• Hey everyone, staff have documented a list of banned content and subject matter that we feel are not consistent with site values, and don't make sense to host discussion of on Famiboards. This list (and the relevant reasoning per item) is viewable here.

StarTopic Future Nintendo Hardware & Technology Speculation & Discussion |ST| (Read the staff posts before commenting!)

Reader mode
I would disregard the A55s. They won't be a thing in the Switch successor's CPU. We can say this with 100% confidence, as Nvidia has a licence for the A78 series. The A78C is gaming-specific, so, it will be a variant of that, unless Nvidia gets a licence for another gaming-specific CPU series. The A78C is hexa-core or octa-core ONLY, according to ARM's official site - That is to say, it is not compatible with A55s at all (One would think they would state it otherwise, right?). I strongly suspect that it will be octa-core because the Switch would've had that if the A57 and A53 clusters could've been used at the same time. Apart from that, octa-core has been the industry standard on consoles for nearly a decade. It's also folly to presume that the dimensions, layout, performance envelope, etc. will follow exactly the same rules as the existing Switch, while the whole "vying for space and something's gotta give" reasoning falls apart when one realises that the A78AE variants carry automotive features that the A78C ones wouldn't. But Yeah, I'm confident that the A55s aren't going to happen. Still, if anybody has a coherent reasoning on why a company which makes video game consoles wouldn't use gaming-centric parts for their video game system, a reasoning which doesn't smack of "Because Nintendo" or "gotta be pessimistic just because....", then I'm all eyes, ears, and tea with the slowest of sips.
 
lemonfresh said:
So when is the next event that we might get more Info on the chips
Click to expand...
Orin AGX goes on sale at the beginning of next year, white paper, benchmarks, a ton of info will come out then, and that will give us a good idea on how Dane could perform. We might also hear about tape out of Dane in the next couple months, but that is for insiders to share, Jetson products people get to just buy.
 
ShadowFox08 said:
Random thought. The more I think if it... The more I think steam deck is going to be lackluster/luke warm at best. This thing won't be released to the majority of the masses until mid 2022. The specs aren't going to change.

AMD themselves likely already have the tech to make a handheld console out by the end of the year they blows the stream deck in efficiency and price if they wanted to.

And considering that switch 2 SOC isn't taped out yet, I'm still holding hope we aren't getting an 8nm chip for holiday 2022-Q1 2023 and could get 5-7nm instead. Though even 7nm feels outdated for 2022. Perhaps Nitnendo is willing to push the date to early 2023 to bank on the chip shortage getting significantly better by then and think they will have enough 6nm or 4nm chips from Samsung (or TSMC).
Click to expand...
Sorry for the late reply but your post reminded of the BlackBerry PlayBook tablet.

Announced in October 2010 with 1GB of RAM, half a year after the first iPad released with 256MB of RAM so the specs were really good for the time. But by the time it released in April 2011 the iPad 2 was already on the market and had 512MB.
The following year in March 2012 the New iPad (iPad 3) released with 1GB of RAM too and closed the gap.

So I think announcing devices too much time in advance can allow the competition to catch up or even surpass a device that’s not even in the market.
 
0
ShadowFox08 said:
Fair enough. Since we are ruling out that the Dane is based off the Orion NX because its too big (like Xavier NX), I think the Orion S can still exist now, despite it not showing up this week's presentation. We got thrown off thinking the Orion S just consolidated and merged into the NX.. But Maybe Nvidia really left this out on purpose because they didn't want us putting 2+2 together, until the time was right.. They still have more to reveal by the end of the year or q1 2022, don't they 🤔? Nintendo would be the ones officially announcing the switch 2 of course.

I think we are all in the ball park on what Dane's power could be. We've discussed this before.. The 1.6 TFLOPs and 6-8 core A78 CPU using 128 bus width of lpddr5 doesn't seem impossible. AMD and Steck are already matching this on 7nm on a form factor slightly bigger than switch, and with the same 15 watts usage.
Click to expand...

With how much Nvidia praises the Switch and how fast its selling, Dane will be much more custom to Nintendo's wants than the TX1 ever was. Switch is a massive success and all of the guess work and uncertainty is gone from where things may have been coming off of the WiiU...

