Support NeoGAF

[asdasdasdbb]

How are people arriving at 30 tflops?

A RDNA3 CU can (in theory) execute 2 FP32 operations where RDNA2 can only do one.

It doesn't work quite right but maybe they can fix it as part of RDNA4.

 

[SHA]

Legendary thread.
We getting 8k 30 fps games!

It's a creative decision, making a120+ fps game on pc running at 30 fps is a creative decision.

 

[SHA]

RTX 7090 wont do that and they use amd so +3 extra years.
Not as funny as "Switch 4K60" with 25w of power and 600+ppi display though.

Either this is BS or i am a clown and they are actually making a 2500W console. Who would they sell those to? power plants?

Don't go hard with switch, cartoonish games gonna look more artistic, it's Nintendo.

 

[James Sawyer Ford]

Console generations were always the chance to do major deviations, they were not slapped together parts designed for a market that offers minor updates / constant iteration.

Look at PS4 Pro vs PS5, look at PS2 vs PS3, look at Xbox One and Xbox Series X|S.

Some features will be reused, many will be new as that is the entire point of the new generation reset: just enough backwards compatibility support to carry the digital library forward (plus some software enhancements you can layer on top), but the focus is to enable new dreams / new horizons (and some new puzzles) in a compact box that developers can optimise for years (that is how it punches above their weight, but being designed/customised at the architecture and components level for the problem at hand).

What major breakthrough created for PS4 Pro enabled PS5?

I am sorry, but I still see the main beneficiary of Pro consoles being the console manufacturer that manages to fill a gap now that console generations have gotten longer… cross generation support has grown and devs are spending (less incentive too) less and less time working on each box…

We still have console generations. PS6 is likely coming out in 2028, which represents a chance for a more significant refresh. You claim it's potential is minimized by a PS5 Pro, I claim that's absolutely hogwash.

Plenty of Polaris features ultimately ended up in PS5, so it's not "wasted" effort. These companies aren't re-inventing the wheel when they create new GPUs. They take what's existing, and re-organize them based on a better understanding of the node capabilities available, and receive feedback from software teams on where bottlenecks are.

The main benefit of Pro consoles IS to extend the 8 year lifecycle, giving consumers and enthusiasts more choice so they feel get improvements sooner. 8 years is too long of a time to wait for better hardware.

 

[DenchDeckard]

Doubling would have made some sense. But they don't have to. The PS5 Pro may end up being not that big of an improvement.

I think it will be a good jump over ps5. It has to.

I think people are going to be disappointed when they hear TFs for this machine as far as pure brute force power goes

I'm hoping they can get to at least 18 on paper. I think they can sell that.

Disagree. And Cerny didn't say more CUs are not the answer, he simply said it makes sense for them to use faster CUs than just use more CUs. And him saying that wouldn't mean that they would never increase the CU count. And in truth, if all they have done is increase it to 60CUs...just 8 more than the XSX, after3+ years, that means that his original stance still holds true. Focusing on faster CUs (assuming those 60CUs clock at over 2.3Ghz) as opposed to just having more of them.

And also... NO. I never felt TFs were the best way to measure performance and I still don't And I didn't need Cerny to say the same thing. I start thinking this was after the 2080 GPU. I mean, Sony could release a PS5pro with a 12TF RDNA 3+ GPU, 96MB of infinity cache and let's say for argument's sake 60 Ml core/units, and that system would run circles around the XSX. Even while having the same `TF` number.

At the same time, no one is saying TFs do not matter. What a lot of people are saying though, is that more than ever, especially when the TF number between the two consoles is not so vastly different... is that it's not `all` that matters. Eg... if the XSX was a 15TF console to the PS5s 10TF? There wouldn't be a single game out here where the PS5 performs better than the XSX. No matter how much better the tools are or whatever the inda IO the PS5 has.


I don't see that happening. Even if that GPU is clocked at 2200mhz (less than the clocks in the PS5, it's still going to be a 16.8TF GPU. If at 2300Mhz, that's a 17.6TF GPU, at 2400Mhz, that's a 18.4TF GPU.

Personally, I see it landing anywhere between a 17 - 18TF GPU. Just on the optics alone. It would kinda be stupid to release a Pro console whose TF number (which unfortunately is what everyone would see first anyways) is just 1-3TF more than the XSX. On the tech side of things though, I can see, and there is proof to back this, that on a 5nm node, they can easily hit clocks of over 2300Mhz.

Yeah 18tf would be a good shout I think. That should give it a considerable jump over the ps5 and be a nice jump over xbox. Then I think they will go for like 549 to 599 in the UK.

Day one for me.

 

[Insane Metal]

I bet this will use Sony's solution for RT somehow. There's no way there's a 'big bump' in RT performance with AMD hardware.

