Support NeoGAF

[Physiognomonics]

If it wasn't for the fact that this will be more about Ray-Tracing I would be worried about the 2 SE


Going from 18 CUs per SE to 30 CUs per SE is crazy

They added more L0 and L1 cache (doubled them) than the increase of CU by SE actually. Knowing Cerny obsession with efficiency, I wouldn't worry.

 

[Rat Rage]

PS5 Pro is gonna be a massive failure. Sony's soulless, trend-chasing business suits have been carried by their "bigger, stronger hardware and bigger, better graphics" strategy for far too long. AAA gaming is dying, nobody needs or actually wants more console power these days - people just want good and fun games that don't fuck them over. Graphics have never mattered less than right now (in the home console space).
Due to this massive lack of forsight, Sony has maneuvered their car into a dead end, and now their strategy is to replace their car engine with a bigger one and continue to drive deeper into the dead end.

 

[Shwing]

PS5 Pro is gonna be a massive failure. Sony's soulless, trend-chasing business suits have been carried by their "bigger, stronger hardware and bigger, better graphics" strategy for far too long. AAA gaming is dying, nobody needs or actually wants more console power these days - people just want good and fun games that don't fuck them over. Graphics have never mattered less than right now (in the home console space).
Due to this massive lack of forsight, Sony has maneuvered their car into a dead end, and now their strategy is to replace their car engine with a bigger one and continue to drive deeper into the dead end.

I do!

 

[hououinkyouma00]

PS5 Pro is gonna be a massive failure. Sony's soulless, trend-chasing business suits have been carried by their "bigger, stronger hardware and bigger, better graphics" strategy for far too long. AAA gaming is dying, nobody needs or actually wants more console power these days - people just want good and fun games that don't fuck them over. Graphics have never mattered less than right now (in the home console space).
Due to this massive lack of forsight, Sony has maneuvered their car into a dead end, and now their strategy is to replace their car engine with a bigger one and continue to drive deeper into the dead end.

This is one of the dumbest posts I've seen in a long time.

 

[twinspectre]

how exactly are their customers being screwed over

I thought it was a simple concept like "You support this BS, then you have no rights to complain".

 

[onQ123]

They added more L0 and L1 cache (doubled them) than the increase of CU by SE actually. Knowing Cerny obsession with efficiency, I wouldn't worry.

I'm not worried this will be a Ray-Tracing monster I just wouldn't expect much more on the Rasterizing & geometry side besides the the gains from having newer hardware for Mesh Shading & VRS + more CUs

If this was the beginning of a generation & random codes was going to be thrown at it I would be thinking that will bite them in the butt in some games but this is a pro console & we know they will not be throwing much more geometry & stuff at the PS5 Pro than they already are with the PS5 .

 

[Leonidas]

Sad to see that PS5 Pro is power limited.

Seems like they really did stick with 6nm, such a shame, if that's the case.

I wonder if there will ever be a 5nm refresh.

 

[HeisenbergFX4]

PS5 Pro is gonna be a massive failure. Sony's soulless, trend-chasing business suits have been carried by their "bigger, stronger hardware and bigger, better graphics" strategy for far too long. AAA gaming is dying, nobody needs or actually wants more console power these days - people just want good and fun games that don't fuck them over. Graphics have never mattered less than right now (in the home console space).
Due to this massive lack of forsight, Sony has maneuvered their car into a dead end, and now their strategy is to replace their car engine with a bigger one and continue to drive deeper into the dead end.

Couldn’t agree more, people don’t buy this piece of shit , show them we won’t stand for this crap

That way I can buy two of everyone else boycotts

 

[Dr.brain64]

Day one. Haters be hating.

 

[Brigandier]

Can't wait for this to finally be announced... Hopefully Cerny does a presentation to explain stuff.

 

[Ashamam]

Sad to see that PS5 Pro is power limited.

Every console ever made is power limited. It goes hand in hand with cost and form factor, two of the most important aspects of a console. If you want unlimited power build a PC.

As an aside upcoming APU's might even let you build your own SFF Steam/GP Box that is equivalent to a PS5 but is customisable. First time I've been tempted to do something like that. Will be watching the Strix Halo closely. Might depend on how FSR progresses though.

 

[reinking]

Sad to see that PS5 Pro is power limited.

Seems like they really did stick with 6nm, such a shame, if that's the case.

I wonder if there will ever be a 5nm refresh.

 

[bitbydeath]

Sad to see that PS5 Pro is power limited.