Blomqvist said:
Orin NX which is a binned version of Orin X (for yields on 8N are low) will hardly be a SOC for the next Nintendo system because it is huge and hence expensive. I recall dismissing Xavier NX as a possibility while it is smaller than this Orin chip. Orin S is not Orin NX, it will be a small and cheap to produce die with good yields and the ability to hit 5W TDP. And it will still be probably binned to make Jetson Nano Next. This thing is probably coming in Q1 2023 and everyone will be disappointed because it will be the lowest end of Nvidias offering.
Click to expand...
There are a few things we can be certain on if nothing else so we don't keep going in circles, Nintendo sees DLSS as a major desire for this new hardware and we have done enough mathematical equations around here to understand that DLSS itself requires a minimum specified performance level to be effective.

Skittzo said:
Yeah that's been the speculation.

Die size is the more interesting variable now IMO, it could go up if this is going to be a pricey product.
Click to expand...

I've said this before but the $350 OLED is a temperature check to see if people will buy a premium priced Switch product.
This next model I'm fully expecting to be $400-$450 and because the specs on paper will look drastic in comparison vs the OG Switch model, it will easily be worth the price increase.
 
0
Z0m3le said:
hopefully we don't have to wait 9 months for the next bit of news, but yes hot chips is always good.
Click to expand...
Funny enough, Nvidia mentioned that the Orin NX devkits will become available on Q4 2022.

And I think there's a fair amount of information about Orin that Nvidia hasn't disclosed. A couple of examples are:
  • How does the RT cores in Orin compare to the RT cores on consumer Ampere and Turing GPUs?
  • How does the Tensor cores in Orin compare to the Tensor cores on consumer Ampere and Turing GPUs, as well as the Tensor cores on Xavier?
I don't know if software benchmarks and/or tests can reveal all the answers to the questions about Orin. There's a possibility the answers to some questions about Orin require seeing a high quality picture of the die shot of the Jetson AGX Orin.

Looking at the Jetson Download Center, the data sheets are probably the closest equivalent to white papers. And there's a possibility that Nvidia's saving the amount of information that Nvidia hasn't disclosed about Orin for Hot Chips 34 on August 2022.
 
Dakhil said:
Funny enough, Nvidia mentioned that the Orin NX devkits will become available on Q4 2022.

And I think there's a fair amount of information about Orin that Nvidia hasn't disclosed. A couple of examples are:
  • How does the RT cores in Orin compare to the RT cores on consumer Ampere and Turing GPUs?
  • How does the Tensor cores in Orin compare to the Tensor cores on consumer Ampere and Turing GPUs, as well as the Tensor cores on Xavier?
I don't know if software benchmarks and/or tests can reveal all the answers to the questions about Orin. There's a possibility the answers to some questions about Orin require seeing a high quality picture of the die shot of the Jetson AGX Orin.

Looking at the Jetson Download Center, the data sheets are probably the closest equivalent to white papers. And there's a possibility that Nvidia's saving the amount of information that Nvidia hasn't disclosed about Orin for Hot Chips 34 on August 2022.
Click to expand...
Orin AGX Jetson boards are available at the beginning of the year, binned Orin chips rebranded to Orin NX will release at the end of next year.
 
Z0m3le said:
Orin AGX Jetson boards are available at the beginning of the year, binned Orin chips rebranded to Orin NX will release at the end of next year.
Click to expand...
Wonder if you could software lock the CPU/GPU Config of Orin AGX to meet the level of Orin NX (Mainly setting it to 8SMs)
 
the hardest part will be people who acquire an AGX system to do relevant testing. people who get their hands on these boards in quite few and far inbetween
 
Z0m3le said:
Orin AGX Jetson boards are available at the beginning of the year, binned Orin chips rebranded to Orin NX will release at the end of next year.
Click to expand...
I'm aware that Jetson AGX Orin devkits will be available by Q1 2022.