 

[DeepEnigma]

Disagree. And Cerny didn't say more CUs are not the answer, he simply said it makes sense for them to use faster CUs than just use more CUs. And him saying that wouldn't mean that they would never increase the CU count. And in truth, if all they have done is increase it to 60CUs...just 8 more than the XSX, after3+ years, that means that his original stance still holds true. Focusing on faster CUs (assuming those 60CUs clock at over 2.3Ghz) as opposed to just having more of them.

And also... NO. I never felt TFs were the best way to measure performance and I still don't And I didn't need Cerny to say the same thing. I start thinking this was after the 2080 GPU. I mean, Sony could release a PS5pro with a 12TF RDNA 3+ GPU, 96MB of infinity cache and let's say for argument's sake 60 Ml core/units, and that system would run circles around the XSX. Even while having the same `TF` number.

At the same time, no one is saying TFs do not matter. What a lot of people are saying though, is that more than ever, especially when the TF number between the two consoles is not so vastly different... is that it's not `all` that matters. Eg... if the XSX was a 15TF console to the PS5s 10TF? There wouldn't be a single game out here where the PS5 performs better than the XSX. No matter how much better the tools are or whatever the inda IO the PS5 has.


I don't see that happening. Even if that GPU is clocked at 2200mhz (less than the clocks in the PS5, it's still going to be a 16.8TF GPU. If at 2300Mhz, that's a 17.6TF GPU, at 2400Mhz, that's a 18.4TF GPU.

Personally, I see it landing anywhere between a 17 - 18TF GPU. Just on the optics alone. It would kinda be stupid to release a Pro console whose TF number (which unfortunately is what everyone would see first anyways) is just 1-3TF more than the XSX. On the tech side of things though, I can see, and there is proof to back this, that on a 5nm node, they can easily hit clocks of over 2300Mhz.

Keep in mind, the clocks will have to be at bare minimum of what the PS5 is now, to not break compatibility within the profiles. I don't see them as being less for this reason, but who knows what they have cooked up with modifications.

 

[THE:MILKMAN]

I don't see that happening. Even if that GPU is clocked at 2200mhz (less than the clocks in the PS5, it's still going to be a 16.8TF GPU. If at 2300Mhz, that's a 17.6TF GPU, at 2400Mhz, that's a 18.4TF GPU.

Personally, I see it landing anywhere between a 17 - 18TF GPU. Just on the optics alone. It would kinda be stupid to release a Pro console whose TF number (which unfortunately is what everyone would see first anyways) is just 1-3TF more than the XSX. On the tech side of things though, I can see, and there is proof to back this, that on a 5nm node, they can easily hit clocks of over 2300Mhz.

Fair points but how does it look factoring in the decreased PPA gains of 7nm > 5nm versus PS4's 28nm > PS4 Pro's 16nm? Is it significant?

 

[HeisenbergFX4]

I think it will be a good jump over ps5. It has to.



I'm hoping they can get to at least 18 on paper. I think they can sell that.



Yeah 18tf would be a good shout I think. That should give it a considerable jump over the ps5 and be a nice jump over xbox. Then I think they will go for like 549 to 599 in the UK.

Day one for me.

Just being totally honest here the talk about the Pro started last year as from what I was told Sony really wanted to launch it along side of the PSVR2 and at the time I was told it was in the 15-18TF range but it was put on hold because of cost all of which I think I mentioned here so not saying anything I haven't said before.

I do get points taken away because I also said around the same time (April ish 2022) that the Series S would be the first console to get a refresh as when I was told that I thought it meant more power and not just double the SSD size so there's that

 

[Mr.Phoenix]

I think it will be a good jump over ps5. It has to.



I'm hoping they can get to at least 18 on paper. I think they can sell that.



Yeah 18tf would be a good shout I think. That should give it a considerable jump over the ps5 and be a nice jump over xbox. Then I think they will go for like 549 to 599 in the UK.

Day one for me.

I don't even think wud be priced that much differently than a PS5 physical edition at launch. Well, technically it would be priced higher because I expect it to come in at the price of the PS5PE but without the disc drive which would be bought separately.

A 5nm chip coming in at around 300-320mm2 in size makes for a lot of APU. And not one that would be that much more expensive than the PS5 APU back in 2020. Also, when considering that they would keep a lot more or less the same... eg. same SSD size, the same amount of RAM (albeit faster RAM), same CPU just clocked higher and with slightly more cache... they are perfectly poised to make a $499 console without a disc drive

 

[Moridianae]

There's always a market for high-end. My late old man was big into audio, so he had all kinds of high-end speakers and devices to play his classical music growing up.

I inherited those speakers after he passed and they still go for a shitton of money despite their age.

Not relevant to PS5 Pro aside from there being enthusiasts. Lol--I remember he got so jealous of me when I had a PS3 at launch and introduced him to Blu Ray. He wanted one so badly.

