• 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!)

It does have more cache than standard ampere, and better clock gating. Its probably a stretch to call these Ada features.
That's unknown, since the Nvidia leaks mention that Drake's GPU either has 1 MB or 4 MB of L2 cache, with 4 MB of L2 cache only being mentioned once outside of the NVN2 folder if I've remembered correctly. (Therefore, Drake's GPU's likely to have 1 MB of L2 cache, which I believe is less cache than the amount of L2 cache on consumer Ampere GPUs.)
 
That's unknown, since the Nvidia leaks mention that Drake's GPU either has 1 MB or 4 MB of L2 cache, with 4 MB of L2 cache only being mentioned once outside of the NVN2 folder if I've remembered correctly. (Therefore, Drake's GPU's likely to have 1 MB of L2 cache, which I believe is less cache than the amount of L2 cache on consumer Ampere GPUs.)
Oh that's a bummer. I suppose it will have Bandwidth issues then.
 
Oh that's a bummer. I suppose it will have Bandwidth issues then.
I will add an addendum to this that despite having less L2 than expected, it does have L3 (Techinally)
By being a modern Tegra SoC, the GPU has access to the CPU L2 and L3 cache, and we know CPU Cache will be hopefully in abundance as the CPU will most likely be A78C, which the C pretty much means Cache.
So while the GPU itself may have less L2 cache exclusive to itself, it has theoretical access to a fair bit more Cache than an RTX 3050 Laptop would.
 
It makes no sense to remove the raytracing units and the matrix units and offer them as chiplet designs, because the way they function is having it right there readily with the shader units.

Splitting them up and which would occur bringing in latency that is completely negligible by being coupled to the shader itself, which has other issues and implications in the design of it. And I’ve seen this idea pop up before of a ray tracing coprocessor or ray tracing chiplet, or a matrix core chiplet, and it makes no sense because the latency would be above what it should be.

High latency would be bad... in consumer cards. However, my primary job is compute farms, where throughput of matrix ops is much more important than their latency which is why, in fact, we did, at one point have pure matrix acceleration cards in the farm. We've moved entirely to NVidia GPU clusters since, and in fact are likely to retire all the Tesla cards for Ampere in the near future, but our need for tensor performance is way out ahead of our need for the rest of the GPU in CUDA operations.

Nvidia has unified their architectures for the consumer and the datacenter, with Nvidia offering DLSS branded tech as a way to use all that silicon that datacenter demands that consumers traditionally haven't had a use for. Medium term, however, I doubt this will scale. Eventually the demands for AI on the consumer end are going to diverge from the demands of the data center. There are huge advantages to having both pieces of hardware share designs, and if the latency problem can be cracked then being able to scale tensor core numbers independently of TPCs, the way we can scale TPCs independently of GPCs, will allow those advantages to last longer.

I didn't expect that AMD would be able to crack it, but I was hopeful. In the case of Ray Tracing, that is less my area of knowledge - potentially if AMD can scale RT performance independently of shader performance, it opens up a new market in render farms with similar needs. The advantage to us being that this would, in turn, strongly incentivize AMD to increase their RT performance for consumers.

Is it really like Ampere? if so, in what way?
The Matrix stuff, fwiw, is only an instruction addition. They have the Matrix in CDNA, the RDNA has none of that. So I’m not sure if that’s what you mean.
If you mean the FP32, it is doing a dual issue FP32, AMD likely leveraged this to add the instruction.

But RDNA3 doesn’t have actual Matrix units in-hardware. Afaik.
The big difference between the two compute unit designs is that an Nvidia SM is broken up into four partitions, of two lanes each, where AMD breaks them into two partitions of two lanes each.

One lane is FP32 logic units. The second lane is integer logic units. The partition also has dedicated hardware for FP64 and transcendental functions (trig accelerator), the register file, a scheduler and scalar dispatch unit, and a shared L0 cache. Nvidia adds a pair of tensor cores per partition.

Ampere changed this design only slightly, having the number of tensor cores, doubling the L0 cache, and changing the dedicated integer lane and ALUs to support mixed FP32/INT workloads. This could double maximum FP32 performance with no additional hardware, though running at full speed sacrifices integer perf.

RDNA 3 did indeed introduce matrix units (though they are quite different from tensor cores), and placed them one per partition, same as Nvidia. They also converted the integer lane to mixed INT/FP32.
 
Does T239 have any Ada Lovelace features?

Unknown, except for AV1 encode support. (There's a possibility Drake's Optical Flow Accelerators (OFA) performs very similarly to Ada Lovelace's OFA, but that's not officially known. But what's known is that Drake shares the same OFA as Orin.)
GA10F is described as the only Ampere GPU to support FLCG (a clock gating feature), and it seems like all Ada and Hopper GPUs support it.