Seems like they really did stick with 6nm, such a shame, if that's the case.

I wonder if there will ever be a 5nm refresh.

It was always just going to be built for Res/FPS upgrades. Don’t need top-end hardware for that.

 

[SonGoku]

That is true. But there has top be something more to it.
Remember that RDNA1 on PC has trouble running a game like Alan Wake, with Mesh Shaders. But RDNA2 runs it just fine.

hmmm if i remember correctly Alan Wake even on its lowest settings uses RT which cannot be disabled
That would be a more likely explanation for the performance discrepancies seen on RDNA1

 

[SlimySnake]

Sad to see that PS5 Pro is power limited.

Seems like they really did stick with 6nm, such a shame, if that's the case.

I wonder if there will ever be a 5nm refresh.

they are not putting a 6nm 60 cu chip in a console at 2.35 ghz. or even 2.185 ghz. A) its too big and expensive. B) it will be well over 300 watts for the GPU alone. C) impossible to cool.

they are probably on 5nm and dont want it to go over 250 watts. ps5 already goes to 230 watts on rare occasions. they will likely cap at 250 watts this time around.

 

[SlimySnake]

hmmm if i remember correctly Alan Wake even on its lowest settings uses RT which cannot be disabled
That would be a more likely explanation for the performance discrepancies seen on RDNA1

no it was the lack of mesh shader support on older GPUs. Not RT. The game does not use RT on consoles.

There was a patch released recently that improved performance on cards that dont support mesh shaders. it runs much better now.

 

[sachos]

I want the PS5 Pro just to hear Cerny give a tech talk again.

 

[Panajev2001a]

That is true. But there has top be something more to it.
Remember that RDNA1 on PC has trouble running a game like Alan Wake, with Mesh Shaders. But RDNA2 runs it just fine.

It also ran quite fine on PS5

So why is the primitive Shaders in RDNA1 not good enough to be used with Mesh Shaders? Is it just drivers? Some hardware feature missing?

It is probably a mix of both. It would not surprise me that RDNA1’s version would lack some small feature or had a HW bug (or both) as AMD was frantically competing with nVIDIA to get the Mesh Shader spec based on their proposal and not nVIDIA’s.

I do wonder what they have added or changed in PS5 Pro to make the performance and compatibility of Mesh Shaders better and if it is just to bring it in line with XSX or beyond (I guess the latter is true because even now the performance on titles using Mesh Shaders was not bad on PS5, but then again… PS5 Pro is a general refinement, not a new generation).

In this article AMD talks about they integrate AMDs Primitive Shaders into the Mesh Shader pipeline.

From vertex shader to mesh shader - Mesh shaders on AMD RDNA™ graphics cards

This post is the start of a new series which aims to demystify mesh shaders through examples and tutorials.

gpuopen.com

.

 

[Panajev2001a]

The problem with the Pentium's 4 pipeline was not the branch predictor, it was the length of the pipeline. It go to the point of being 30+ stages.
So when there was a stall because of a miss with the OoO, the resulting pipeline flush was catastrophic for all operations already in the pipeline.

I know, but that was the point. Even 95% accuracy can not be enough if the cost of a mistake is very dire. Either a hazard when getting to instruction execution or a branch misprediction or a cache miss could wreak havoc (isolated or together).

Now mind you, I'm not saying that a good frontend can nullify cache and memory latencies. I'm saying it can help to hide it.

OoO is technically seen as backend, but the point was that OoO (Issue, Execution but in order retire) with regards to L2 misses is more similar to trying to empty the sea with a bucket. It helps too .

When it works well, a good branch predictor, pre-fetcher and OoO, can do wonders to keep a superscaler CPU with several pipelines, all running at peak performance.

Agreed .

You might be mistaking what OoO means. It's just Out of Order execution.

I am not confused . OoO Issue and Execute looks as far ahead in the instruction stream (speculating away when it comes to branches) reordering instructions, analysing dependencies between instructions, resources conflict (limited architectural registers) and detecting hazards that could cause a pipeline flush. It does it… to extract work… to extract available ILP.

The feature in modern CPUs that allow for greater parallelism is superscaler, meaning there are several pipelines in the CPU. Though, of course that each stage has different amounts of units.

Superscalar, ability to have multiple concurrent instructions being processed at any one time came much earlier than OoO.
It is a mechanism that allows parallel work to occur and in a way this means we are extracting ILP, but it is not the only contributing factor. Pentium was superscalar and so was Itanium. In either cases further work to extract ILP was left to the compiler without runtime assistance.