 

[DenchDeckard]

Just being totally honest here the talk about the Pro started last year as from what I was told Sony really wanted to launch it along side of the PSVR2 and at the time I was told it was in the 15-18TF range but it was put on hold because of cost all of which I think I mentioned here so not saying anything I haven't said before.

I do get points taken away because I also said around the same time (April ish 2022) that the Series S would be the first console to get a refresh as when I was told that I thought it meant more power and not just double the SSD size so there's that

Ahhh ok, wouldn't launching it alongside the psvr have been a little early?

Well fingers crossed it ends up being a decent jump

I don't even think wud be priced that much differently than a PS5 physical edition at launch. Well, technically it would be priced higher because I expect it to come in at the price of the PS5PE but without the disc drive which would be bought separately.

A 5nm chip coming in at around 300-320mm2 in size makes for a lot of APU. And not one that would be that much more expensive than the PS5 APU back in 2020. Also, when considering that they would keep a lot more or less the same... eg. same SSD size, the same amount of RAM (albeit faster RAM), same CPU just clocked higher and with slightly more cache... they are perfectly poised to make a $499 console without a disc drive

Ahh yeah, maybe 499 without a disk drive. I deffo want the disk drive so I hope it's readily available.

 

[Mr.Phoenix]

Fair points but how does it look factoring in the decreased PPA gains of 7nm > 5nm versus PS4's 28nm > PS4 Pro's 16nm? Is it significant?

Not really. Take for instance, as it stands, the current PS5s on the market are on a 6nm process. That already lets them drop the APU size from 320mm2 it launched at to around 260mm2 currently.My numbers may be a little off, can't remember the specifics.

When you look at all that, and consider a PS5pro would at least be on a 5nm node, you can start to see why they are only going for a 32WG (64CU) GPU. Yes, thats what this GPU really is before they start disabling WGs/CUs.

Then considering that the CPU will pretty much be untouched, and even if the individual CUs are bigger in RDNA 3+ compared to RDNA2, its nothing that would be relevant considering the node shrink. All that would fit into a 300-340mm2 chip easily. Only thing that really even stands to make that chip big (340mm2+) is if they opt to add more cache and how much of it they opt to add.

 

[DrKeo]

To be honest, this leak is probably fake or at least misguided.

 

[PSYGN]

I want to see games that really push the PS5 before forking out more. Third party will still have to make things run on base PS5 and XSS so it seems like a waste for 95% of games.

Edit: Wow, a lot in the poll voted for "For sure". Scary.

 

[StereoVsn]

I'd guess it will use Zen 4c or 5c. Smaller footprint and better power consumption at lower clocks.
The console is always going to be GPU limited, there's no need for a big CPU boost.

Sure there is a need. Look at Starfield, most likely some of the reasons for 30fps limit is CPU not being powerful enough.

 

[shamoomoo]

54 CUs active is more likely.
One WGP disabled for each Shader Engine.

Why would Sony disable 6 compute units if the total is 60? Both the PS5 and the series consoles have 4 out of their total CUs disabled and 2 in the previous gen.

 

[Tripolygon]

I bet this will use Sony's solution for RT somehow. There's no way there's a 'big bump' in RT performance with AMD hardware.

This is AMD engineers doing the work on building a custom SoC for Sony. Whatever idea Sony engineers have if it is good enough or aligns with AMDs strategy, you'll see it in future AMD products. Moreover, AMD is constantly evolving their products, just like Nvidia, they are just using a different approach and dedicating less silicon to it. RDNA 3 is already a big bump compared to RDNA 2 in RT.

Raytracing on AMD’s RDNA 2/3, and Nvidia’s Turing and Pascal

Note: Jake has commented that Nvidia’s tools may not show the true BVH structure.

chipsandcheese.com

 

[StereoVsn]

Ahhh ok, wouldn't launching it alongside the psvr have been a little early?

Well fingers crossed it ends up being a decent jump

Ahh yeah, maybe 499 without a disk drive. I deffo want the disk drive so I hope it's readily available.

Unless Sony are crazy, it will be the same optional disk drive as the one Slim will use. There is no need of having 2 SKUs.

My personal opinion is $599 Pro in Q4 2024, Slim will be $399 or so giving enough differentiation.

As far as power, who the heck knows other then Cerny, lol. Guessing ~18TF, but we shall see.

 

[THE:MILKMAN]

Not really. Take for instance, as it stands, the current PS5s on the market are on a 6nm process. That already lets them drop the APU size from 320mm2 it launched at to around 260mm2 currently.My numbers may be a little off, can't remember the specifics.

When you look at all that, and consider a PS5pro would at least be on a 5nm node, you can start to see why they are only going for a 32WG (64CU) GPU. Yes, thats what this GPU really is before they start disabling WGs/CUs.

Then considering that the CPU will pretty much be untouched, and even if the individual CUs are bigger in RDNA 3+ compared to RDNA2, its nothing that would be relevant considering the node shrink. All that would fit into a 300-340mm2 chip easily. Only thing that really even stands to make that chip big (340mm2+) is if they opt to add more cache and how much of it they opt to add.

