StarTopic Future Nintendo Hardware & Technology Speculation & Discussion |ST| (New Staff Post, Please read)

[Shoulder]

What makes PS4 and Xbone CPU so bad? Since both of these platforms delivered some of the most impressive looking games of the generation.
Like, what were AMD failings with the Jaguar, and how will the Switch 2 not face these issues?

Like... If these were such huge problem, then why did Sony go with it, is it because of cost, since i remember Nvidia or Sony mentioning the reason Xbox One and PS4 didn't go with Nvidia is because of how expensive these Nvidia chips would be? If so, then what makes Nintendo partnership different then them?

There’s going to be some nuance here, but by comparison, the “Espresso” CPU in the Wii U, despite being only a Tri-Core setup, wasn’t far behind the jaguar CPUs in PS4, and Xbone. The additional cores is really what helped push it above the Wii U CPU.

And this is despite the Espresso CPU was effectively a heavily modified IBM PowerPC 750 CPU from the late 90s. Core for Core, it was about the same as the Jag cores, but again, there are nuances here.

This is where someone else might have to step in for corrections, but I believe the Espresso CPU benefited greatly in Integer point calculations, whereas like most x86 CPUs, floating point calculations.

For all the Wii U’s faults, say Nintendo decided to go with an 8-core variant of the Wii CPU, it probably would’ve overall been more capable than the Jaguar CPU. But we can only imagine what a Frankenstein CPU using the PowerPC 750 would’ve been like.

for more information, just saying "netbook cpus" doesn't really explain things. this era of AMD cpus was bad. the bulldozer family couldn't be saved in any way other than budget pricing, which is where AMD gets their current reputation of being the budget option from

Prior to my Intel 11400F, I was rocking the AMD FX-8320, and while Bulldozer wasn’t particularly great, I had hoped given it was AMD, and that Sony, and Microsoft were also using AMD, it would’ve been a good PC equivalent at the time. Boy was I wrong. Intel’s quad-core CPUs at the time, such as the legendary 4770k, would run circles around the AMD FX, and especially so the Jag cores.

The only saving grace for AMD FX was when it came to overlocking, which when done right, would produce some great results. It even had the side benefit of practically being a space heater, so it was great in the winter time!

 

[Darknut85]

There’s going to be some nuance here, but by comparison, the “Espresso” CPU in the Wii U, despite being only a Tri-Core setup, wasn’t far behind the jaguar CPUs in PS4, and Xbone. The additional cores is really what helped push it above the Wii U CPU.

And this is despite the Espresso CPU was effectively a heavily modified IBM PowerPC 750 CPU from the late 90s. Core for Core, it was about the same as the Jag cores, but again, there are nuances here.

This is where someone else might have to step in for corrections, but I believe the Espresso CPU benefited greatly in Integer point calculations, whereas like most x86 CPUs, floating point calculations.

For all the Wii U’s faults, say Nintendo decided to go with an 8-core variant of the Wii CPU, it probably would’ve overall been more capable than the Jaguar CPU. But we can only imagine what a Frankenstein CPU using the PowerPC 750 would’ve been like.



Prior to my Intel 11400F, I was rocking the AMD FX-8320, and while Bulldozer wasn’t particularly great, I had hoped given it was AMD, and that Sony, and Microsoft were also using AMD, it would’ve been a good PC equivalent at the time. Boy was I wrong. Intel’s quad-core CPUs at the time, such as the legendary 4770k, would run circles around the AMD FX, and especially so the Jag cores.

The only saving grace for AMD FX was when it came to overlocking, which when done right, would produce some great results. It even had the side benefit of practically being a space heater, so it was great in the winter time!

If you overclock it enough, it may can be used for some graphical effects…

 

[PedroNavajas]

Regarding CPUs.

Quick refresher. Geekbench is not a perfect benchmark, but it's a pretty okay one, and it happens to have a lot of public data. Unfortunately, it doesn't run on consoles. Fortunately, AMD has sold console CPUs in the past as "desktop kits" which means that we have PC benchmarks of some last gen/current gen machines.

Unfortunately, there is no system with Switch 2's exact CPU setup, and we're only guessing at clocks. Fortunately, Drake, the chip in Switch 2, has a "sister chip", Orin, which is well benchmarked, and has a configuration that is a pretty good approximation.

Here is a list of consoles, their hardware equivalents, and a sample geekbench score, for those of you playing along at home.

