Support NeoGAF

[StreetsofBeige]

We know PS5 will have the best RDNA 2 features as Sony would have chosen them. And the gap will be small especially with Sony likely having an even better checkerboard implementation for the PS5

Who knew gamers put checkerboarding on their priority list of console features.

lol

 

[Hobbygaming]

Right. The point is that the game's best version will be on PC.

This didn't affect my enjoyment of Death Stranding a year ago or Horizon Zero Dawn 3 years ago lol it won't change anything. It's all about consoles, I'm a console guy. I still consider Xbox One games to be exclusives

 

[The Alien]

Based off the math in this thread, XBox Infinite was indeed the better name as XBox is infinitely more powerful.

 

[Bryank75]

Jim is filing the paperwork as we speak!

Which issue of Wired will it appear in?

Is he using carrier pigeon or gyrocopter this time?

Remember there is a nasty looking guy with a mustache and a dog after the pigeon...saw them in an air balloon!

 

[Hobbygaming]

Who knew gamers put checkerboarding on their priority list of console features.

lol

Why not when there's things like DLSS out there that can look better than native resolutions

 

[Bryank75]

Based off the math in this thread, XBox Infinite was indeed the better name as XBox is infinitely more powerful.

 

[OptimistPrime]

Those were conservative numbers? What are you even basing those numbers on? Neither of these consoles are out, nor is an actual full RDNA2 GPU. Hell, we still don't know exactly what's in these chips just yet. It's absurd to just go, well one is 1.5 and the other is clearly 2.0, so math. Considering the long development cycles of these machines, both are probably hybrids, and by cutting the fat they gain a little.

I accumulated 3 performance uplifts: ML upscale (like DLSS) VRS, and RDNA2 IPC. The first two have been extensively looked at using RTX cards that have been out for almost 2 years. DLSS routinely gives a 50-100% performance improvement. Early looks at VRS (using a lesser version) show around 25% improvement.

The RDNA IPC improvement was a guess. AMD said RDNA1 is a 50% improvement over GCN, and RDNA2 is a 50% improvement over RDNA1, That second one is just a claim from AMD, but the first claim has been tested and is essentially accurate. I made a guess that if PS5 actually is a RDNA 1.5 GPU that the improvement would be about 25%. It might actually be better or much worse depending on how close it is to RDNA2 in IPC.

 

[hyperbertha]

PS5 fans still have plenty of reason for hope. He has no reason to make this slide unless the ps5 didn't get the flop benefits of RDNA2 :

Remember, in Lord Cerny you trust. He has never let you down. He is strong.

 

[SantaC]

This is a custom gpu, not something you see being sold as a PC graphic card.

RDNA2 isnt ready yet, so it makes sense that it has some mixed artchitecture there.

 

[VFXVeteran]

This didn't affect my enjoyment of Death Stranding a year ago or Horizon Zero Dawn 3 years ago lol it won't change anything. It's all about consoles, I'm a console guy. I still consider Xbox One games to be exclusives

I got ya man. I'm just saying that the PC versions of the PS exclusives will be able to keep pace with the XSX exclusives (of course the PC versions of XSX will be better than the XSX too). All in all, it will be a glorious time for PC owners that have the hardware to push all of these games to their limits to see what they really will look like. I'm a games guy. I want the best iteration of a game and that's why I pick the PC.

 

[Xplainin]

Why is she giving the old guy head?

Good on the old fella.

 

[Dave_at_Home]

I accumulated 3 performance uplifts: ML upscale (like DLSS) VRS, and RDNA2 IPC. The first two have been extensively looked at using RTX cards that have been out for almost 2 years. DLSS routinely gives a 50-100% performance improvement. Early looks at VRS (using a lesser version) show around 25% improvement.

The RDNA IPC improvement was a guess. AMD said RDNA1 is a 50% improvement over GCN, and RDNA2 is a 50% improvement over RDNA1, That second one is just a claim from AMD, but the first claim has been tested and is essentially accurate. I made a guess that if PS5 actually is a RDNA 1.5 GPU that the improvement would be about 25%. It might actually be better or much worse depending on how close it is to RDNA2 in IPC.

Again, all of this is incredibly speculative.

 

[OptimistPrime]

PS5 fans still have plenty of reason for hope. He has no reason to make this slide unless the ps5 didn't get the flop benefits of RDNA2 :

Remember, in Lord Cerny you trust. He has never let you down. He is strong.