But I don't know if software benchmarks and/or tests can reveal all the answers to the questions about Orin with regards to information about Orin Nvidia hasn't disclosed yet. There's a possibility the answers to some questions about Orin require seeing a high quality picture of the die shot of the Jetson AGX Orin.
 
note : concerning the backward compatibility. Différents ways to get it but far from impossible the question is which method
 
0
Also, I'm trying to figure out how capable 4 RT Cores (Assuming they're at the performance level of Ampere) would be relative to the Series S's 20 Ray Accelerators.

Now, If anyone can help, I've done one set of number crunching based on DF's 6700XT RT performance page comparing against the RTX 2080TI Yes, I corrected it by subtracting 10% from the 2080Ti's numbers, and got 30%, so the Turing RT cores are 30% faster than the RDNA2 Ray Accelerators, so that would mean that Ampere is 130% faster in RT per-core?

But something just feels off about that number (The reason I did Turing VS RDNA2 then doubled the number is because devs haven't had a chance to really pull the full RT Acceleration ability out of the Ampere RT cores, thus why they perform similarly to the Turing ones for a lot of games despite their doubling).

Does anyone have better data to compare and draw a conclusion from?

Also note, maybe not using the RTX 3050 and 3050Ti as the 4GB Frame buffer in those seemingly is limiting RT Performance hard as RT likes memory.
 
Last edited:

Alovon11 said:
Also, I'm trying to figure out how capable 4 RT Cores (Assuming they're at the performance level of Ampere) would be relative to the Series S's 20 Ray Accelerators.

Now, If anyone can help, I've done one set of number crunching based on DF's 6700XT RT performance page comparing against the RTX 2080TI Yes, I corrected it by subtracting 10% from the 2080Ti's numbers, and got 30%, so the Turing RT cores are 30% faster than the RDNA2 Ray Accelerators, so that would mean that Ampere is 130% faster in RT per-core?

But something just feels off about that number (The reason I did Turing VS RDNA2 then doubled the number is because devs haven't had a chance to really pull the full RT Acceleration ability out of the Ampere RT cores, thus why they perform similarly to the Turing ones for a lot of games despite their doubling).

Does anyone have better data to compare and draw a conclusion from?

Also note, maybe not using the RTX 3050 and 3050Ti as the 4GB Frame buffer in those seemingly is limiting RT Performance hard as RT likes memory.
Click to expand...

Sad thing is Ampere in its current state is over engineered and not balanced to fully execute what potential it really has or is capable of.
So less Tensor cores and RT cores than Turing might just streamline the workload enough to keep the GPU consistently fed for the given memory bandwidth as well (which is another reason Nvidia keeps putting out these alternative cards with increased RAM).

Infinity Cache definitely caught Nvidia off guard and was the equivalent of AMD saying they only have a V6 under the hood, but keeping secret that it's also turbo charged as well. Funny thing is Infinity Cache is pretty much an evolution on eDRAM used in both the 360 and WiiU by AMD in which the WiiU had 32MB of eDRAM at that time which was extremely expensive.
 
NineTailSage said:
Sad thing is Ampere in its current state is over engineered and not balanced to fully execute what potential it really has or is capable of.
So less Tensor cores and RT cores than Turing might just streamline the workload enough to keep the GPU consistently fed for the given memory bandwidth as well (which is another reason Nvidia keeps putting out these alternative cards with increased RAM).

Infinity Cache definitely caught Nvidia off guard and was the equivalent of AMD saying they only have a V6 under the hood, but keeping secret that it's also turbo charged as well. Funny thing is Infinity Cache is pretty much an evolution on eDRAM used in both the 360 and WiiU by AMD in which the WiiU had 32MB of eDRAM at that time which was extremely expensive.
Click to expand...
Still doesn't help their Ray Accelerators be much good though versus NVIDIA's.

Not having BVH Traversal really limits them versus the RT cores.

(The 6600 Drops over 40% where its effective performance should be...down below a laptop RTX 2060 in Port Royal)

I feel that Sony and Msoft are pulling the stops out to make everything work (Wouldn't be surprised if a lot of the CPU power in the Next-Gen systems is just for BVH in RT applications for the future)

But that would give Nintendo and NVIDIA a gap to slide in with the BVH-Accelerated RT cores in Dane freeing up CPU resources.
 