Honestly, I'm not convinced...yet. From what I can find online (Anandtech) the PPA gains of 7nm > 5nm are much reduced compared to previous node improvements: 15% frequency gain or 20% power reduction and 45% area reduction.

Interesting info from Heisenberg above. Could indicate a delay (to launch on 3nm?) or even that it is scrapped altogether?

 

[ToTTenTranz]

Sure there is a need. Look at Starfield, most likely some of the reasons for 30fps limit is CPU not being powerful enough.

Starfield's 30FPS are definitely not there because of the CPU.

 

[PeteBull]

I want to see games that really push the PS5 before forking out more. Third party will still have to make things run on base PS5 and XSS so it seems like a waste for 95% of games.

Edit: Wow, a lot in the poll voted for "For sure". Scary.

I encourage u to check DF vids and see resolution/performance of many games that push graphics, u will notice we need ps5pr0 asap, we will get it likely 1,5years from now so it will be extremly needed .

 

[Mr.Phoenix]

Honestly, I'm not convinced...yet. From what I can find online (Anandtech) the PPA gains of 7nm > 5nm are much reduced compared to previous node improvements: 15% frequency gain or 20% power reduction and 45% area reduction.

Interesting info from Heisenberg above. Could indicate a delay (to launch on 3nm?) or even that it is scrapped altogether?

The facts are out here though. it's a fact that the current PS5 APU is only around 260mm2. Compared to the arod 320mm2 it launched at. And hats going from 7nm to 6nm.

The PS5pro is going to be on at worst a 5nm process.

Simple deduction would indicate that on a 5nm node, the og PS5 chip would be no bigger than say, 220mm2? So what do you think they could fit in a 320-340mm2 5nm chip? 24 more CUs? 4x the cache?

And this further supports the theory that they aren't even going to be doing anything major with their clocks. So GPU goes from 2.23Ghz to say, 2.35Ghz. CPU goes from 3.6 GHz to 4Ghz? Or 4.2Ghz?

I can't speak to it coming at all, dont care either way, I can see why they would do the same thing they did with the PS4 though, doesn't take anything out of them having an entry and pro grade PS5 on the market.

 

[PeteBull]

Starfield's 30FPS are definitely not there because of the CPU.

Starfield on xsx/xss is 30fps target(might dip below even, its bethesda after all, lets not forget that) coz of cpu, and its 1296p on xsx coz of its gpu, if it had much stronger gpu but same cpu it could be 4k30, which obviously isnt the case, if it had much stronger cpu it would have performance mode with 60fps but much lower resolution.

Both xsx and ps5 arent some infinite power monsters, they got highly downclocked/downvolted r7 3700x as cpu, with less cache, and 330$ gpu so rx 6700xt is stronger from both in any scenario, and just look how much stronger top pc gpus are from that 6700xt https://www.techpowerup.com/gpu-specs/radeon-rx-6700-xt.c3695

Now fastforward 1,5years and tell me- once nvidia launches their 50xx series and amd launches their rdna4 archi gpu's. wont console only players need/want much stronger option vs current ps5/xsx?

 

[Physiognomonics]

I think people are going to be disappointed when they hear TFs for this machine as far as pure brute force power goes

Not if they use RDNA3 Tflops. Anyways what's going to be interesting is how high they are going to clock that thing. If they reach 3ghz then we'll get double PS5: +20tflops using RDNA2 tflops (and much more using RDNA3 tflops). I think they'll reach 3ghz as one of the leak was talking about 2x more raster.

At those clocks this thing is going to be made for higher framerates and it's will very likely be a very balanced hardware like PS5 is.

 

[DuduIsDisappointed]

Ps5 pro is around 22TF.

Won't share any more info

 

[Investor9872]

Personally, I see it landing anywhere between a 17 - 18TF GPU. Just on the optics alone. It would kinda be stupid to release a Pro console whose TF number (which unfortunately is what everyone would see first anyways) is just 1-3TF more than the XSX. On the tech side of things though, I can see, and there is proof to back this, that on a 5nm node, they can easily hit clocks of over 2300Mhz.

I think all of us here who are interested in the PS5 Pro genuinely hope that it will have at least 17TF - 18TF GPU. But I wouldn't be surprised at all if the Pro will be nearer to 11-13TF, since Cerny has already filed a patent which seems to prioritize for an increase in Ray-tracing over Teraflops.

Now I would love it if Cerny puts out a high-performance specs PS5 Pro for around $1,000 (and I'll be there day-1), but that's just a pipe-dream. It will be around $550 to $600 max if Sony wants to have the same sales numbers as the PS4 Pro.

 

[Imtjnotu]

Ps5 pro is around 22TF.

Won't share any more info

Dads cousins girlfriend shared the info?