RDNA 1 FLOPs would be about 50% improved over GCN.
RDNA 2 FLOPs would be about 125% improved over GCN (50% over RDNA1) according to an AMD slide

 

[NullZ3r0]

This is consistent with what Cerny said. He said:

• RDNA2 is a malleable framework not a de facto set of features.
• PS5's GPU would have some features in other RDNA2 chips and some features not in RDNA2.
• Sony would contribute some features that might show up in other chips.

Source:

 

[Xplainin]

PS5 doesnt have Machine Learning. ML is more than FP16. Way more.

 

[Hobbygaming]

I got ya man. I'm just saying that the PC versions of the PS exclusives will be able to keep pace with the XSX exclusives (of course the PC versions of XSX will be better than the XSX too). All in all, it will be a glorious time for PC owners that have the hardware to push all of these games to their limits to see what they really will look like. I'm a games guy. I want the best iteration of a game and that's why I pick the PC.

That's great for PC guys but the majority of us will still be on consoles and PC won't get every game on PS5 so given that I'm a console guy and I prefer PS I'll still be on PS5 to get every game released on the console

 

[alstrike]

Edit: My post was meaner than I like. Good luck to you!

Judging by the post I first quoted I'm not the one who needs it.

= 23.4TF or 130% performance difference over PS5

I'm not laughing any harder because I'm scared if I do my head would fall off...

 

[01011001]

PS5 fans still have plenty of reason for hope. He has no reason to make this slide unless the ps5 didn't get the flop benefits of RDNA2 :

Remember, in Lord Cerny you trust. He has never let you down. He is strong.

what people need to learn is to stop listening to marketing blurps and take everything a company says during a press conference with a massive grain of salt.

during the PS3 reveal Sony literally had a slide up saying the PS3 is a 2 teraflops machine.

 

[OptimistPrime]

Judging by the post I first quoted I'm not the one who needs it.

I'm not laughing any harder because I'm scared if I do my head would fall off...

I don’t mean literally 23TF, I’m talking about relative performance. If PS5 truely lacks VRS, good ML support, and RDNA2 IPC improvements, that would be a big deal.

I actually think it’s unlikely to be the case. I’m only saying that if it’s true it would be a big deal as those features add up to quite a lot.

 

[StreetsofBeige]

what people need to learn is to stop listening to marketing blurps and take everything a company says during a press conference with a massive grain of salt.

during the PS3 reveal Sony literally had a slide up saying the PS3 is a 2 teraflops machine.

Yup. This beauty.

Never knew PS3 had more TF than PS4.

 

[Xplainin]

Reminds me of when it became clear that PS5 was 10TF so people started saying SX MUST be only 10 TF and 6x2 wasn’t 12 and Phil was lying

Ah, the old 12tflop is GCN flops, not RDNA flops.
They stuck to that shit to the death.

 

[Grumpy Toe Nail]

For decades of experience console warring, I doubt we'll notice the difference. Microsoft would have made a bigger deal out of this if they could, or perhaps I missed the memo.

Microsoft isnt known to shit talk the competition like Sony does. I think its mostly because Microsoft joined the scene with Sony already ahead.
Since people be LoLing @ me, this was Sony after microsoft's E3. Their strategy was reactionary and toxic. X1X was when Microsoft grew some balls. There's a reason Sony is being quiet this time...LOL

 

[GamerEDM]

stop fucking panicking and rewatch The Road to PS5 video where Mark Cerny addressed all these concerns

 

[Bryank75]

I don’t mean literally 23TF, I’m talking about relative performance. If PS5 truely lacks VRS, good ML support, and RDNA2 IPC improvements, that would be a big deal.

I actually think it’s unlikely to be the case. I’m only saying that if it’s true it would be a big deal as those features add up to quite a lot.

Well actually X is the 24th letter...so it's probably 24 Tflops!

 

[Shakka43]

Wait, so the PS5 is gonna be weaker than Series X? Who knew.

 

[kyoji]

can someone link me to the tweet itself?

 

[Rikkori]

Sounds like every GPU tier for Ampere just went up $100

 

[Neo Blaster]

Damn, Cerny should take a look at some threads in this forum, he would definitely laugh his ass out on the sheer stupidity of Xbox Defense Force trying to downplay everything about PS5.

 

[Bryank75]

The XBox lads right now...