0
Z0m3le said:
Got a number? I'm thinking 10-15 watts for the entire system when docked... Switch OG was ~13 watts for the entire system when not charging the battery.
Click to expand...

General system design question, vut what is the reason they choose a 15W max over e.g. 20, or 25 or anything else?

I understand in handheld mode wanting to keep it low. But when docked, drawing power from the wall and being able to e.g. place more heat mitigation within the dock itself, why cap at such a low number?
 
Okay, so how many RT cores would it take to match the 20 Ray Accelerators in the Series S

If Ampere RT cores were 180% to 210% faster per-Core than the RA's?

unknown.png
 
0
Terrell said:
Well, that's a bit of an unknowable variable. It's right to assume it will be as lightweight as possible like Horizon, but whether or not it's the exact same? Who knows.
Click to expand...
I think it will be an improved form of the current Switch OS

davec00ke said:
$400-$450 is DOA
Click to expand...

Alovon11 said:
People said the same about the 350$ SWOLED.

And the SWOLED is selling like hotcakes
Click to expand...

My theory is that Switch OLED is a trial run for Switch 2, which will use many of the same improved components 64GB storage, OLED screen. That way they can segue into Switch 2 manufacturing with a smaller step up and get some economies of scale on certain components that could be shared.

The $350 pricing IMHO for OLED is also to prep the groundwork for either a $399 Switch 2, or for Switch 2 to occupy the $349 slot. I'm a pessimist so I'm assuming Switch 2 will be $399.
 
0
uncleoptimus said:
General system design question, vut what is the reason they choose a 15W max over e.g. 20, or 25 or anything else?

I understand in handheld mode wanting to keep it low. But when docked, drawing power from the wall and being able to e.g. place more heat mitigation within the dock itself, why cap at such a low number?
Click to expand...
It needs to be immediately undockable at any point, so you can't be generating too much heat even when docked.

Also in a device that size there's always the possibility of it simply breaking/failing due to excessive heat.
 
NineTailSage said:
Sad thing is Ampere in its current state is over engineered and not balanced to fully execute what potential it really has or is capable of.
So less Tensor cores and RT cores than Turing might just streamline the workload enough to keep the GPU consistently fed for the given memory bandwidth as well (which is another reason Nvidia keeps putting out these alternative cards with increased RAM).

Infinity Cache definitely caught Nvidia off guard and was the equivalent of AMD saying they only have a V6 under the hood, but keeping secret that it's also turbo charged as well. Funny thing is Infinity Cache is pretty much an evolution on eDRAM used in both the 360 and WiiU by AMD in which the WiiU had 32MB of eDRAM at that time which was extremely expensive.
Click to expand...
I remember Kopite saying that Ampere Super line will bridge Ampere and Lovelace. given the changes in cache sizes for Orin, I wonder if the Super line will see similar changes
 
ILikeFeet said:
I remember Kopite saying that Ampere Super line will bridge Ampere and Lovelace. given the changes in cache sizes for Orin, I wonder if the Super line will see similar changes
Click to expand...
I doubt it since the consumer Ampere Super GPUs seem to be using the same dies as the regular consumer Ampere GPUs. I think kopite7kimi's comment about the consumer Ampere Super GPUs being a smooth transition to consumer Lovelace GPUs is in reference to the power consumption, considering that the RTX 3090 Ti is rumoured to have a TDP of up to 450 W.
 
0
Dakhil said:
I'm not sure that the types of benchmarks and tests ETA PRIME does, at least when looking at his videos on the devkits for the Jetson Nano, Jetson AGX Xavier, and Jetson Xavier NX, are going to answer any questions in regards to the performance of the Tensor cores and the RT cores.
Click to expand...
Well, we could just ask him to run Port Royal on it and see how it performs.

Either that or some other RT benchmark that can be used to compare against AMD.
 
0
davec00ke said:
$400-$450 is DOA
Click to expand...
Alovon11 said:
People said the same about the 350$ SWOLED.