SMT is just an opportunistic feature that tries to fit secondary instructions, in a pipeline that already has different thread on it, but that ha left unused units.
In a way, good frontend and having SMT are counter productive, because a good frontend will leave fewer stages open in the pipeline.

SMT allow, for like 4-5% area cost, to take advantage of the existing mechanism for ILP maximisation to take advantage of multiple threads of execution to keep the instruction units fed as the current mechanisms to do so with a single stream were hitting diminishing returns.

It is to be expected that as technology changes assumptions are tested / challenged. Depending on your design and intended workload SMT may make or may not make sense.

That is why Intel is ditching SMT for it's future CPU architectures. And why Apple already left it behind.

Apple has not implemented it yet, it remains to be seen if they may introduce it later on given how wide their cores are they seem to be prime to do it (they are milking CPU core redesigns quite a lot… both M2 and M3 feel very slowly evolutionary rather than revolutionary designs… GPU cores and NE is where they invested more of their design resources, CPU wise they rebalanced it and added resources [beefing everything up, tons and tons of cache and putting optimised memory very very close to the SoC]).

For a while, many supposed that X86 could only have an 4-wide decode stage. While ARM isa, could have significantly more.
Intel proved everyone wrong, by having a 6-wide decode stage since 12th gen.
AMD is still with a 4-wide decode, but with strong throughput. I wonder what they will do with Zen5 regarding this.

Yes, it is a very interesting area of evolution .

 

[Panajev2001a]

If it wasn't for the fact that this will be more about Ray-Tracing I would be worried about the 2 SE


Going from 18 CUs per SE to 30 CUs per SE is crazy

They have beefed up the efficiency of the memory controllers and increased RAM bandwidth so maybe the global shared L2 graphics cache not being boosted to over 4 MB was not something that would hurt them too much (nice to have, but maybe not worth the cost in a Pro design).

L0 vector cache inside the CU being doubled and the L1 cache per Shader Array (each shader array only grew by 3 CUs over PS5’s ones and their L1 doubled).

 

[winjer]

OoO is technically seen as backend, but the point was that OoO (Issue, Execution but in order retire) with regards to L2 misses is more similar to trying to empty the sea with a bucket. It helps too .

The slip between the CPU frontend and backend is done by the units it has.
The Frontend is the fetch+decode stages. The backend are the execution units.
Something like the OoO is in the frontend.

 

[Panajev2001a]

The slip between the CPU frontend and backend is done by the units it has.
The Frontend is the fetch+decode stages. The backend are the execution units.
Something like the OoO is in the frontend.

OoO issue and execute are firmly backend. It is not involved before the instruction issue stage which is not part of the front end. Again, minor details , seems more philosophical than anything.

AMD’s Zen 4 Part 1: Frontend and Execution Engine

AMD’s Zen 4 architecture has been hotly anticipated by many in the tech sphere; as a result many rumors were floating around about its performance gains prior to its release. In February 2021…

chipsandcheese.com

 

[Leonidas]

Every console ever made is power limited. It goes hand in hand with cost and form factor, two of the most important aspects of a console. If you want unlimited power build a PC.

It could have been much less power limited if it used a more advanced node.

Last gen both Pro consoles got bigger boosts than what we're seeing this time.

If you want unlimited power build a PC.

I did, feels good to not be held back by console APU power limits

they are not putting a 6nm 60 cu chip in a console at 2.35 ghz. or even 2.185 ghz. A) its too big and expensive. B) it will be well over 300 watts for the GPU alone. C) impossible to cool.

they are probably on 5nm and dont want it to go over 250 watts. ps5 already goes to 230 watts on rare occasions. they will likely cap at 250 watts this time around.

Its a shame that going to 5nm didn't improve things more.

 

[FireFly]

Its a shame that going to 5nm didn't improve things more.

The 7800 XT only averages 2425 MHz at 250W. I don't see how it's realistic to expect a ~180W part to achieve similar (or greater) clock speeds without a big boost in performance per watt. The PS5 also clocked below equivalent PC GPUs.

 

[MisterXDTV]

Sad to see that PS5 Pro is power limited.

LOL

A console is power limited....

Big news at 11

 

[Panajev2001a]

It could have been much less power limited if it used a more advanced node.