Xbox One X/PS4 Pro: A9-9820 "Cato"
"Cato" is the product name of a set of desktop CPUs made out of scrapped Xbox One APUs. The PS4 had a very similar CPU. They would clock the CPUs higher in their mid-gen versions, and Cato here is default clocked to the One X number

Single core: 250
Multi-core: 1244

Switch 2: Orin NX, 8 Core
Orin runs a variant of the CPU for the automotive industry. Drake uses a laptop/gaming variant, but it's the same core technology. This configuration of Orin runs 8 cores, same as Switch 2, at 1.98 GHz, which I think is probably slightly higher than Nintendo will go. But I'm not made of money, I can't buy my own for a benchmark.

Single core: 912
Multi-core: 4914

Uhh, wait, are you serious, a 4x leap from the PS4 Pro??? Yes. This is not because the CPU in Switch 2 is miraculous, it's because the CPU in the last gen consoles were hot garbo. I cannot emphasize this enough. AMD made a bunch of gambles with the Jaguar cores and not a single one paid off.
Steam Deck: Valve Jupiter aka The Steam Deck
We don't need a proxy here, folks have just benchmarked the real thing

Single core: 1400
Multi-core: 4845

Uhh, wait, single core is much higher than Switch 2, but multi-core is basically the same??? Steam Deck only has 4 CPU cores. Technically each of those cores can run two threads at the same time, but it's not a super efficient performance gain. Switch 2 only has one thread per core, but it has 8 of them. There is no one number that describes performance, or even one aspect of performance.

...

Moreover, according to Geekbench, that score has the Jaguars clocked at @2.34 GHz, which is over 40% higher than PS4 1.6Ghz. and about 10% over the PS4. So, 912/((1.6/2.34)*250) = 5.335 . Orin's CPU is over 5 time faster than PS4's.

Also, Orin has an IO chip, further freeing CPU resources.

 

[Pokemaniac]

No, not really. The reason for the burst mode on the final unit is to drastically reduce loading times by allowing the CPU to go full throttle and decompress assets. Switch 2 has a dedicated engine for this that should drastically alleviate the bottleneck and greatly reduce the need to even do such a thing.

I don't know that it's obvious that decompression accelerators would remove the need for a CPU boost during loading. There's a lot of other stuff going on, like parsing data, mutating the internal scene state in the engine, procedural generation, and initialization.

 

[ERezaSwitch]

Pyoro is back

 

[eye0h]

Pyoro is back

the media said he was done, that his bag was empty.

pyoro.... WHAT WILL YOU DO??

 

[Hermii]

Where is the human size?

That's what came out on Google translate.

 

[RennanNT]

Where is the human size?

That's what came out on Google translate.

The Nindai most likely stands for Nintendo D(a)irect.

But they're just making fun of the whole Midori mess.

 

[Aufhebung RPG]

Where is the human size?

That's what came out on Google translate.

Nindai = Nintendo Direct, and doko literally means "where" in Japanese.

 

[oldpuck]

Thanks for this. I will say though, at least in terms of the Steam Deck, it has high numbers, but those also seem to be at max clock of 3.5Ghz, which is only possible when the GPU is not pushed heavily. But of course these are CPU benchmarks, not full system benchmarks. And while the "Switch 2" comparison via Orin NX is used, could the actual multi-core number end up being a bit higher because of the single cluster design vs two cluster? Or does that only really affect power consumption?

Multi-core would probably be a little higher. The single cluster design means that sharing data is faster amongst threads.

 

[ItWasMeantToBe19]

Where is the human size?

That's what came out on Google translate.

"Where Nintendo Direct"

He's probably saying it's not happening this week.

(Probably because he sees nothing on the eShop backend)

 

[LiC]

 

[Hocot8]

"Where Nintendo Direct"

He's probably saying it's not happening this week.

(Probably because he sees nothing on the eShop backend)

you should probably not jump to such a conclusion.

 

[LiC]

Pyoro is back

Reminder that Direct posts and especially spoilers belong in their own thread.

 

[Concernt]

* Hidden text: cannot be quoted. *

 

[Shoulder]

If you overclock it enough, it may can be used for some graphical effects…

Fun fact: in the original Gekko CPU in the GameCube, some games did run graphical effects off of it. Most famously, Factor 5 was able to run some effects in Star Wars Rogue Squadron 3 Rebel Strike.

I do remember reading an article on IGN back in the day about that with Factor 5. I want to say it was some of the lighting effects that were run on the CPU vs. the GPU, and this is on top of all the AI scripts, physics, and any other executions the CPU had to run. Sheer wizards those at Factor 5 I tell ya.

Factor 5 got about every bit of juice out of that GameCube at the time, though admittedly, I would love to see the likes of Kaze, or James Lambert who have made their names on optimizing crazy stuff on the N64, do so on the GameCube.