And the SWOLED is selling like hotcakes
Click to expand...
same thing was said about original switch lol


So I think Nintendo has wanted to raise price for a while now. They made bank on Wii at $250

The wii U was raised to $300 and $350 for the DX version.. Wii U had a value perception problem obviously it sold poorly. But the DX version was the one that sold most.

They took a step back and priced the switch at $300 to ensure people saw value in that. They did. And now they are walking back up to $350
I wouldn't be surprised if Dane switch was $399 ... I don't see more than that. I know people talk of price drops ... and yeah ... that COULD happen... But I don't think they will with OLED. at least not soon. maybe after DANE launches... lite $200, OLED $300, DANE $400?

My whole point is I think they're trying to climb the price ladder.
 
0
Skittzo said:
It needs to be immediately undockable at any point, so you can't be generating too much heat even when docked.

Also in a device that size there's always the possibility of it simply breaking/failing due to excessive heat.
Click to expand...
Very true. It’s also that you’re just not gonna see a huge gap between docked and handheld. You need the performance gap to be small enough that games can reasonably run in both modes and once you start optimizing the unit for power draw you’re building in certain decisions that constrain docked mode no matter what.

Even if Nintendo goes with an active cooling solution in the dock to open up the heat budget it seems highly likely that docked power draw is still going to be reasonably close to undocked
 
Summary of the very few info we have on next switch hardware:

1) Ampere architecture
2) 8nm for node process
3) New SoC (not Tegra X1+)
4) Will have DLSS support via featuring Tensor Cores
5) (Very) Limited Ray Tracing support, few RT cores on SoC
6) Its still an hybrid console, power output limited to max 15W when docked
7) It will have a (few?) number of (third party only?) exclusive titles.

What we don’t know:
1) Console form factor, dimensions, weight, screen size, resolution and technology.
2) New joycons or will be the same
3) SoC specifics (GPU SMs, CPU core and which cores, clock, bus bandwidth)
4) RAM: amount, technology, speed, etc
5) Dock, the same as OLED model or different?
 
oldpuck said:
Very true. It’s also that you’re just not gonna see a huge gap between docked and handheld. You need the performance gap to be small enough that games can reasonably run in both modes and once you start optimizing the unit for power draw you’re building in certain decisions that constrain docked mode no matter what.

Even if Nintendo goes with an active cooling solution in the dock to open up the heat budget it seems highly likely that docked power draw is still going to be reasonably close to undocked
Click to expand...
Well, the thing there is that the gap can be wider than it is on the current switch due to DLSS.

Like the current Switch, it would likely have the same CPU Frequencies in Portable and Docked, but the GPU can be scaled more easily, and the range on the current Switch is halving it so that likely is that limit.

So if Dane is 2TFLOPs docked, Portable mode would have a bottom-end of 1TFLOP.
Before DLSS.
 
0
LoneRanger said:
Summary of the very few info we have on next switch hardware:

1) Ampere architecture
2) 8nm for node process
3) New SoC (not Tegra X1+)
4) Will have DLSS support via featuring Tensor Cores
5) (Very) Limited Ray Tracing support, few RT cores on SoC
6) Its still an hybrid console, power output limited to max 15W when docked
7) It will have a (few?) number of (third party only) exclusive titles.

What we don’t know:
1) Console form factor, dimensions, weight, screen size, resolution and technology.
2) New joycons or will be the same
3) SoC specifics (GPU SMs, CPU core and which cores, bus bandwidth)
4) RAM: amount, technology, speed, etc
5) Dock, the same as OLED model or different?
Click to expand...
I would ammend #5 on the What we know section to be "RT Support at unknown value"

Again, NVIDIA's RT cores are over twice as fast as the Ray accelerators in the AMD GPUs right now (and that is only Ampere, we don't know if they doubled it again and that is why they halved the RT Core amount in Orin).

So 4RT cores in an 8SM config, would be able to do actual RT work, maybe not RT Reflections (Still trying to crunch the numbers on how exactly they are versus RDNA2's Ray Accelerators to figure out where 4 RT cores would compare to the 20 Ray Accelerators in the Series S), But RTGI via RTXGI (A solution that can be scaled to run on the OG Xbox One), Lumen, RT Shadows, RTAO, RT Audio are certainly possible to be accelerated.

