Moore's Law is Dead teasing a potential PS Vita successor

anonpuffs

Veteran
Icon Extra
29 Nov 2022
10,140
11,522
I can't see a PS6 being stronger than PS5 Pro being smaller than your TV hack in your living room 🤷‍♂️
If Sony continue with AMD there is no option for anything at all.. neither console stronger than PS5 Pro neither a portable.

What I expect from PS5 Pro?

Lower jump than PS4 to PS4 Pro and the biggest console ever created.
I believe they will use the old shell of the launch ps5 for the pro
 

ToTTenTranz

Veteran
Icon Extra
4 Aug 2023
1,295
1,307
I can't see a PS6 being stronger than PS5 Pro being smaller than your TV hack in your living room 🤷‍♂️
If Sony continue with AMD there is no option for anything at all.. neither console stronger than PS5 Pro neither a portable.

Sony isn't launching a PS6 that is less powerful than the PS5 Pro. Technology will evolve within the next 4 years.
 
  • Like
Reactions: Umar

kaluas

Member
3 Oct 2022
71
73
There are…. very little.
It is the smaller jump in a generation ever.
1,8TF to 10,8?
The controller was the best jump in their history alone.
Let's not forget SSDs, which changes game design entirely, 3D audio, RayTracing among other things.

Sure, Uncharted 4 God Of War and Detroit BH running on a PS4 is poetry, this doesn't mean that whatever comes after that is nothing.
 

Kokoloko

Veteran
Icon Extra
21 Jun 2022
5,293
4,184
I think something that is missing here in the discussion about there not being a dedicated library is that people can still make "handheld" oriented PlayStation games with lower budget that just happen to work on PlayStation consoles too.

How many PSP games ended up getting ported to PS2 in order to improve their margins? You had a game like Gravity Rush which was released on PS Vita and almost 4 years later get a PS4 remaster.

A PlayStation handheld might allow for A and AA games to thrive and could rescue the industry from falling into a spiral of pure AAA gaming.

As PlayStation becomes one platform across generations as alluded to in the thread about there no longer being hard resets, a handheld offering makes way more sense.

There will undoubtedly be PS6 exclusives that don't work on PS5 at some point particularly cross-gen PS7 games or games that make heavy use of next-gen AI, but going forward games will probably be released across two generations of consoles, maintaining a stronger userbase for game sales and handheld will further drive those margins.
I like the idea of smaller “handheld” oriented games. I guess indies have been the replacement for that in the last decade but Id still love that. Just depends if Sony gets there act together to have enough studios to do this. And Im sure there will be smaller studios who make these games for Nintendo that can do it for PS too.

A shared library is a must though.

PS6 games that wont run on PS5/ PSP would just either have to be skipped or drastically downgraded at that point.
After they release on PS6 ( no points stopping the full potential of the PS6 game to cater for PS5/PSP3 owners) They’ve have to be like 720p, and reduced in features if ported at all
 

ethomaz

Rebolation!
Icon Extra
21 Jun 2022
10,859
8,886
Brasil 🇧🇷
PSN ID
ethomaz
1,8TF to 10,8?
The controller was the best jump in their history alone.
Let's not forget SSDs, which changes game design entirely, 3D audio, RayTracing among other things.

Sure, Uncharted 4 God Of War and Detroit BH running on a PS4 is poetry, this doesn't mean that whatever comes after that is nothing.
It is 10.3TFs.
But hey let's look at the whole configuration.

PS3
Cell CPU
192GFs (some says that Cell add 230GFs when used to render graphics but nobody knows for sure and what extend you can use Cell for that and hot it affected the CPU performance side... you probably could use only a fraction of these 230GFs for graphics)
512MB RAM

PS4
Jaguar CPU is inferior to Cell.
1.84TFs (+9.6x)
8GB RAM (+16x)

PS5
Zen2 CPU (not sure how it is compared to Cell)
10.3TFs (5.6x)
16GB RAM (+2x)

If you go back the jump was even bigger from PS1 to PS2 to PS3.