 

[RedSpring]

Also thanks for the response, I would have never expected these CPU to bad, since Sony and Microsoft main shtick, is that everything about theirs consoles is next level, without any bottlenecks.

That's the power of marketing. If people feel you are next level then you are regardless of the specs.

What makes PS4 and Xbone CPU so bad? Since both of these platforms delivered some of the most impressive looking games of the generation.
Like, what were AMD failings with the Jaguar, and how will the Switch 2 not face these issues?

Like... If these were such huge problem, then why did Sony go with it, is it because of cost, since i remember Nvidia or Sony mentioning the reason Xbox One and PS4 didn't go with Nvidia is because of how expensive these Nvidia chips would be? If so, then what makes Nintendo partnership different then them?

It's the best AMD can offer at the time and they went with AMD because they can put in a bad CPU with a pretty good GPU into a single chip versus Intel putting a good CPU and down right horrible GPU into a single chip. The graphic is what sell so AMD it is. The single chip is great for cost and space saving but it's one of the compromise that Sony and Microsoft have to do for their console.

As for why their games are some of most impressive games of the generation. One part is that CPU doesn't have to do much for graphic to put it very simply if a bit inaccurate. Another part is the amount of resources. PS4 and Xbox One are the primary platforms for these games so developers would make sure that their games would look good on those consoles. The lightning, the texture, the size would all be designed in consideration of the hardware limitations. If they didn’t, you will have something similar to some of Switch's ports. Impressive but ugly.

There were some problems with working with NVDIA. We don't know for sure what happened since it's behind the scene but we know companies tried to work with NVDIA but backed out later. AMD probably have a more mature team for custom chip development and offering a better deal at the time.

NVDIA also does not have the right to design x86 chips. The chips that AMD and Intel are producing. They do have the right to design ARM chip but ARM can't compete with the x86 chip in power but much more energy efficient. For Nintendo making a hybrid, the energy efficiency was crucial.

Will Sony and Microsoft switch to NVDIA in the future now that ARM is catching up in power? Probably not, AMD can also design ARM CPU and switching to ARM and NVDIA GPU would probably means no Backward Compatibility.

 

[LiC]

 

[Giancarlo]

Pyoro is back

Ninkai Doko= to a new Donkey Konga game?

 

[Dakhil]

Samsung's 4 nm* process node still doesn't look good, assuming that this isn't Samsung's incompetence at integrating Arm's IPs at play.

* → a marketing nomenclature used by all foundry companies

 

[ItWasMeantToBe19]

Ninkai Doko= to a new Donkey Konga game?

The translation is "Where Direct"

 

[Concernt]

* Hidden text: cannot be quoted. *

 

[ShadowFox08]

I think you misinterpreted my context.

My point is we won’t be getting N64 to GCN leaps in graphics. At best, it’ll be Wii to Wii U.

SNES > N64 > GC were the 3 biggest leaps.

I think in part because the industry itself was going through a shift. Both SNES/Genesis/TG-16 and consoles designed in the 1980s focused on pulling off the shelf parts and adding a custom chip to do what the mfg wanted to do. The Motorola 68000 in the Genesis was 10 years old by the time Genesis launched in 1989 and was widely adopted by home computers like Atari ST/Amiga as well as arcades at the time.

The SNES CPU design was 6 years old by the time it released and contained a core going back to the 6502 from the 1970s as well.

Starting with the N64, game companies starting adopting bleeding edge, sometimes not yet released chips to their consoles and it contributed to the leaps we saw at the time

I don't know how strong n64 to GC is on the top of my head...

But I will say Wii to Wii U is pretty huge. Just comparing on paper speeds speeds alone , xbox 360 GPU is 12x faster then Wii 12gflops vs 240). 360's GPU has 36% more speed over Wii U's, but Wii U's is newer and edges out with efficiency and new tech. i dunno how much, But you can clearly see it in games. So it's

In regards to CPU, it's 729MHz single core vs 1.24Ghz tri core. So easily 5x CPU wise.

Switch to Switch 2 will be like N64 to Gamecube

And Switch 2 to Next-gen will be like Gamecube to Wii with VR being the main thing most likely

My two yens.

For Switch to Switch 2, geez.. I don't think it will be that bad. Wii was essentially a gameCube 1.5.. 1.5x the speed in GPU and CPU, but almost 3x more memory.