 

[Investor9872]

Dads cousins girlfriend shared the info?

And did they wink wink at him as well?

 

[SlimySnake]

Disagree. And Cerny didn't say more CUs are not the answer, he simply said it makes sense for them to use faster CUs than just use more CUs. And him saying that wouldn't mean that they would never increase the CU count. And in truth, if all they have done is increase it to 60CUs...just 8 more than the XSX, after3+ years, that means that his original stance still holds true. Focusing on faster CUs (assuming those 60CUs clock at over 2.3Ghz) as opposed to just having more of them.

Does it still hold true when the XSX is beating the PS5 in so many games? Despite having the same power draw? higher clocks mean way higher power draw. wide and slow wouldve got them the same performance, if not better. yes, they saved on the price of the die, but it didnt trickle down to the customer. everyone paid the same $500 for both consoles.

And also... NO. I never felt TFs were the best way to measure performance and I still don't And I didn't need Cerny to say the same thing. I start thinking this was after the 2080 GPU.

I dont understand what the 2080 did that convinced you that tflops werent the answer. if anything the 2080 performed just like the 1080 ti in standard rasterized games and was roundly criticized for its non RT performance when it came out. It also performed 35% worse than the 2080 Ti which had around 40% more tflops.

I am not sure how ANYONE on this board could look at the PS4 and X1 tflops/performance gap then look at the PS4 and PS4 Pro tflops/performance gap and finally the PS4 Pro and X1X performance gap literally translating 1:1 and say hey I never felt TFs were the best way to measure performance. Never? really?

The fact that we still do it on PCs should tell you everything. We arent even talking about the XSX here where performance comparisons are skewed at times due to DX12 fuck ups that we see on PCs as well as XS consoles. We are talking about Tflops of the PS5 Pro compared to the PS5's. If you want 2x more performance than the PS5, you need 2x more tflops. Real tflops, not the fake RDNA3.0 tflops. Plain and simple. Really doesnt matter how you get there. higher clocks or more CUs. Computing power is what matters most.

Now if cerny fucks up and bottlenecks himself like he did with the PS4 Pro's memory bandwidth or forgets to upgrade the vram like nvidia does with their midrange GPUs then thats a different discussion.

 

[JerryinSoCal]

Does it still hold true when the XSX is beating the PS5 in so many games? Despite having the same power draw? higher clocks mean way higher power draw. wide and slow wouldve got them the same performance, if not better. yes, they saved on the price of the die, but it didnt trickle down to the customer. everyone paid the same $500 for both consoles.

I dont understand what the 2080 did that convinced you that tflops werent the answer. if anything the 2080 performed just like the 1080 ti in standard rasterized games and was roundly criticized for its non RT performance when it came out. It also performed 35% worse than the 2080 Ti which had around 40% more tflops.

I am not sure how ANYONE on this board could look at the PS4 and X1 tflops/performance gap then look at the PS4 and PS4 Pro tflops/performance gap and finally the PS4 Pro and X1X performance gap literally translating 1:1 and say hey I never felt TFs were the best way to measure performance. Never? really?

The fact that we still do it on PCs should tell you everything. We arent even talking about the XSX here where performance comparisons are skewed at times due to DX12 fuck ups that we see on PCs as well as XS consoles. We are talking about Tflops of the PS5 Pro compared to the PS5's. If you want 2x more performance than the PS5, you need 2x more tflops. Real tflops, not the fake RDNA3.0 tflops. Plain and simple. Really doesnt matter how you get there. higher clocks or more CUs. Computing power is what matters most.

Now if cerny fucks up and bottlenecks himself like he did with the PS4 Pro's memory bandwidth or forgets to upgrade the vram like nvidia does with their midrange GPUs then thats a different discussion.

During last gen a teraflop or two difference was a high percentage of the overall power, that hasn't scaled with the current gen. If one was a 4.2TF machine and the other 6tf (pro/one x) that was basically a 30% increase in performance and that's not counting the extra memory it had. The current gen the TF difference is roughly the same base amount but the percentage is a lot smaller, also the PS5 regularly beats Series X in head to heads when it comes to overall performance and in some cases graphical settings as well. The only time you see the Series X really win out in anything it's usually a slightly higher native resolution but in most cases when that happens the framerate isn't as stable as it is on PS5. If they lowered the resolution they'd probably end up with the same performance overall which shows both machines are basically equal even though one has a lower TF count than the other. The PS5 also does some things better than Series X unlike the PS4 Pro and One X where the Pro was outclassed in every way.

BTW You get the PS5 digital for $100 less than the Series X, for people who only buy digital games these days that's a no brainer.

 

[Panajev2001a]

We still have console generations. PS6 is likely coming out in 2028, which represents a chance for a more significant refresh. You claim its potential is minimized by a PS5 Pro, I claim that's absolutely hogwash.