If you add RayTracing to the comparison then PS5 is probably weaker than PS4... I mean rendering a RT scene in PS5 will be slower than rendering the same scene with rasterization on PS4.
 

xollowsob

Veteran
6 Jan 2024
1,017
840
PS4
Jaguar CPU is inferior to Cell.
1.84TFs (+9.6x)
8GB RAM (+16x)

PS5
Zen2 CPU (not sure how it is compared to Cell)
10.3TFs (5.6x)
16GB RAM (+2x)
Bear in mind that the ram type is faster than that of the ram in PS4. While the number may only double, the performance is much higher.

Likewise with TF. They aren't comparable.
 

ethomaz

Rebolation!
Icon Extra
21 Jun 2022
10,859
8,886
Brasil 🇧🇷
PSN ID
ethomaz
Bear in mind that the ram type is faster than that of the ram in PS4. While the number may only double, the performance is much higher.

Likewise with TF. They aren't comparable.
176GB/s vs 440GB/s (+2.5x)

My point is that hardware is not moving like it moved before.
Chips are not becoming smaller.

We can talk about the Moore Law... it say you need double the numbers of transistors each 2 years... basically in the past a move from a node to a new node allowed you to have double the transistors using the same space... and that happened each 2 years.

Today that doesn't happen anymore.
Moving from 7nm to 5nm that physical should cutting the size in half or putting twice the transistor in the same size because a marketing name only... what happens today in the 7nm to 5nm are just improvements in the same process node and not actually a new node.
7nm to 5nm is like 10 to max 20% more transistors in the same size... it is far way from double (100%).

That is a big issue to the hardware industry that slowed down and are having to put bigger and bigger chips to reach the desired performance.

But hey... what you expect from PS6?

40TFs RDNA3? That is less than 20TFs from RDNA2 that means less than double of PS4 is.
We are taking at very small increase in raw power compared with previous generations... hell it is even small compared with mid-gen refresh PS4 Pro.

Memory? Do you believe the memory size will reach 64GB or the speed reach 2TB/s anytime in the future? No it will probably be around 24-32GB with 700-900GB/s sppeeds.

CPU? We know each new Zen add like less than 10% IPC increase.
They try to counter that with both clock and core count.
I can't see they putting more than 16 cores CPU on PS6... and that is like 2x increase in CPU power... added the higher clock you will have something like 2.5x to 3x max.

The hardware is basically stagnated with minor improvements.
Why GPU manufacturers pushed DLSS, TAA, FSR, etc... these techs were all created to free up GPU because the couldn't envolve GPU hardware at the paces they need anymore... and most of these techs give worst results in terms of image quality than what you had before.
But they can't stop to release new hardware because... if they wait 4-5 years to release a new GPU they will be dead as company.... so they have to do what they are doing... minor improvement in raw power but with new software techs to help to mitigate the gap. and give the ideia of big upgrade (when I say big I mean 30% max in a new GPU generation in each 2 years... that is what is considered big nowdays).

I said in 2018 in GAF that we didn't have tech to a Switch Pro or Switch 2 release.
We are in 2024 and Nintendo is still suffering to release a successor that was again delayed to 2025 by the rumors.

And you can already expect the Switch 2 will have small improvement over Switch... it will at max reach the same raw power found in Steam Deck and similar... to be fair for that Steam Deck had to be big in size something that Switch 2 can't.
So I expect Switch 2 to have less raw power than Steam Deck.

Because that where hardware is today.

PS6... either Sony push a not ever seem giant size of console (and finally breaks that 250W power draw limit... something hard because that is what most common plugs give you in the world... and use special plugs requires changes in most people homes)... or they will have not even close jump in performance like previous generations.
 
  • they're_right_you_know
Reactions: xollowsob

ToTTenTranz

Veteran
Icon Extra
4 Aug 2023
1,295
1,307
There are…. very little.
It is the smaller jump in a generation ever.