Isn't AV1 encode supported on Orin? The documentation says it is.
Orin has the 8th gen NVENC hardware, yes. In the Linux driver, 8 gen NVENC is listed as a "t23x" class feature. So I suspect Drake will too.

It does have more cache than standard ampere, and better clock gating. Its probably a stretch to call these Ada features.

Tensor cores and RT cores are clearly Ampere generation.

Edit: corrected by Dakhil
That's unknown, since the Nvidia leaks mention that Drake's GPU either has 1 MB or 4 MB of L2 cache, with 4 MB of L2 cache only being mentioned once outside of the NVN2 folder if I've remembered correctly. (Therefore, Drake's GPU's likely to have 1 MB of L2 cache, which I believe is less cache than the amount of L2 cache on consumer Ampere GPUs.)

I will add an addendum to this that despite having less L2 than expected, it does have L3 (Techinally)
By being a modern Tegra SoC, the GPU has access to the CPU L2 and L3 cache, and we know CPU Cache will be hopefully in abundance as the CPU will most likely be A78C, which the C pretty much means Cache.
So while the GPU itself may have less L2 cache exclusive to itself, it has theoretical access to a fair bit more Cache than an RTX 3050 Laptop would.

Oh that's a bummer. I suppose it will have Bandwidth issues then.
It's actually not that bad. The Ampere white paper lists a massive cache advantage for Ampere, but that was for their very first AI focused workstation GPU. In practice, Ampere cards had much much less L1 cache. 1MB for 12 SMs isn't the best ratio for an Ampere card, but it's not actually the worst either.

Of course if we get Orin's fatter cache, I will not complain :)
 
Yeah it must have been click bait.

The Ryzen 5700g has a really strong CPU. At least when compared to jaguar. The APU itself is a desktop processor from 2021, while Vega 8 is from 2018. 3.8 GHz base clocks with 8 CPU cores and 16 threads on a 7nm tsmc node with 65 watts power draw. Zen 3 CPU. Shiit thats newer and faster than PS5/X series X.

The GPU might actually be 2 tflops though, and released in a January 2021. So an updated Vega 8?


Quoting you again.. A 2 tflop GPU (more advanced than PS4/xbone) with a CPU that is more powerful than current gen Xbox series X/PS5.. But it ends up running at 1080p and 30 fps with lumens on or 1080p 98 fps without (but with high textures). Soo it's GPU bottlenecked to not reach 60fps on Lumens.

while Unreal Engine 5 is supported on PS4 and xbone, 30 fps is likely the limit then on Fortnite if supported with that engine and Lumen settings.
I wonder how much impact there may be from it being potentially under-fed.
In that particular video, the creator doesn't mention the RAM speed, so I can only list some possibilities:
  • 5700G officially supports up to DDR4-3200 (as JEDEC profiles stop there for DDR4). Assuming two sticks/'dual channel'/128-bit, that's 51.2 GB/s. ~25.6 GB/s per tflop, which seems low for Vega, and we aren't accounting for how much the CPU is using.
  • Worst case: the creator forgot to enable XMP profile, so the ram stick(s) fall back to some lower speed profile for compatibility, probably 2133 MT/s. So that's 1/3 less bandwidth, or ~34.13 GB/s. Worse worst case would be single stick for half that, of course.
  • Great case: the creator went with the typical 'sweet spot' for Zen 3 (at least as far as keeping the memory controller:ram speed ratio 1:1) and used a pair of DDR4-3600. That's 1/8 more bandwidth, or 57.6 GB/s
  • Best/lucky case (for 1:1 ratio): 4000 MT/s sticks, or 64 GB/s. Bandwidth per tflop is still under average for GCN?
 
I wonder how much impact there may be from it being potentially under-fed.
In that particular video, the creator doesn't mention the RAM speed, so I can only list some possibilities:
  • 5700G officially supports up to DDR4-3200 (as JEDEC profiles stop there for DDR4). Assuming two sticks/'dual channel'/128-bit, that's 51.2 GB/s. ~25.6 GB/s per tflop, which seems low for Vega, and we aren't accounting for how much the CPU is using.
  • Worst case: the creator forgot to enable XMP profile, so the ram stick(s) fall back to some lower speed profile for compatibility, probably 2133 MT/s. So that's 1/3 less bandwidth, or ~34.13 GB/s. Worse worst case would be single stick for half that, of course.
  • Great case: the creator went with the typical 'sweet spot' for Zen 3 (at least as far as keeping the memory controller:ram speed ratio 1:1) and used a pair of DDR4-3600. That's 1/8 more bandwidth, or 57.6 GB/s
  • Best/lucky case (for 1:1 ratio): 4000 MT/s sticks, or 64 GB/s. Bandwidth per tflop is still under average for GCN?
since randomgaminginhd is a regular hardware reviewer for old stuff, I'd assume he would have settings set appropriately
 