We will see I guess, so far thinned are still following this pattern. It is a big investment that could be focused on games, tools, and R&D for PS6. This is getting diverted and helping cross generation periods.

Plenty of Polaris features ultimately ended up in PS5, so it's not "wasted" effort. These companies aren't re-inventing the wheel when they create new GPUs. They take what's existing, and re-organize them based on a better understanding of the node capabilities available, and receive feedback from software teams on where bottlenecks are.

Right now, they need more days in the base PS5 rather than launching a PS5 Pro, AMD still ships on PC so it is not like Sony needs the PS5 Pro to make the PS6. They make it because some people will buy it, period.

The main benefit of Pro consoles IS to extend the 8 year lifecycle, giving consumers and enthusiasts more choice so they feel get improvements sooner. 8 years is too long of a time to wait for better hardware.

Do you think 3-4 years is enough to give you a meaningful update unless you raise the price and size/power envelope of the console by quite a bit (and then once you plane a $399-450 PS6 you have some problem to differentiate).

Chances are that PS5 Pro will be less of a jump from PS5 than PS4 to PS4 Pro and PS6 to PS6 Pro will be even smaller.

I feel this calls for consoles to keep changing to accommodate the frequent refreshes we have on PC or phones nowadays, but this is not the way tech is changing.

 

[asdasdasdbb]

To be honest, this leak is probably fake or at least misguided.

It's probably just 18 Gbps memory.

Unless Sony are crazy, it will be the same optional disk drive as the one Slim will use. There is no need of having 2 SKUs.

My personal opinion is $599 Pro in Q4 2024, Slim will be $399 or so giving enough differentiation.

As far as power, who the heck knows other then Cerny, lol. Guessing ~18TF, but we shall see.

Do you think Sony would avoid $599 because of the memes?

 

[shamoomoo]

Does it still hold true when the XSX is beating the PS5 in so many games? Despite having the same power draw? higher clocks mean way higher power draw. wide and slow wouldve got them the same performance, if not better. yes, they saved on the price of the die, but it didnt trickle down to the customer. everyone paid the same $500 for both consoles.

I dont understand what the 2080 did that convinced you that tflops werent the answer. if anything the 2080 performed just like the 1080 ti in standard rasterized games and was roundly criticized for its non RT performance when it came out. It also performed 35% worse than the 2080 Ti which had around 40% more tflops.

I am not sure how ANYONE on this board could look at the PS4 and X1 tflops/performance gap then look at the PS4 and PS4 Pro tflops/performance gap and finally the PS4 Pro and X1X performance gap literally translating 1:1 and say hey I never felt TFs were the best way to measure performance. Never? really?

The fact that we still do it on PCs should tell you everything. We arent even talking about the XSX here where performance comparisons are skewed at times due to DX12 fuck ups that we see on PCs as well as XS consoles. We are talking about Tflops of the PS5 Pro compared to the PS5's. If you want 2x more performance than the PS5, you need 2x more tflops. Real tflops, not the fake RDNA3.0 tflops. Plain and simple. Really doesnt matter how you get there. higher clocks or more CUs. Computing power is what matters most.

Now if cerny fucks up and bottlenecks himself like he did with the PS4 Pro's memory bandwidth or forgets to upgrade the vram like nvidia does with their midrange GPUs then thats a different discussion.

But RDNA3 TFLOPS,even at the PS5 clock speed is already 3x the RDNA2 TFLOPS 30+ vs 10.

The PS5 pro is a close platform, maybe RDNA3 TFLOPS will be close to real there.

 

[Mr.Phoenix]

Does it still hold true when the XSX is beating the PS5 in so many games? Despite having the same power draw? higher clocks mean way higher power draw. wide and slow wouldve got them the same performance, if not better. yes, they saved on the price of the die, but it didnt trickle down to the customer. everyone paid the same $500 for both consoles.

It allowed Sony to offer up 10TF console for $399. Let not forget that version of h PS5 exists. I would know being thats the one I got. AndIwoudgo as far as say that f the 38M PS5s sod, at east 6Mof those are PS5 digitals. lastly, you would expect the XSX to beat the PS5, but that sony was able to match it in most cases (at least far better than how the XB1 matched the PS4) speaks volumes of their approach. Wide and slow would have not let them do a PS5D.

I dont understand what the 2080 did that convinced you that tflops werent the answer. if anything the 2080 performed just like the 1080 ti in standard rasterized games and was roundly criticized for its non RT performance when it came out. It also performed 35% worse than the 2080 Ti which had around 40% more tflops.

Exactly, the 2080 was basically a 1080ti. And I am not specifically speaking to the 2080... but more to what Nvidia did then. They dint just double down on raster performance AMD. They pushed things ie RT and dlss... things that at the time were scoffed at, but now define the tech of the industry. Very good case for it not being all about TFs wouldn't you say? lets face it, AMD is still trailing behind primarily because of their stubbornness of not coloring outside the lines of their TF-chasing habits. That a first-generation intel GPU outperforms nearly every AMD GPU in RT and/or reconstruction quality is embarrassing.