Despite the transistor/node slowing down, I do think In 7-8 years time a minimum of a 2.5-3x upgrade (Wii to Wii u? will be possible for the bare minimum for the worse case scenario, and 4-5x being possible (half of switch vs switch 5)

if we look at tsmc's road map in power efficiency gains:

Node N5 vs N3E->N3E vs N2P->N2P- vs A16

Power 34%->30-40%-> 15-20%
or
Performance 18%->15-20%->8-10%

5nm to A16 has some significant power draw and/or power changes. 1.55x speed and/or 3x power efficiency gains if my math is right.A16 is expected to release in 2026. it we'll likely see at least one or two more node and architecture gains beyond that if switch 3 releases in 2031-2033.

TSMC's Roadmap at a Glance: N3X, N2P, A16 Coming in 2025/2026

www.anandtech.com

Only thing I don't know is (and has confused me) is if both power and performance gains are received at the same, or it's one or the other.

So in short.. I think a 2033 Switch 3 release with 10-12 TFLOPs GPU docked, 3x CPU speed (via higher raw numbers or architectural and/or both) and 24 GB RAM with lpddr6x (300GB/s bandwidth) is plausible.

4-5 TFLOPs GPU handheld should be doable with 3x power efficiency and architectural gains (and worse case scenario: more efficient battery tech) with 3 hours battery. Hard to imagine at 8-10 watts I know, but some handheld PCs are already at that range with the GPU, with 25 watts or more.

There’s going to be some nuance here, but by comparison, the “Espresso” CPU in the Wii U, despite being only a Tri-Core setup, wasn’t far behind the jaguar CPUs in PS4, and Xbone. The additional cores is really what helped push it above the Wii U CPU.

And this is despite the Espresso CPU was effectively a heavily modified IBM PowerPC 750 CPU from the late 90s. Core for Core, it was about the same as the Jag cores, but again, there are nuances here.

This is where someone else might have to step in for corrections, but I believe the Espresso CPU benefited greatly in Integer point calculations, whereas like most x86 CPUs, floating point calculations.

For all the Wii U’s faults, say Nintendo decided to go with an 8-core variant of the Wii CPU, it probably would’ve overall been more capable than the Jaguar CPU. But we can only imagine what a Frankenstein CPU using the PowerPC 750 would’ve been like.

Eh.. I don't know how much faster the Jaguars are vs the gen before, but the Wii U's CPU struggled against the 360's CPU. It was quite embarrassing in multiplatform games, as it showed the former had better frame rate and/or more objects on screen for the former.

Despite the Wii U having a stronger GPU due to better architectural advantages (176gflops vs 360's 240 TFLOPs) and more RAM (512 vs 2GB) some AAA parties were just not running as good as 360. The CPU was apparently easy to work and it was the clockspeeds. i don't know how much newer Wii U's CPU was.. But that 1.24 GHZ tri core CPU vs 3.2 GHz tri core CPU could not compete

 

[monkeygotyou]

The Games drought of the Wii U era has been replaced by the News drought of the Switch era

 

[Shish]

Off topic:
I'm curious how people will compare Switch 2 with PlayStation 6 and Xbox next gen.

 

[oldpuck]

Off topic:
I'm curious how people will compare Switch 2 with PlayStation 6 and Xbox next gen.

unfairly and with derision

 

[Shoulder]

Eh.. I don't know how much faster the Jaguars are vs the gen before, but the Wii U's CPU struggled against the 360's CPU. It was quite embarrassing in multiplatform games, as it showed the former had better frame rate and/or more objects on screen for the former.

Despite the Wii U having a stronger GPU due to better architectural advantages (176gflops vs 360's 240 TFLOPs) and more RAM (512 vs 2GB) some AAA parties were just not running as good as 360. The CPU was apparently easy to work and it was the clockspeeds. i don't know how much newer Wii U's CPU was.. But that 1.24 GHZ tri core CPU vs 3.2 GHz tri core CPU could not compete

Agreed. That's also why I left some nuance there because I know the Xenon CPU in the 360 also ran nearly 3x the frequency. One advantage though the Espresso had over Xeon was out-of-order executions, which 360 could not do. How this actually translates into gaming I could not explain well, though I do remember it was discussed by Microsoft and IBM when designing the chip. To save on costs though, they removed that feature.

There was also the power consumption aspect too. Wii U ran at most 40 watts, but realistically was 30-35 watts under most loads. Xbox 360 by contrast was over 150 watts, though by the time the Slim model came out, it was cut down in half or more. It's possible the Espresso could’ve ran at a higher frequency, though it doesn’t appear any of the 750 family of PPC processors ever ran at higher frequencies. At this time, Xbox was about brute force, and Wii U was more about efficiency, and Nintendo continues this route to this day.