 

[farmerboy]

I got ya man. I'm just saying that the PC versions of the PS exclusives will be able to keep pace with the XSX exclusives (of course the PC versions of XSX will be better than the XSX too). All in all, it will be a glorious time for PC owners that have the hardware to push all of these games to their limits to see what they really will look like. I'm a games guy. I want the best iteration of a game and that's why I pick the PC.

And supposedly a grown man.

 

[DonJuanSchlong]

Regardless if it's RDNA 1/1.5/2/3, it will be a much appreciated upgrade. Bring on that SSD powah!

 

[Jtibh]

The way i feel about ps5 not having rdna2 is like how chubby women feel when they find out that ribbed condoms dont even taste like ribbs.

 

[RCU005]

I don't know why people always expect alien tech every generation shift, yet everyone complains about the specs of each console. With PS4, people complained that the specs were already outdated. Now, they are doing it again!

The sad thing about all this, it's that ~95% of people who buy the console, DON'T CARE AT ALL! So the remaining ~5% just waste their time complaining.

 

[Jtibh]

Jim is filing the paperwork as we speak!

Magical pink slip here i come hahaha

 

[Heimdall_Xtreme]

Rosario Leonardi, Principal Graphics Engineer at Sony Interactive Entertainment Europe, had this interesting tweet...

He also confirmed there is no machine learning...

In english or spanish please.

Ok brainy tell me slow down whats that mean.

 

[BattleScar]

Not this shit again.

 

[thicc_girls_are_teh_best]

This is consistent with what Cerny said. He said:

• RDNA2 is a malleable framework not a de facto set of features.
• PS5's GPU would have some features in other RDNA2 chips and some features not in RDNA2.
• Sony would contribute some features that might show up in other chips.

Source:

No, not really. Again, aside from a couple rather glaring things, that Twitter post is saying PS5 is custom RDNA2, which we knew from Day 1. We also knew Series X was custom RDNA2, also more or less from day one.

What's always been up for contention is what does that "custom" really entail, because we don't quite know what all the standard/generic RDNA2 features are. And even things we can try saying are standard/generic, after some looking it'd appear that could be down to a specific platform holder's partnership in bringing it to standard (such as Mesh Shaders likely being adapted at the behest of Microsoft).

The reason "custom" RDNA2 has always been a hot issue is because, at least regarding PS5, we have had Ariel and Oberon leaks that moreso fit the profile of RDNA1, particularly Ariel. 40 CU limit (same as RDNA1), mention of Primitive Shaders (same as RDNA1, albeit notably customized by Sony), mention of Geometry Engine (also RDNA1, though it's probably present in RDNA2 under the same name so that really doesn't tell us anything), etc.

Of course you can counter-balance that with confirmation of hardware-based RT, something RDNA1 definitely doesn't have, and even further with mention of cache scrubbers which aren't in RDNA1 or standard RDNA2, but could be present in RDNA3 PC GPUs as something Sony brought to the table, similar to Mesh Shaders in RDNA2 PC GPUs and Series GPUs brought forward by Microsoft. Seeing mention of no VRS by this graphics engineer doesn't really surprise me; speaking semantics VRS is MS's terminology for such a technique and they may have an implementation specific to them. Seeing though that Foveated Rendering is a basis of inspiration for VRS and FR was already in PS4 Pro for PSVR support, Sony might have an evolved implementation of FR in PS5 that's roughly similar to VRS, but with some differences.

(*Also regarding Oberon, its listing as RDNA1 can be rationalized as referring to an Ariel iGPU regression testing profile. Oberon was a major revision after all, and had at least two smaller revisions following it.*)

That's a bit similar to talk of if they have an equivalent of SFS; there isn't any evidence pointing to it, but the closest equivalent would be some evolution of PRT, which was already in the PS4. Worth stressing though is that PRT and SFS are not necessarily equivalents to one another as SFS seems like a very evolved offshoot that takes PRT as a foundational inspiration and went off from there. Even if there's a PRT "2.0" in PS5 there's no guarantee it's doing what SFS is doing, part of the reason being they may not need it when you look at their own focus of the SSD I/O and how it's prioritizing big bandwidth to try moving data in and out of RAM as quickly as it can, regardless of the type, up to a reasonable compression level. SFS is a much more discriminate approach that probably requires a bit more familiarity by devs to get a handle on, but mastery of it should yield very effective use of the Series' SSD I/O and memory systems to stay closer to PS5's SSD I/O despite what the paper specs show.

