sony let AMD use something it patented outside of a PlayStation GPU? If they did it must be because Sony thought of the idea but needed AMD to develop the implementation and then allowed them to practice the invention as "payment" but then reserved the right to let nobody else use it.

I think it is more likely that it is not something that will show up in RDNA3 and will be Sony only.

 

[Concern]

what crow? ACKTHUALLY, I've gone through your post history...

Consoles - Sony is the better product. Microsoft offer me nothing of interest. IF THEY DID, I would not fight against it. I dont care about RDNA2 feature sets.. I am getting a 3080. I dont care about NATIVE 4k. because I dont even game at 4K on PC. I dont care about BC because I dont play old games. I haven't used a Console in 4ish years? Instead of pandering to their zealots, how about Microsoft produce SOME ACTUAL FUCKING GAMES.

"Fight against" lol. You dont care about consoles but are "fighting against" some imaginary war?

You don't care about RDNA2 but joined this hilarious dick measuring contest based off a fanboy youtubers rumors.

Again you claim you don't care about consoles but are crying about console games. Although I agree on this front. This is the weakest next-gen launch ever. That's why I ain't buying neither till next year at minimum.

 

[Audiophile]

Both systems are RDNA2-based. But there's nothing odd about some specific featuresets from RDNA3 being in the PS5 or XSX. Nor would there be anything surprising about PS5 features making their way to RDNA3. There's precedent in both respects and it's even more likely now than in the past given the increasingly collaborative nature of the relationship between AMD and Sony/MS.

Also, Cerny very clearly alluded to this point in "The Road to PS5".

 

[geordiemp]

No way I am reading this. Why would Sony let AMD use something it patented outside of a PlayStation GPU? If they did it must be because Sony thought of the idea but needed AMD to develop the implementation and then allowed them to practice the invention as "payment" but then reserved the right to let nobody else use it.

I think it is more likely that it is not something that will show up in RDNA3 and will be Sony only.

Same for both consoles, there are 2 teams lets call them green and blue.

If team blue has an idea, AMD implement with team blue on silicon for them and can use it later if AMD wish in PC designs. ]

Same for team green. 2 teams that are independent for a few years.

So both consoles will have difefrences, and some of them will be used by AMD for PC, some wont.