 

[Steve]

Off topic:
I'm curious how people will compare Switch 2 with PlayStation 6 and Xbox next gen.

I don't think the leap of PS5 to PS6 will be huge, with the exception of better fps.

Hopefully native 4K will be possible.

but the comparison will be there and most Sony and Xbox fans will just clown on the Switch 2, until they realise they don't have games.

 

[Danny]

Off topic:
I'm curious how people will compare Switch 2 with PlayStation 6 and Xbox next gen.

Doesn't matter, the Switch 2 will have exclusive games day 1. Something their new consoles won't have for years.

 

[Steve]

I'm sometime surprised how small the Switch Dev-Kit is, especially considering, it's pretty much a regular Switch with a large back.

Like, i'm personally intrigued how big the Switch 2 Dev-Kit looks like, if it's similar to the Switch Dev-Kit.

 

[ItWasMeantToBe19]

I don't think the leap of PS5 to PS6 will be huge, with the exception of better fps.

Hopefully native 4K will be possible.

but the comparison will be there and most Sony and Xbox fans will just clown on the Switch 2, until they realise they don't have games.

PS6 should allow much higher detail via virtualized geometry, 1080p upscaled to 4K with good methods, and full ray tracing.

Should be a pretty big leap. Basically Hellblade 2 tier visuals but for real games with real level design and stuff happening on screen.

 

[Concernt]

I'm sometime surprised how small the Switch Dev-Kit is, especially considering, it's pretty much a regular Switch with a large back.

Like, i'm personally intrigued how big the Switch 2 Dev-Kit looks like, if it's similar to the Switch Dev-Kit.

There are usually several iterations of development kit for a given console. The most recent Nintendo Switch development kit is, basically, an OLED Model with more RAM and different software!

Early dev kits can be pretty sketchy contraptions. I'd imagine developers on board for launch more or less have final-ish units (like the one pictured) kept under lock and key at the moment, while others that are on board but maybe not trusted as much might have a black box with a Nintendo Switch Pro Controller hanging off it. Others still might have access to NVN2 and tools but not hardware, and told to develop using a virtual environment for now.

 

[Concernt]

virtualized geometry, 1080p upscaled to 4K with good methods, and full ray tracing.

Honestly I'd expect some games to do this on the successor; though maybe with a considerable sacrifice to internal resolution.

It has plenty of bells and whistles that we already know of.

 

[ShadowFox08]

Off topic:
I'm curious how people will compare Switch 2 with PlayStation 6 and Xbox next gen.

I'm thinking like switch vs x series s in power in some areas. A 10x power gap vs switch 2 in GPU and RAM bandwidth.. CPU probably won't get a 5x jump from PS5, but I dunno. And many of us think Switch 2 will be close to 50% of X Series S/PS5's CPU performance, but we'll see.

PS6 will come out no less than 3 years after PS5 Pro. I'm really hoping PS5 Pro is a Q4 2025 release, so they could use a 2nm node. if PS5 Pro ends up at 15-20 TFLOPs, I'm thinking PS5 will be 40 TFLOPs or a bit more. Going from 7nm to 2nm will be significant.

So close to 40 TFLOPs GPU minimum with better ray tracing and a.i. to punch above it's weight (better than PS5 Pro as well of course) in performance... Maybe a 3x faster/efficient CPU vs PS5, and 32GB of RAM?

it's going to be the shortest power jump for Sony for sure, but A.I. will help. 3-4x GPU speed in on paper is doable with node changes + newer architectural efficiencies.

Agreed. That's also why I left some nuance there because I know the Xenon CPU in the 360 also ran nearly 3x the frequency. One advantage though the Espresso had over Xeon was out-of-order executions, which 360 could not do. How this actually translates into gaming I could not explain well, though I do remember it was discussed by Microsoft and IBM when designing the chip. To save on costs though, they removed that feature.

There was also the power consumption aspect too. Wii U ran at most 40 watts, but realistically was 30-35 watts under most loads. Xbox 360 by contrast was over 150 watts, though by the time the Slim model came out, it was cut down in half or more. It's possible the Espresso could’ve ran at a higher frequency, though it doesn’t appear any of the 750 family of PPC processors ever ran at higher frequencies. At this time, Xbox was about brute force, and Wii U was more about efficiency, and Nintendo continues this route to this day.

Totally. It really was shame the Wii U's CPU was the largest bottleneck. They could have up clocked it at least. The OS was god awful slow too.