It's not very little at all. The difference in effective IPC between GCN2 and RDNA2 is enormous, even at similar TFLOPs throughput. The Steam Deck is head and shoulders above the PS4 and it usually clocks around 1.2GHz (1.2 TFLOPs). The PS5's GPU isn't just 5.5x faster than the PS4's, it's probably over 7x faster in modern games.
The CPU is also heaps and bounds faster, with one Zen2 core being around 3x faster than a Jaguar core at ISO clocks, meaning it's almost 7x faster if you compare them at their respective clocks.

And then we're getting >50x faster storage with >1000x lower latency, which allows for lower amount of data to cache in the RAM/VRAM, together with all the advantages from that.


176GB/s vs 440GB/s (+2.5x)

The PS4's GPU didn't use delta color compression, so the effective bandwidth increase is a lot higher than those 2.5x. Even the GCN3, AMD's first architecture to use color compression, showed an enormous improvement in effective bandwidth, which is why the Tonga GPUs (R9 280, R9 380X) had practically the same performance as the older Tahiti (HD 7970), despite only getting 66% of the memory bandwidth.
When they got to RDNA2, AMD is probably taking twice the effective bandwidth out of each GB/s, than they did with GCN2.
There's also a generational upgrade in geometry performance.
 

ethomaz

Rebolation!
Icon Extra
21 Jun 2022
10,859
8,886
Brasil 🇧🇷
PSN ID
ethomaz
It's not very little at all. The difference in effective IPC between GCN2 and RDNA2 is enormous, even at similar TFLOPs throughput. The Steam Deck is head and shoulders above the PS4 and it usually clocks around 1.2GHz (1.2 TFLOPs). The PS5's GPU isn't just 5.5x faster than the PS4's, it's probably over 7x faster in modern games.
The CPU is also heaps and bounds faster, with one Zen2 core being around 3x faster than a Jaguar core at ISO clocks, meaning it's almost 7x faster if you compare them at their respective clocks.

And then we're getting >50x faster storage with >1000x lower latency, which allows for lower amount of data to cache in the RAM/VRAM, together with all the advantages from that.




The PS4's GPU didn't use delta color compression, so the effective bandwidth increase is a lot higher than those 2.5x. Even the GCN3, AMD's first architecture to use color compression, showed an enormous improvement in effective bandwidth, which is why the Tonga GPUs (R9 280, R9 380X) had practically the same performance as the older Tahiti (HD 7970), despite only getting 66% of the memory bandwidth.
When they got to RDNA2, AMD is probably taking twice the effective bandwidth out of each GB/s, than they did with GCN2.
There's also a generational upgrade in geometry performance.
GPU

RDNA2 IPC is 4-5% slower than RDNA.
RDNA2 IPC is 36% faster than GCN2.

GPU terms Steam Deck is still slower than PS4... it should be similar to PS4 running at 90% of it clock.

In raw power PS5 should be around 7.6x more than PS4.
5.6x or 7.6x continues being small jump for a generation.

CPU

Zen2 IPC is 49.5% faster than Jaguar's IPC (same clock).
PS4's Jaguar runs at 1.6Ghz.
PS5's Zen 2 runs at 3.5Ghz.
They have the same amount of cores... and SMP is not used in games.

Relatively the difference is around 3.3x.
3.3x is still a small jump for CPU in a generation... it is even smaller than GPU/RAM jump.

MEMORY BANDWIDTH


I have no evidence how much the memory compression really helps in that.
In fact most stronger GPUs in the market have significant increase in bandwidth even with memory compression... that lead me believe the key factor is still raw memory bandwidth.
 

kaluas

Member
3 Oct 2022
71
73
5.6x or 7.6x continues being small jump for a generation.
And they'll probably keep getting smaller, especially when you compare them like that. I don't know how you get your numbers, and i highly doubt their validity, that isn't the case though. You can't compare FLOPS between Series X and PS5, let alone PS5 and PS4. A bunch of new tech was introduced with these consoles and a shit ton of QoL improvements.

Nothing will beat PS1 to PS2 jump, that is for certain, for a number of reasons. The most important of all is that companies then had wider profit margins regarding hw sales. Another important reason, in regards to an earlier post is that nobody will give you 16 times the RAM of PS4, as they gave you the previous generations, cause no console software would or should need 128GBs or RAM.
 