I am not sure how ANYONE on this board could look at the PS4 and X1 tflops/performance gap then look at the PS4 and PS4 Pro tflops/performance gap and finally the PS4 Pro and X1X performance gap literally translating 1:1 and say hey I never felt TFs were the best way to measure performance. Never? really?

Again.. ps read what I at least have said, I can't speak for everyone else. I used very simar examples too. TF is a great way to measure performance when there is a wide disparity in TF between the hardware. On top all the other benefits of the hardwrae itself outside the APU. So it's easy to see how 1.2TF vs 1.8TF vs 4.2 vs 6TF can be tabulated. These TFs become less relevant when comparing 10 to 12.That's a 20% difference. A far less difference than any of the previous gen consoles compared to each other. When TF differences are that small... all the other smaller things that make up an APU or platform... start to become more relevant. This should be obvious and easy to understand when you consider that he whole TFs thing, only accounts for just one of ie 5 parts of a frame's render time.

Again, as I said.. if the XSX was 15TF to the PS5s 10TF... we won't even be having this conversation.

The fact that we still do it on PCs should tell you everything.

We actually don't do it on PCs nearly much. Especially with the new rampant TF obfuscation thats going on now with regard to institutions. Now we have GPUs that for x code can be X TF and for y code can be Y TF...

We arent even talking about the XSX here where performance comparisons are skewed at times due to DX12 fuck ups that we see on PCs as well as XS consoles. We are talking about Tflops of the PS5 Pro compared to the PS5's. If you want 2x more performance than the PS5, you need 2x more tflops. Real tflops, not the fake RDNA3.0 tflops. Plain and simple. Really doesnt matter how you get there. higher clocks or more CUs. Computing power is what matters most.

Now if cerny fucks up and bottlenecks himself like he did with the PS4 Pro's memory bandwidth or forgets to upgrade the vram like nvidia does with their midrange GPUs then thats a different discussion.

Come on man... you and I know that's just not true. You do not need 2x the TFs to get 2 times the performance. That's not the only way to achieve that. The easiest way to do it sure... but not the only way, not even the cheapest way either.

 

[cesar2xtreme]

I would lovely get a PS5 only for MGSDelta and FF7R2

 

[SlimySnake]

So it's easy to see how 1.2TF vs 1.8TF vs 4.2 vs 6TF can be tabulated. These TFs become less relevant when comparing 10 to 12.That's a 20% difference.

And yet we are talking about a 10 tflops PS5 and a 15 tflops PS5 Pro, same 50% difference. so why is it not relevant in this case?

Unless we are hoping for AMD to suddenly find 50% IPC gains, that 15 tflops PS5 Pro will not perform like a 20 tflops PS5.

If Sony wants 2x more performance from this thing then they need 2x more computing power. i dont care if it comes from extra CUs or higher clocks or both. They need the extra GPU power. You mentioned PCs dont do it? What? Even if both nvidia and AMD inflating their tflops, their GPU performance is still characterized by tflops. A 4070 has fewer tflops than a 4080 and performs worse. Same goes for a 4090 which has more tflops than a 4080 and performs better. The 7900xt even with its fake tflops is 52 tflops and performs worse than the 61 tflops 7900xtx. They are both effectively 26 and 30.5 RDNA2 tflops GPUs that line up 1:1 with the tflops difference between them and their rdna 2.0 predecessors.

If you want double the performance of a 10.7 tflops 6600xt, you will need a 6950xt equivalent which is a 23 tflops card. If AMD manages 25% IPC gains then you are looking at maybe a 17-18 tflops card. Somewhere between a 6800 and a 6800xt. 60 CUs, probably around 2.4 tflops.

 

[James Sawyer Ford]

We will see I guess, so far thinned are still following this pattern. It is a big investment that could be focused on games, tools, and R&D for PS6. This is getting diverted and helping cross generation periods.


Right now, they need more days in the base PS5 rather than launching a PS5 Pro, AMD still ships on PC so it is not like Sony needs the PS5 Pro to make the PS6. They make it because some people will buy it, period.


Do you think 3-4 years is enough to give you a meaningful update unless you raise the price and size/power envelope of the console by quite a bit (and then once you plane a $399-450 PS6 you have some problem to differentiate).

Chances are that PS5 Pro will be less of a jump from PS5 than PS4 to PS4 Pro and PS6 to PS6 Pro will be even smaller.

I feel this calls for consoles to keep changing to accommodate the frequent refreshes we have on PC or phones nowadays, but this is not the way tech is changing.

It's not that big of an investment. It keeps the hardware team constantly engaged, otherwise they wouldn't be doing much at all. It also provides for research into future products even if they don't get implemented into the PS5 Pro. It staves off platform migration. It provides a relatively low cost, high performance solution earlier for their most enthusiastic fans.