Really I think the more alarming thing here is no mention of ML being present, as a few other posters have mentioned. Not even having FP16 is a surprise. Another poster kept mentioning about there being RDNA1 cards with ML on them, but I think those were niche cards, not the mainline, mainstream offerings. It's plausible the Sony graphics engineer was referring to the main cards in what they said, which would be true, since for RDNA2 AMD've spun off compute-focused designs to CDNA2 cards, they mention this in their shareholders call back in...early March I think it was?

I can only imagine things like ML missing because Sony needed to prioritize importance of features vs. silicon budget and maybe they felt making room for things like cache scrubbers were more important and would benefit their design over having ML support that maybe could be done anyway (albeit less efficiently) in software on the GPU. Who knows. I just find omissions like that extremely surprising and it does bring into question about what other possible RDNA2 features could be missing or, present yet not as a big focus in the design of PS5.

So, uh, hopefully Cerny or someone else high up on the engineering team start doing a couple of interviews that can specifically clear the air on this. If it's "bad" news, then they need to try spinning discourse to something else or find a more clever way to dispel growing doubts, like with a visually arresting gameplay demo, real-time, played by a real person.

Or something (granted, I do just kinda want actual gameplay footage for next-gen now, even beyond Microsoft's show because after the 23rd, who knows when we'll get more from either).

 

[kyoji]

so we all just going to pretend like this isn't clearly photoshopped? and fake? why isn't this thread locked yet?

 

[MasterCornholio]

Not this shit again.

Not this shit again.

Agreed we are practically going around in circles at this point.

 

[MasterCornholio]

so we all just going to pretend like this isn't clearly photoshopped? and fake? why isn't this thread locked yet?

Has that been proven yet?

At the moment it's just a theory. I haven't seen any proof that it's fake yet. But then again I haven't seen that engineer confirm the information on Twitter yet.

 

[Grinchy]

This must be why Far Cry 6 is 4K on XSX and 240p on PS5.

 

[Dave_at_Home]

No, not really. Again, aside from a couple rather glaring things, that Twitter post is saying PS5 is custom RDNA2, which we knew from Day 1. We also knew Series X was custom RDNA2, also more or less from day one.

What's always been up for contention is what does that "custom" really entail, because we don't quite know what all the standard/generic RDNA2 features are. And even things we can try saying are standard/generic, after some looking it'd appear that could be down to a specific platform holder's partnership in bringing it to standard (such as Mesh Shaders likely being adapted at the behest of Microsoft).

The reason "custom" RDNA2 has always been a hot issue is because, at least regarding PS5, we have had Ariel and Oberon leaks that moreso fit the profile of RDNA1, particularly Ariel. 40 CU limit (same as RDNA1), mention of Primitive Shaders (same as RDNA1, albeit notably customized by Sony), mention of Geometry Engine (also RDNA1, though it's probably present in RDNA2 under the same name so that really doesn't tell us anything), etc.

Of course you can counter-balance that with confirmation of hardware-based RT, something RDNA1 definitely doesn't have, and even further with mention of cache scrubbers which aren't in RDNA1 or standard RDNA2, but could be present in RDNA3 PC GPUs as something Sony brought to the table, similar to Mesh Shaders in RDNA2 PC GPUs and Series GPUs brought forward by Microsoft. Seeing mention of no VRS by this graphics engineer doesn't really surprise me; speaking semantics VRS is MS's terminology for such a technique and they may have an implementation specific to them. Seeing though that Foveated Rendering is a basis of inspiration for VRS and FR was already in PS4 Pro for PSVR support, Sony might have an evolved implementation of FR in PS5 that's roughly similar to VRS, but with some differences.

(*Also regarding Oberon, its listing as RDNA1 can be rationalized as referring to an Ariel iGPU regression testing profile. Oberon was a major revision after all, and had at least two smaller revisions following it.*)

That's a bit similar to talk of if they have an equivalent of SFS; there isn't any evidence pointing to it, but the closest equivalent would be some evolution of PRT, which was already in the PS4. Worth stressing though is that PRT and SFS are not necessarily equivalents to one another as SFS seems like a very evolved offshoot that takes PRT as a foundational inspiration and went off from there. Even if there's a PRT "2.0" in PS5 there's no guarantee it's doing what SFS is doing, part of the reason being they may not need it when you look at their own focus of the SSD I/O and how it's prioritizing big bandwidth to try moving data in and out of RAM as quickly as it can, regardless of the type, up to a reasonable compression level. SFS is a much more discriminate approach that probably requires a bit more familiarity by devs to get a handle on, but mastery of it should yield very effective use of the Series' SSD I/O and memory systems to stay closer to PS5's SSD I/O despite what the paper specs show.