It's hard to imagine he effects of If things went differently. If Nintendo ditched the Wii U pad and got a much faster CPU and GPU, the Switch might not have existed and we could have gotten another home and dedicated handheld console , Or if it did, the jump in power wouldn't have been as large and it might have ended up as a Wii u 1.5 in power and made lower sales than current switch timeline. The Wii U's failure really blessing in disguise. Wouldn't go back and change anything other than the name and boosted CPU clocks I think.

 

[omniryu]

...

Moreover, according to Geekbench, that score has the Jaguars clocked at @2.34 GHz, which is over 40% higher than PS4 1.6Ghz. and about 10% over the PS4. So, 912/((1.6/2.34)*250) = 5.335 . Orin's CPU is over 5 time faster than PS4's.

Also, Orin has an IO chip, further freeing CPU resources.

Out of curiosity which is more

There’s going to be some nuance here, but by comparison, the “Espresso” CPU in the Wii U, despite being only a Tri-Core setup, wasn’t far behind the jaguar CPUs in PS4, and Xbone. The additional cores is really what helped push it above the Wii U CPU.

And this is despite the Espresso CPU was effectively a heavily modified IBM PowerPC 750 CPU from the late 90s. Core for Core, it was about the same as the Jag cores, but again, there are nuances here.

This is where someone else might have to step in for corrections, but I believe the Espresso CPU benefited greatly in Integer point calculations, whereas like most x86 CPUs, floating point calculations.

For all the Wii U’s faults, say Nintendo decided to go with an 8-core variant of the Wii CPU, it probably would’ve overall been more capable than the Jaguar CPU. But we can only imagine what a Frankenstein CPU using the PowerPC 750 would’ve been like.

Why did they chose a weak cpu? Everyone I mean?

 

[Aufhebung RPG]

I'm thinking like switch vs x series s in power in some areas. A 10x power gap vs switch 2 in GPU and RAM bandwidth.. CPU probably won't get a 5x jump from PS5, but I dunno. And many of us think Switch 2 will be close to 50% of X Series S/PS5's CPU performance, but we'll see.

Switch2 will have better textures and ray tracing than xss, which is really due to the tensor cores and larger RAM, and because of the presence of upgraders like dlss that also flatten out the resolution gap for example.So this power comparison is meaningless because it's never a 1:1 comparison.

 

[ShadowFox08]

Switch2 will have better textures and ray tracing than xss, which is really due to the tensor cores and larger RAM, and because of the presence of upgraders like dlss that also flatten out the resolution gap for example.So this power comparison is meaningless because it's never a 1:1 comparison.

I think you misread.

i wasn't comparing switch 2 vs XSS. They asked about the power gap between switch 2 vs PS6. And said it would probably be similar to Switch (current models) vs XSS. To to be clear, I'm just comparing TFLOPs numbers, and not taking newer architectural differences and efficiencies into account, which like you said, is very hard to compare, especially between AMD vs Nvidia, and architectures that are several generations newer to boot. 10x power gap in just raw on paper specs only.


Now that you mention it, conna be really interesting to see how Switch 2 vs XSS compares though on light CPU load games.

 

[PedroNavajas]

Out of curiosity which is more

Why did they chose a weak cpu? Everyone I mean?

Sony and MS? They didn't had many options for a 64-bit SoCs at the time. Nintendo? Because low power consumption and perfect backward compatibility with the Wii were main design priorities.

 

[albumin]

I think you misread.

i wasn't comparing switch 2 vs XSS. They asked about the power gap between switch 2 vs PS6. And said it would probably be similar to Switch (current models) vs XSS. To to be clear, I'm just comparing TFLOPs numbers, and not taking newer architectural differences and efficiencies into account, which like you said, is very hard to compare, especially between AMD vs Nvidia and architectures that are several generations to boot. 10x power gap in just raw on paper specs only.

This reminds me of something; what exactly is the difference between the TFLOP numbers in an ampere vs an RDNA GPU? I ask because
I'm wondering what ~4 TFLOPS in an ampere switch 2 would look like vs 4 TFLOPS in an XSS. If we assume for a second that the TFLOP count is the same, would that give more or less performance to a switch 2 and by how much? I'm aware that an exact answer would be tough to get but I'm a little clueless here so I at least want to get the gist lol

 

[Aufhebung RPG]

This reminds me of something; what exactly is the difference between the TFLOP numbers in an ampere vs an RDNA GPU? I ask because
I'm wondering what ~4 TFLOPS in an ampere switch 2 would look like vs 4 TFLOPS in an XSS. If we assume for a second that the TFLOP count is the same, would that give more or less performance to a switch 2 and by how much? I'm aware that an exact answer would be tough to get but I'm a little clueless here so I at least want to get the gist lol

Floating point comparisons make little sense, t239 is lower than xss in physical specs and shader cores, so even if floating point was identical, it would still be lower than xss at the upper limit, we're really moving away from traditional power comparisons but power still determines the upper limit of graphics quality.