3-4 years is enough to give a meaningful update. Closer to 4K native, better Ray Tracing, better AA methodologies, higher framerates - similar enhancements the PS4 Pro offerred. And many people really enjoyed having the PS4 Pro, I am one of them. I am very glad I was able to experience games starting with Horizon running at much better fidelity even if it wasn't a generational leap.

 

[Mr.Phoenix]

And yet we are talking about a 10 tflops PS5 and a 15 tflops PS5 Pro, same 50% difference. so why is it not relevant in this case?

Unless we are hoping for AMD to suddenly find 50% IPC gains, that 15 tflops PS5 Pro will not perform like a 20 tflops PS5.

If Sony wants 2x more performance from this thing then they need 2x more computing power. i dont care if it comes from extra CUs or higher clocks or both. They need the extra GPU power. You mentioned PCs dont do it? What? Even if both nvidia and AMD inflating their tflops, their GPU performance is still characterized by tflops. A 4070 has fewer tflops than a 4080 and performs worse. Same goes for a 4090 which has more tflops than a 4080 and performs better. The 7900xt even with its fake tflops is 52 tflops and performs worse than the 61 tflops 7900xtx. They are both effectively 26 and 30.5 RDNA2 tflops GPUs that line up 1:1 with the tflops difference between them and their rdna 2.0 predecessors.

If you want double the performance of a 10.7 tflops 6600xt, you will need a 6950xt equivalent which is a 23 tflops card. If AMD manages 25% IPC gains then you are looking at maybe a 17-18 tflops card. Somewhere between a 6800 and a 6800xt. 60 CUs, probably around 2.4 tflops.

Errr I am lost.

Who said 10 vs 15TF is not relevant? What we have said is 10 vs 12 is not that relevant. 12 vs 13/14... not that relevant. !0/12 vs 16+ is very relevant.

 

[jroc74]

PS4 vs XBO also had a ram difference between the consoles. Just like the One X had more ram than the PS4 Pro.

TF did not tell the whole story.

 

[shamoomoo]

PS4 vs XBO also had a ram difference between the consoles. Just like the One X had more ram than the PS4 Pro.

TF did not tell the whole story.

There was a capacity and bandwidth difference between the pro and XboneX.

 

[Urban]

1440p/60fps thats all i need on console and thats the minimum i want!

 

[DenchDeckard]

And yet we are talking about a 10 tflops PS5 and a 15 tflops PS5 Pro, same 50% difference. so why is it not relevant in this case?

Unless we are hoping for AMD to suddenly find 50% IPC gains, that 15 tflops PS5 Pro will not perform like a 20 tflops PS5.

If Sony wants 2x more performance from this thing then they need 2x more computing power. i dont care if it comes from extra CUs or higher clocks or both. They need the extra GPU power. You mentioned PCs dont do it? What? Even if both nvidia and AMD inflating their tflops, their GPU performance is still characterized by tflops. A 4070 has fewer tflops than a 4080 and performs worse. Same goes for a 4090 which has more tflops than a 4080 and performs better. The 7900xt even with its fake tflops is 52 tflops and performs worse than the 61 tflops 7900xtx. They are both effectively 26 and 30.5 RDNA2 tflops GPUs that line up 1:1 with the tflops difference between them and their rdna 2.0 predecessors.

If you want double the performance of a 10.7 tflops 6600xt, you will need a 6950xt equivalent which is a 23 tflops card. If AMD manages 25% IPC gains then you are looking at maybe a 17-18 tflops card. Somewhere between a 6800 and a 6800xt. 60 CUs, probably around 2.4 tflops.

I feel this is a great post. I think they can get to a high clocked 6800XT by Xmas next year.

 

[bender]

45 is the number that comes at 44 and is also the number Michael Jordan wore after he came out of retirement for this first time.

Won't share any more info.

 

[asdasdasdbb]

45 is the number that comes at 44 and is also the number Michael Jordan wore after he came out of retirement for this first time.

Won't share any more info.

For 45 TF it'd have to be well over 3 Ghz for the GPU clock speed.

 

[bender]

For 45 TF it'd have to be well over 3 Ghz for the GPU clock speed.

 

[StereoVsn]

Starfield's 30FPS are definitely not there because of the CPU.

Its very probably that it actually is. Starfield is going to be tracking a lot of objects and in general will have fairly heavy CPU load in places judging by previous Bethesda games.

 

[Barakov]

O fuck off with that 8k bullshit

8k is still pretty far off. PS5 Pro is going to be 6.72k checkerboard up to 8k.

 

[cash_longfellow]

Trinity?

Playstation, Xbox, and Switch games available on the same system?

 

[Neo_game]

It looks like the worst case scenario is approx 50% more performance in raster 80% more in RT and 30% in BW. This is not a big upgrade and sound like 499$ console.