Really I think the more alarming thing here is no mention of ML being present, as a few other posters have mentioned. Not even having FP16 is a surprise. Another poster kept mentioning about there being RDNA1 cards with ML on them, but I think those were niche cards, not the mainline, mainstream offerings. It's plausible the Sony graphics engineer was referring to the main cards in what they said, which would be true, since for RDNA2 AMD've spun off compute-focused designs to CDNA2 cards, they mention this in their shareholders call back in...early March I think it was?

I can only imagine things like ML missing because Sony needed to prioritize importance of features vs. silicon budget and maybe they felt making room for things like cache scrubbers were more important and would benefit their design over having ML support that maybe could be done anyway (albeit less efficiently) in software on the GPU. Who knows. I just find omissions like that extremely surprising and it does bring into question about what other possible RDNA2 features could be missing or, present yet not as a big focus in the design of PS5.

So, uh, hopefully Cerny or someone else high up on the engineering team start doing a couple of interviews that can specifically clear the air on this. If it's "bad" news, then they need to try spinning discourse to something else or find a more clever way to dispel growing doubts, like with a visually arresting gameplay demo, real-time, played by a real person.

Or something (granted, I do just kinda want actual gameplay footage for next-gen now, even beyond Microsoft's show because after the 23rd, who knows when we'll get more from either).

It's possible ML just doesn't bring the performance people think it does? I dunno. Maybe Sony has some sort of custom implementation? Probably not, but if there's any custom hardware we don't know about, I bet it's VR related.

 

[Tripolygon]

what people need to learn is to stop listening to marketing blurps and take everything a company says during a press conference with a massive grain of salt.

during the PS3 reveal Sony literally had a slide up saying the PS3 is a 2 teraflops machine.

Its almost as if they were responding to Microsoft fuzzy math when it came to Xbox 360 performance.

Xbox 360 Ushers in the Future of Games and Entertainment - Stories

Highly anticipated video game console is unveiled on an MTV special.

news.microsoft.com

Luckily that all stopped and actual performance was revealed. I find Microsoft and Sony have been honest with their systems performance giving actual real numbers without fudging this gen and next gen. You can calculate their numbers and arrive at the same result given published data out there. So we can actually take marketing blurps at face value if they are actual real numbers. Leave any qualifiers out.

 

[kyoji]

Has that been proven yet?

At the moment it's just a theory. I haven't seen any proof that it's fake yet. But then again I haven't seen that engineer confirm the information on Twitter yet.

the guy willydearborn who is listed in one of the photos as one of the people who the ps5 principle guy is responding too has never even tweeted at him before. https://twitter.com/willydearborn/with_replies

If you know you know.

 

[MasterCornholio]

the guy willydearborn who is listed in one of the photos as one of the people who the ps5 principle guy is responding too has never even tweeted at him before. https://twitter.com/willydearborn/with_replies

If you know you know.

Can you build a solid case around that theory? It would be useful because the last thing gaf needs is FUD.

 

[kyoji]

Can you build a solid case around that theory? It would be useful because the last thing gaf needs is FUD.

look at the images in the original post then click that link i put in my last post, the guy has never even had a conversation with imal3 on twitter. so how is he in the picture posted in the original post? twitter does not lie

 

[Hosam6xavi]

Maybe it's a fud

No wonder all they have been talking about is the SSD

 

[saintjules]

This is consistent with what Cerny said. He said:

• RDNA2 is a malleable framework not a de facto set of features.
• PS5's GPU would have some features in other RDNA2 chips and some features not in RDNA2.
• Sony would contribute some features that might show up in other chips.

Source:

I bet you there are people on here with fingers in their ears, lalalalala.

 

[kyoji]

https://twitter.com/willydearborn/with_replies

 

[TigerKnee]

I bet a large number of people here making an argument don't even know what the hell RDNA 1 and 2 does!!

(admittedly I don't which is why I am not saying anything!!)