Design and manufacturing costs for those nodes and availability are a constraint (easier when you are trying to target a much much much higher MSRP, but you know that). The free lunch is over everywhere. If you still design a chip thinking semiconductor scaling will save you, you are doing it wrong.

Thanks for the PCMR trolling contribution as it would have been expected by your or rnvival .

 

[The ninja farmer]

they are not putting a 6nm 60 cu chip in a console at 2.35 ghz. or even 2.185 ghz. A) its too big and expensive. B) it will be well over 300 watts for the GPU alone. C) impossible to cool.

they are probably on 5nm and dont want it to go over 250 watts. ps5 already goes to 230 watts on rare occasions. they will likely cap at 250 watts this time around.

This for me. Although I do think that when df first looked through the docs. They said power and clock speed look to be thermally limited by the cooling and node being used

(I might be wrong as to who said it)

I am surprised that the rumours talk of 6 process for the pro. They perhaps the design is not fully out there, in this regard. Although Sony obviously know exactly what the target is

This is I guess the key part of the pro and what it can do, for me. As the chip size and layout dictates the key concept of the ps5 family. Efficiency in power & data management. this controls everything else in their design i believe

I would have thought 5np or possibly 4 would have allowed for greater clocks speeds at the same or less power as the ps5. Maybe that’s where the 2500-2800 frequency rumours came from. Chip testing and cost analysis from amd/sony engineering

I get that this is also a live beta test for pssr and to see what devs can do with some extra rt grunt

I will be getting. But just surprised that the node drop is more limited (guess cost or availability might factor here). Maybe again cost and impact on any ps6 has to be considered (like the previous pro and what it did or didn’t bring and what shaped the Sony design from there to the ps5)

 

[Zathalus]

The Pro might very well be a combination of 6nm/5nm using chiplets. Like RDNA3 is. CPU/IO/MCD is all on 6nm while the GCD is 5nm.

 

[mckmas8808]

So 15tf on top of hw acceleration for compute work that teraflop metrics traditionally cover. So more like 50+TF equivalent.



considering base PS5 often performs bettet than Series X Cyberpunk RT games, I'll say you're coming in WAY too low with the 70% higher than Series X" comparison

Now, lets not get crazy.

 

[ChiefDada]

Now, lets not get crazy.

Too late

But seriously, I'm just estimating equivalent compute performance that would be needed if the Pro didn't have RT hardware. No different than how Nvidia communicate is it. And tbf, it is a reasonsable way to look at it.

 

[Unrivaledx88]

Tom Henderson is never wrong.

Was it him that said Switch 2 this yr as well ? Doesn’t look likely now

 

[Euler007]

PS5 Pro is gonna be a massive failure. Sony's soulless, trend-chasing business suits have been carried by their "bigger, stronger hardware and bigger, better graphics" strategy for far too long. AAA gaming is dying, nobody needs or actually wants more console power these days - people just want good and fun games that don't fuck them over. Graphics have never mattered less than right now (in the home console space).
Due to this massive lack of forsight, Sony has maneuvered their car into a dead end, and now their strategy is to replace their car engine with a bigger one and continue to drive deeper into the dead end.

This has to be satire.

 

[twinspectre]

Day one. Haters be hating.

And this is how this company continues to screw people.

 

[Zathalus]

And this is how this company continues to screw people.

How does a fully optional upgrade screw anyone? Your regular PS5 won't suddenly stop working.

 

[SlimySnake]

Was this posted here?

I am not seeing a big upgrade here. Were there other screenshots in his video?

 

[twinspectre]

How does a fully optional upgrade screw anyone? Your regular PS5 won't suddenly stop working.

I think "explaining it" could potentially waste my time.

 

[Gaiff]

Was this posted here?

Yes. On the day of the leak by MLID.

I am not seeing a big upgrade here. Were there other screenshots in his video?

No, but this is a shitty screenshot of a still image.

 

[Metnut]

Is this going to be easy to buy at launch or will it be like the PS5 launch all over again?

 

[saintjules]

That way I can buy two of everyone else boycotts

This you?

 

[HeisenbergFX4]

This you?

 

[farmerboy]

No, but this is a shitty screenshot of a still image.

Even in this shitty screenshot of a still image, there is a clear improvement. Just look around Clank's eyes and Rachet"s armband.

Not that I've ever really cared about these kinds of nitpicks. On a moving screen 8ft away on the couch to my 48yo eyes it's all the same.