The disadvantage of switch2 over xss will result in geometric quantities that won't be as high as xss.

So whether switch2 is 3.5tflops or 4tflops makes practically no difference, the impact is minimal.

 

[TLZ]

if PS5 Pro ends up at 15-20 TFLOPs

I think the leak says around 33tf if I'm not mistaken.

 

[Look over there]

That ~33 tflops is with the dual-issue capability of certain instructions.

 

[albumin]

Floating point comparisons make little sense, t239 is lower than xss in physical specs and shader cores, so even if floating point was identical, it would still be lower than xss at the upper limit, we're really moving away from traditional power comparisons but power still determines the upper limit of graphics quality.

The disadvantage of switch2 over xss will result in geometric quantities that won't be as high as xss.

So whether switch2 is 3.5tflops or 4tflops makes practically no difference, the impact is minimal.

understandable, thank you

 

[kasparov77]

the media said he was done

 

[Discostew]

Despite the Wii U having a stronger GPU due to better architectural advantages (176gflops vs 360's 240 TFLOPs) and more RAM (512 vs 2GB) some AAA parties were just not running as good as 360. The CPU was apparently easy to work and it was the clockspeeds. i don't know how much newer Wii U's CPU was.. But that 1.24 GHZ tri core CPU vs 3.2 GHz tri core CPU could not compete

Here is my understanding, so correct me if anything is wrong.

One of the big differences between the Xbox 360's CPU and the Wii U's CPU has to do with the instruction pipelines. For the 360, it's a long pipeline of around 21 stages. For the Wii U, it's around 4 major stages. My understanding is that each stage takes 1 cycle to complete, so for the 360's CPU, it takes 21 cycles to complete an instruction, whereas the Wii U's CPU takes 4 cycles. But, while as each instruction completes, there are other instructions in the pipeline, so the 360's CPU is working on 21 instructions in sequence all on different stages, while the Wii U's CPU is working on only 4 in sequence. At the same clock frequency, they would process instructions at the same rate, except the 360's CPU would have an initial delay of 21 cycles where the Wii U's CPU has an initial delay of 4 cycles. This delay comes from beginning at a new point in the code, like after branching, but more on that later. So again, at the same frequency, the Wii U would be ahead by 17 cycles. But, the 360's CPU is clocked at 3.2Ghz whereas the Wii U's CPU is clocked at 1.24Ghz, giving the advantage of the 360 of around 2.58x the number of instructions per second. So at the point the Wii U finishes its first instruction at 4 cycles, the 360 will still be in the process of its first instruction, but it will roughly be on stage 10. When another instruction on the Wii U complete one cycle after, the first instruction of the 360's CPU will be around stage 12-13. By the time the 360 finishes its first instruction, the Wii U will have completed around 5 instructions as opposed to 17 instructions if they were at the same clock frequency. The 360 quickly catches up, and by the time the Wii U completes 8 instructions, they are already matched, with the 360 further along in completing the next instruction.

Now, 8 instructions is a VERY short length for the Wii U to have any advantage, but this is with regards to the simplest of sequential execution. There are many other things to consider, like if instructions require the results of others, which tacks on delays. For an in-order execution CPU like the 360, this can be a problem, but not so much with an out-of-order execution CPU like the Wii U where it can process other instructions in the meantime. Then there's branching (such as "if-then-else"), where if the branch predictor guesses wrong, then the pipeline would basically be flushed, and the new instruction would have to make its way through the pipeline from the start, causing the delay again. For a 21-stage pipeline, that's pretty bad, but not as severe for a 4-stage pipeline.

So for all this, could the Wii U with its 1.24Ghz tri core beat out the 360's 3.2Ghz tri core? Possibly, but that becomes a question of whether the devs would optimize for it. My thoughts are that they had to designed their code a certain way with the 360, and that code, while passable on the Wii U, isn't really as effective. To improve on Wii U's code execution would require writing the code to Wii U's benefit rather than stick with what benefitted the 360, and at that time, devs simply wanted to get their games ported over. This doesn't take into account that most of the time, the original developers weren't making the ports to Wii U.

 

[Discostew]

Floating point comparisons make little sense, t239 is lower than xss in physical specs and shader cores, so even if floating point was identical, it would still be lower than xss at the upper limit, we're really moving away from traditional power comparisons but power still determines the upper limit of graphics quality.