The question of it can do RT Reflections though depends on where exactly it is versus the Series S RT Acceleration-wise as the main reason DOOM Eternal doesn't have RT Reflections seems to be a memory issue which Dane can side-step using DLSS.

So, if 4 Ampere RT Cores actually can get in the ballpark of the Series S's 20 Ray Accelerators (Still trying to figure that out) then RT Reflections can be a thing Dane could do as we saw in Watch Dogs Legion.
 
0
Glom said:
But any RT support is better than current situation, right?
Click to expand...
We don't know what kind of RT support the Dane SoC will have, but it will likely have some support now.
The big news as I understand it this week is it was revealedthe Orin design has RT cores (1 per 2 SM) so maybe we'll get 2-3 RT cores in the final Nintendo Switch 2 SoC, which isn't a lot, but it is better than 0 we were expecting back earlier this year.
 
LoneRanger said:
What we don’t know:
1) Console form factor, dimensions, weight, screen size, resolution and technology.
2) New joycons or will be the same
3) SoC specifics (GPU SMs, CPU core and which cores, clock, bus bandwidth)
4) RAM: amount, technology, speed, etc
5) Dock, the same as OLED model or different?
Click to expand...
I think the RAM technology being used for the DLSS model* is known, considering that the Jetson AGX Orin and the Jetson Orin NX use LPDDR5. And considering the Nintendo Switch and the Jetson Tegra X1 use 4 GB of LPDDR4, I think expecting the DLSS model* to also use LPDDR5 is a reasonable expectation.
 
0
Dekuman said:
We don't know what kind of RT support the Dane SoC will have, but it will likely have some support now.
The big news as I understand it this week is it was revealedthe Orin design has RT cores (1 per 2 SM) so maybe we'll get 2-3 RT cores in the final Nintendo Switch 2 SoC, which isn't a lot, but it is better than 0 we were expecting back earlier this year.
Click to expand...
Again 8SMs is the most likely config cost-wise at this point due to Orin NX being right there as Z0m3le has pointed out on a few occasions.

And while 4RT cores may not sound like a lot there are some things to consider.
  • NVIDIA's RT Cores are far ahead of the Ray Accelerators from AMD.
    • Assuming worst-case for NVIDIA vs AMD RT (Which likely isn't the case as Turing RT cores are stronger than Ray Accelerators) You'd only need 10 RT cores to Match the Series S's 20 Ray Accelerators.
      • And like I mentioned earlier, Turing is actually better at RT than RDNA2 in most benchmarks/tests, so that number is actually smaller to match the Series S's 20.
        • Some number configurations actually put 4RT cores as matching the 20RA's but still trying to number-crunch to verify that, but still it may be close.
  • RT has gotten far more scalable
    • RTXGI (NVIDIA's own GI solution) is scalable enough to run in software on the OG Xbox One, and considering Dane when docked will be stronger than the OG PS4, those RT cores can be used to accelerate RTXGI.
      • Same thing with the RT cores accelerating things that are in software like Lumen or CryTek's RT.
    • RT Shadows are able to be run on smartphone ARM SoC GPUs in software now (With crap Mali drivers), so they can be accelerated perfectly fine to run on Switch on a low number of RT cores.
    • And even if not RT shadows, there is RTAO, RT Audio, and other usages of RT that can be used (Maybe even acceleration of SSRT effects like SSR, SSRTGI.etc)
  • DLSS
    • Back to the Series S point, even assuming Dane has half the theoretical RT performance of the Series S, Dane can easily run at half the internal resolution or even a quarter of the resolution and get that performance back and then some and jump to match or surpass the Series S in output (outside of the resolution of the RT effect suffering, but for RTGI, RTAO, and RT Shadows that shouldn't be too noticeable)
 
Reader mode
You must log in or register to reply here.
Please read this staff post before posting.

Furthermore, according to this follow-up post, all off-topic chat will be moderated.
Last edited:


Back
Top Bottom