ToTTenTranz

Veteran
Icon Extra
4 Aug 2023
1,295
1,307
Zen2 IPC is 49.5% faster than Jaguar's IPC (same clock).

Where are you even taking these numbers from? It's way faster than 49.5%.

Here are the Geekbench results between a 8-core Jaguar A9-9820 and the PS5 SoC on a PC:


Single-threaded score is 250 on the Jaguar and 1290 on PS5's Zen2. It's 5.16x faster.
Normalize for frequency and it's 4.3x faster.

And this is just single-threaded. The difference in multi-threading is even bigger due to Zen2 cores running two threads.



You're off by an order of magnitude.
 

anonpuffs

Veteran
Icon Extra
29 Nov 2022
10,140
11,522
Where are you even taking these numbers from? It's way faster than 49.5%.

Here are the Geekbench results between a 8-core Jaguar A9-9820 and the PS5 SoC on a PC:


Single-threaded score is 250 on the Jaguar and 1290 on PS5's Zen2. It's 5.16x faster.
Normalize for frequency and it's 4.3x faster.

And this is just single-threaded. The difference in multi-threading is even bigger due to Zen2 cores running two threads.



You're off by an order of magnitude.
Geekbench isn't indicative of real world performance. But I'd expect zen2 to at least double a jaguar in games.
 

ethomaz

Rebolation!
Icon Extra
21 Jun 2022
10,859
8,886
Brasil 🇧🇷
PSN ID
ethomaz
Where are you even taking these numbers from? It's way faster than 49.5%.

Here are the Geekbench results between a 8-core Jaguar A9-9820 and the PS5 SoC on a PC:


Single-threaded score is 250 on the Jaguar and 1290 on PS5's Zen2. It's 5.16x faster.
Normalize for frequency and it's 4.3x faster.

And this is just single-threaded. The difference in multi-threading is even bigger due to Zen2 cores running two threads.



You're off by an order of magnitude.
Your comparison link is not even putting them at same clock... it is not comparing IPC and depends of others parts.
IPC shows how much an architecture increased performance over the other... it is not based in clock or external componentes.
So these Geekbench results are pretty useless for comparisons because there is no way to put that old CPU in a fast motherboard with faster RAM... so the comparison is biased due the bottlenecks from the components.

BTW these are a compilation of IPC comparison and it includes PS4 CPU via hacking:

Cortex-A73 (ARMv8-A) 64.4
Pentium M 69.0
Bulldozer (AMD FX) 76.2
K10 (Phenom II) 76.7
Goldmont Plus (Atom 2nd gen) 77.5
Core 80.0
Bobcat 81.3
Nehalem 90.0
Jaguar (PS4) 93.5
Sandy Bridge 100.0
Tremont (Atom 3rd gen) 100.7
Ivy Bridge 105.8
Cortex-A76 (ARMv8.2-A) 114.7
Zen 115.9
Haswell 117.6
Zen+ 119.3
Broadwell 121.5
Skylake 124.0
Zen 2 139.8
A10 (ARMv8.1-A) 139.8
Cortex-A77 (ARMv8.2-A) 141.0
Sunny Cove Ice Lake 146.3
Cypress Cove Rocket Lake 147.6
Zen 3 167.4
A11 (ARMv8.2-A) 186.0
A12 (ARMv8.3-A) 213.8
A13 (ARMv8.4-A) 235.0
A14/M1 (ARMv8.5-A) 236.7

BTW it close to what AMD PR told us via press release.

I have no ideia where people got these 2x, 3x, 4x, 5x, etc in CPU increase with new Archs... AMD talks at max 40% increase from Jaguar to Zen.

amd-40-ipc-zen-zen.jpg


AMD and Intel CPUs got increase like 10-20% each year with some bigger increase when you have big changes in the Arch like happened with Zen... after that Zen, Zen 2, Zen 3 you are looking at 10-20% in performance increase.

For exemple AMD about Zen 3 "major changes" to increase in performance:

Zen3_arch_7.jpg


A CPU new Arch doubling the performance of the previous Arch I don't see since well ever? I really don't remember it.
 