The disadvantage of switch2 over xss will result in geometric quantities that won't be as high as xss.

So whether switch2 is 3.5tflops or 4tflops makes practically no difference, the impact is minimal.

Just a slight correction, but the Series S's GPU has 20 CUs, which equates to 1280 shader cores, whereas the T239's GPU has 12 SMs, which equates to 1536 CUDA cores. Series S's GPU clock is higher at 1.565Ghz than what the T239 would be at when docked. While the Series S has the advantage of raw GPU performance, the use of any RT or FSR cuts into that advantage where Switch 2 isn't affected nearly as much because of dedicated hardware.

 

[ShadowFox08]

This reminds me of something; what exactly is the difference between the TFLOP numbers in an ampere vs an RDNA GPU? I ask because
I'm wondering what ~4 TFLOPS in an ampere switch 2 would look like vs 4 TFLOPS in an XSS. If we assume for a second that the TFLOP count is the same, would that give more or less performance to a switch 2 and by how much? I'm aware that an exact answer would be tough to get but I'm a little clueless here so I at least want to get the gist lol

The RDNA GPU architecture in XLSS is supposed to be better in raster performance than Switch 2 per FLOP. Some p eople did some comparisons and found that AMD's was around the ball park of 25- 30% better or something like that per flop.

This is also why a lot of people are also saying steam deck would out perform switch 2 in handheld mode (on GPU at least) if Switch 2 was 1.6 TFLOPs like SD... SD would out perform it (not counting DLSS) .It would take 2 tflops on Ampere to be on par.

 

[Discostew]

The RDNA GPU architecture in XLSS is supposed to be better in raster performance than Switch 2 per FLOP. Some p eople did some comparisons and found that AMD's was around the ball park of 25- 30% better or something like that per flop.

This is also why a lot of people are also saying steam deck would out perform switch 2 in handheld mode (on GPU at least) if Switch 2 was 1.6 TFLOPs like SD... SD would out perform it (not counting DLSS) .It would take 2 tflops on Ampere to be on par.

But isn't that better raster performance through the use of Infinity Cache, which is only used by their discrete GPUs? None of the APUs have it, which includes those in the consoles and portable PCs.

 

[Dakhil]

I'm really hoping PS5 Pro is a Q4 2025 release, so they could use a 2nm node.

The PlayStation 5 Pro is rumoured to be released in holiday 2024.

But even in the scenario the PlayStation 5 Pro is planned to release during Q4 2025, TSMC's N2 process node is only planned to start high volume manufacturing (HVM) in 2H 2025.

And considering that TSMC said that TSMC's N3 process node is planned to start HVM in 2H 2022, but TSMC made an official announcement about starting HVM for TSMC's N3 process node on 29 December 2022, two days before the beginning of 2023, there's a possibility that HVM for TSMC's N2 process node won't practically start until early 2026.

So TSMC's N2 process node won't be ready by the time the PlayStation 5 Pro's APU is taped out. And I don't think TSMC's N2 process node is a realistic possibility to begin with.

 

[NineTailSage]

I'm thinking like switch vs x series s in power in some areas. A 10x power gap vs switch 2 in GPU and RAM bandwidth.. CPU probably won't get a 5x jump from PS5, but I dunno. And many of us think Switch 2 will be close to 50% of X Series S/PS5's CPU performance, but we'll see.

PS6 will come out no less than 3 years after PS5 Pro. I'm really hoping PS5 Pro is a Q4 2025 release, so they could use a 2nm node. if PS5 Pro ends up at 15-20 TFLOPs, I'm thinking PS5 will be 40 TFLOPs or a bit more. Going from 7nm to 2nm will be significant.

So close to 40 TFLOPs GPU minimum with better ray tracing and a.i. to punch above it's weight (better than PS5 Pro as well of course) in performance... Maybe a 3x faster/efficient CPU vs PS5, and 32GB of RAM?

it's going to be the shortest power jump for Sony for sure, but A.I. will help. 3-4x GPU speed in on paper is doable with node changes + newer architectural efficiencies.

This is the problem Sony will run into.
The PS5 Pro will only complicate even more how the PS6 will not look that impressive over the current systems (without going extreme to make a grande leap over PS5/PS5 Pro).

They can barely afford to make cutting-edge games for PS5, so what would the budget of these games look like on a 40Tflop system?
Not to mention after the success of the Switch it sounds like every manufacturer is interested in a portable variation in their portfolio and how would this stay close in spec to a stationary console...