 

[Lysandros]

Why is everyone taking Dictator's comments/framing at face value and assuming PS5 Geometry Engine is a vanilla RDNA1 part and "less advanced than XSX' RDNA2" and making analysis based on that all of a sudden? This was always his narrative from the beginning of the generation, along with other claims about variable clocks, hardware rt etc. all proven to be fase. Is this the "give credit time", is he redeemed now?..

 

[IDWhite]

Let's make something clear - caches are designed for >90% hitrates. Making them larger doesn't dramatically change the hitrates (unless the balance was really broken somewhere) - it just smooths out the 'bumps'. And where and how often such 'bumps' typically happen is discovered through profiling and statistics, which as you say - Sony and other companies have tons of data on to base such decisions from.
Eg. PS360 CPUs had massive (for their time) L2 caches but that made precious little difference because their memory subsystem design made them fall over on anything that missed L1. Even without going to main-memory you could get code running slower than 300mhz CPU of the previous gen.


Yea but that's just stating the obvious. Caches themselves are a trade-off - if high performance memory was plausible/not cost prohibitive, hw makers would gladly skip the caches alltogether.

 

[Physiognomonics]

The Pro might very well be a combination of 6nm/5nm using chiplets. Like RDNA3 is. CPU/IO/MCD is all on 6nm while the GCD is 5nm.

I think It would create a more expensive packaging overall. An APU is still the less expensive solution for a console.

Why is everyone taking Dictator's comments/framing at face value and assuming PS5 Geometry Engine is a vanilla RDNA1 part and "less advanced than XSX' RDNA2" and making analysis based on that all of a sudden? This was always his narrative from the beginning of the generation, along with other claims about variable clocks, hardware rt etc. all proven to be fase. Is this the "give credit time", is he redeemed now?..

I think some dev might have told him PS5 "mesh shaders" lack one small feature vs RDNA2 which is possibly true (but never was confirmed AFAIK by any dev). Nonetheless even if that is true it's not making a noticeable difference as PS5 still have mesh shaders well implemented in a few games and was actually the first console to use primitive shaders in games. Dictator is just using this to imply PS5 being the "inferior" hardware versus his PCMR model. He is doing the same shit with VRS and the other useless feature I don't remember the name, litterally an useless feature not even used in MS own games.

Was this posted here?

I am not seeing a big upgrade here. Were there other screenshots in his video?

MLID showed the original comparison pic to a dev who said there was a noticeable difference. But even in this compressed shot there is actually a noticeable difference when you know what to look for (Clank's body and lips). Realistically PSSR should have similar results as the first implementations of DLSS. At this point I don't think DF are even doubting about that.

 

[TheRedRiders]

I think some dev might have told him PS5 "mesh shaders" lack one small feature vs RDNA2 which is possibly true (but never was confirmed AFAIK by any dev). Nonetheless even if that is true it's not making a noticeable difference as PS5 still have mesh shaders well implemented in a few games and was actually the first console to use primitive shaders in games. Dictator is just using this to imply PS5 being the "inferior" hardware versus his PCMR model. He is doing the same shit with VRS and the other useless feature I don't remember the name, litterally an useless feature not even used in MS own games.

The whole Primitive Shader RDNA 1 thing was pretty much debunked by how AW2 performed on PS5 vs Series X, it performed exactly as expected if we consider the PS5's GPU and Geometry Engine the same as RDNA 2.

We already saw how AW2 performed on the 5700 XT (RDNA 1 with Primitive Shaders only), and it performed very poorly.

The only conclusion we can come too, and which is consistent with developer comments, AMD documentation and driver code is that the Primitive Shaders on RDNA 2 (including the PS5) run and perform the same as Mesh Shaders. The difference lies in the software implementation.

 

[Zathalus]

Why is everyone taking Dictator's comments/framing at face value and assuming PS5 Geometry Engine is a vanilla RDNA1 part and "less advanced than XSX' RDNA2" and making analysis based on that all of a sudden? This was always his narrative from the beginning of the generation, along with other claims about variable clocks, hardware rt etc. all proven to be fase. Is this the "give credit time", is he redeemed now?..

Primitive shaders appear to be less flexible in usage compared to mesh shaders. They end up doing the same thing but programmatically it appears the mesh shader is just more flexible in usage. The reported statement on the Sony developer platform is that it no longer has stricter requirements. I don't think there in any major performance difference between the two, but this might lead to wider adoption.