Last edited:

ToTTenTranz

Veteran
Icon Extra
4 Aug 2023
1,295
1,307
BTW these are a compilation of IPC comparison and it includes PS4 CPU via hacking:


That isn't a compilation of "IPC comparison", it's a compilation of SPECint benchmarks that specifically measures integer performance like it says in the benchmark's own name. Games depend mostly on floating point performance (higher range of values), not integer, and FP operations even occur in a different part of the chip, different ALUs.



Your comparison link is not even putting them at same clock...
And neither does yours.. he says right there he used Anandtech results and then normalized by clock frequencies, just like I did.




I have no ideia where people got these 2x, 3x, 4x, 5x, etc in CPU increase with new Archs... AMD talks at max 40% increase from Jaguar to Zen.

amd-40-ipc-zen-zen.jpg

Dude where is Jaguar in that graphic?
Jaguar isn't Excavator, it's not even based on Bulldozer. It was an iteration over Bobcat, a completely different CPU architecture aimed at low-power operation to compete with Intel's Atom cores. It had much lower IPC than Excavator.

Excavator cores are almost 90% faster than Jaguar cores after normalizing for frequency.
And the IPC upgrade betweem Excavator and Zen1 wasn't 40%, it ended up being much closer to 60%.
1.9 (Jaguar vs. Excavator) * 1.6 (Excavator vs. Zen1) * 1.29 (Zen1 vs Zen2) = 3.92x higher single-core IPC between Jaguar and Zen2.


Couple that with twice the frequency and you're getting ~8x the performance on single-thread alone.



Geekbench isn't indicative of real world performance. But I'd expect zen2 to at least double a jaguar in games.
It's among the best there is for measuring IPC. Probably better than Cinebench, at least for reporting.
 

ethomaz

Rebolation!
Icon Extra
21 Jun 2022
10,859
8,886
Brasil 🇧🇷
PSN ID
ethomaz
That isn't a compilation of "IPC comparison", it's a compilation of SPECint benchmarks that specifically measures integer performance like it says in the benchmark's own name. Games depend mostly on floating point performance (higher range of values), not integer, and FP operations even occur in a different part of the chip, different ALUs.




And neither does yours.. he says right there he used Anandtech results and then normalized by clock frequencies, just like I did.






Dude where is Jaguar in that graphic?
Jaguar isn't Excavator, it's not even based on Bulldozer. It was an iteration over Bobcat, a completely different CPU architecture aimed at low-power operation to compete with Intel's Atom cores. It had much lower IPC than Excavator.

Excavator cores are almost 90% faster than Jaguar cores after normalizing for frequency.
And the IPC upgrade betweem Excavator and Zen1 wasn't 40%, it ended up being much closer to 60%.
1.9 (Jaguar vs. Excavator) * 1.6 (Excavator vs. Zen1) * 1.29 (Zen1 vs Zen2) = 3.92x higher single-core IPC between Jaguar and Zen2.


Couple that with twice the frequency and you're getting ~8x the performance on single-thread alone.




It's among the best there is for measuring IPC. Probably better than Cinebench, at least for reporting.
Man 😂😂😂😂

Jaguar IPC is stronger than Bulldozer.

Zen to Bobcat is 42% increase in IPC.
Jaguar to Bobcat is 15% increase IPC.
Zen to Bulldozer is 4x% increase in IPC (by AMD… it is probably a bit lower).

You come with these weird 4x, 8x, etc that never existed… even AMD doesn’t claim even when their own inflated %.

And just to stop the bullshiting here.
There is Jaguar in the place I said it is.

images


Jaguar to Escavator IPC is less than 10% difference.

Edit - I realized you can even read your own link… it shows 35% of Escavator performance with with 33% less clock in single core performance… with big component difference.

RAM speed alone is what makes the big difference between performance… not CPU… I understood now?

CPU performance at the same clock (IPC) they are very close to each other.
 
Last edited:
  • they're_right_you_know
Reactions: flaccidsnake