 

[Lysandros]

I think some dev might have told him PS5 "mesh shaders" lack one small feature vs RDNA2 which is possibly true (but never was confirmed AFAIK by any dev). Nonetheless even if that is true it's not making a noticeable difference as PS5 still have mesh shaders well implemented in a few games and was actually the first console to use primitive shaders in games. Dictator is just using this to imply PS5 being the "inferior" hardware versus his PCMR model. He is doing the same shit with VRS and the other useless feature I don't remember the name, litterally an useless feature not even used in MS own games.

The thing is, even if that's the case, i don't believe for sec. that he is relaying the info 'as is' without adding his own spin/framing to support his original narrative. His level of bias and unreliability in regards to playstation hardware is such a known quantity.

The very first narrative was that PS5 didn't support Mesh Shaders altogether. When developer statements proved the contrary in cases like Avatar, Alan Wake 2 and UE5 prior like you pointed out the narrative shifted to 'full support', alluding to the amplification stage/'amplification shaders' (championed in places like beyond3d). As it happens, it appears that on AMD/RDNA2 amplification shaders don't map 1:1 to the hardware either, but instead, this stage called 'task shaders' is processed via the async compute queues (ACEs) using the exact same RDNA/PS5/XSX hardware.

 

[Bojji]

Why is everyone taking Dictator's comments/framing at face value and assuming PS5 Geometry Engine is a vanilla RDNA1 part and "less advanced than XSX' RDNA2" and making analysis based on that all of a sudden? This was always his narrative from the beginning of the generation, along with other claims about variable clocks, hardware rt etc. all proven to be fase. Is this the "give credit time", is he redeemed now?..

https://www.neogaf.com/threads/alan-wake-ii-doesnt-officially-support-rdna1-or-pascal-and-older-gpus.1662607/reply?quote=268545604

You can literally get the PS5 APU as a PC with a functional GPU (sold as a cryptomining rig) and the GPU identifies as gfx1013 which is a RDNA1 GPU albeit with RT hardware.

It's RDNA1 plus RT, it even has no DP4A support that was added in RDNA1 revision (like 5500XT). It's pretty much confirmed that it doesn't support VRS, MS and SFS.

The whole Primitive Shader RDNA 1 thing was pretty much debunked by how AW2 performed on PS5 vs Series X, it performed exactly as expected if we consider the PS5's GPU and Geometry Engine the same as RDNA 2.

We already saw how AW2 performed on the 5700 XT (RDNA 1 with Primitive Shaders only), and it performed very poorly.

The only conclusion we can come too, and which is consistent with developer comments, AMD documentation and driver code is that the Primitive Shaders on RDNA 2 (including the PS5) run and perform the same as Mesh Shaders. The difference lies in the software implementation.

They just can't use PS on PC, or don't care to write code for just one series of graphics cards (5xxx). On PC this game is running in vertex shaders mode, that's why it performs much worse than 6600XT. On PS5 they use PS in similar way they are using MS on Xbox and PC, Sony provided Api for that. DX12 doesn't allow that (most likely).

There are no other games with MS on PC to compare.

 

[Lysandros]

The whole Primitive Shader RDNA 1 thing was pretty much debunked by how AW2 performed on PS5 vs Series X, it performed exactly as expected if we consider the PS5's GPU and Geometry Engine the same as RDNA 2.

We already saw how AW2 performed on the 5700 XT (RDNA 1 with Primitive Shaders only), and it performed very poorly.

The only conclusion we can come too, and which is consistent with developer comments, AMD documentation and driver code is that the Primitive Shaders on RDNA 2 (including the PS5) run and perform the same as Mesh Shaders. The difference lies in the software implementation.

I think Avatar is an even more relevant case since the game is using mesh shaders on both PS5/XSX as per developer's statement and is running a bit better on PS5.

 

[ChiefDada]

https://www.neogaf.com/threads/alan-wake-ii-doesnt-officially-support-rdna1-or-pascal-and-older-gpus.1662607/reply?quote=268545604

It's RDNA1 plus RT, it even has no DP4A support that was added in RDNA1 revision (like 5500XT). It's pretty much confirmed that it doesn't support VRS, MS and SFS.



They just can't use PS on PC, or don't care to write code for just one series of graphics cards (5xxx). On PC this game is running in vertex shaders mode, that's why it performs much worse than 6600XT. On PS5 they use PS in similar way they are using MS on Xbox and PC, Sony provided Api for that. DX12 doesn't allow that (most likely).

There are no other games with MS on PC to compare.

Are there cache scrubbers in that "literal" PS5 equivalent APU?