I think Nintendo would put preference towards their next generation console releasing day and date with a strong launch title. I’m sure they don’t mind missing this “golden week” for a chance to repeat 2017 by having Zelda and Switch 2 coming out the gate side-by-side. New hardware + big launch title = more sales. And it doesn’t get much bigger than a new mainline Zelda.
This is what they're doing imo to try and replicate the exact scenario of Switch and BotW. All the console needs is TotK as it's launch title and it will sell 5 million units in a couple of months Worldwide imo (if they can build that many). The upsides to releasing it in March are blown away when compared to launching it with Zelda in May. It also gives them another 60 days to manufacture another 2 million consoles.
 
This is what they're doing imo to try and replicate the exact scenario of Switch and BotW. All the console needs is TotK as it's launch title and it will sell 5 million units in a couple of months Worldwide imo (if they can build that many). The upsides to releasing it in March are blown away when compared to launching it with Zelda in May. It also gives them another 60 days to manufacture another 2 million consoles.

Yep! Having a good launch title is extremely important, and if history has proven anything, Zelda gets the ball rolling. Twilight Princess on Wii, BotW on Switch, and now potentially TotK on Switch 2.
 
0
RDNA 3 did indeed introduce matrix units (though they are quite different from tensor cores), and placed them one per partition, same as Nvidia. They also converted the integer lane to mixed INT/FP32.
But it doesn’t have matrix units? that is a misconception. All they did was introduced more instructions, they did not include any physically new hardware. The only GPU that they have they have with actual matrix unit hardware is cDNA. Which isn’t for consumers.


High latency would be bad... in consumer cards. However, my primary job is compute farms, where throughput of matrix ops is much more important than their latency which is why, in fact, we did, at one point have pure matrix acceleration cards in the farm.
I’m confused as to why this was brought up then considering the context of the video was focusing on the gaming performance increase gen-on-gen and making comparisons, the idea of a chiplet design where it has the ray tracing on a separate chiplet would be a nightmare, because of the latency that it would incurred. if it’s for something not game related then, so be it that is not really my scope and I don’t really know much about that so I’ll take your word for it but that shouldn’t come as a surprise as that’s parallelized work and latency isn’t really an issue there, even for gaming it isn’t really an issue that is that bad. It’s an issue for the CPUs more so.

But with respect to the video, and the comment about what would be a chiplet RT mentioned in it, Latency for ray tracing would be detrimental to the GPU, not beneficial.


They can decouple shader clocks from the front end, but there seems to be so many bugs with it…. N31 weeps in the corner…

It's actually not that bad. The Ampere white paper lists a massive cache advantage for Ampere, but that was for their very first AI focused workstation GPU. In practice, Ampere cards had much much less L1 cache. 1MB for 12 SMs isn't the best ratio for an Ampere card, but it's not actually the worst either.

Of course if we get Orin's fatter cache, I will not complain
I don’t really think that’s the issue and I have an issue(?) with it and that I find it weird, the ampere white papers it mentions that they actually couple 512 kB of level 2 cache to each 32-bit memory controller. Or it’s equivalent of 32 bit memory controller pairing. Every GPU follows the same schematic except Drake. drake should have 2 MB of level 2 cache following the same set up as the desktop, but it does not. It is the only GPU that is different from the rest. Even ORIN which has a 256-bit interface has 4MB.

Drake has 8x 16-bit memory controllers. 1MB L2

Orin has 16x 16-bit memory controllers. 4MB L2



Even if ORIN doesn’t use the 32-bit, it uses 16-bit and has 256kB coupled to each MC, giving 4096 kB of L2. If you want to look at it like that…


And in case one is curious on this, you can check literally every GA10x chip and count the memory controllers (it’s the bus width bit divided by 32) and multiply that by 512 and you get the kB of L2 cache it has.


GA102 has 12 MCs, has 6144kB
GA104 has 8MCs, has 4096kB. This one is similar in configuration to ORIN, and also has the 4MB of L2 cache.

But Drake has less cache than any other ampere based silicon relative to its MCs. I don’t think 4MB makes sense, but I also don’t think 1MB of L2 makes sense, it should be 2MB.


That or nvidia has a SLC that leverages that… but that’s… an extra layer.
 
I think '4K' will be a sexy marketing bullet point but I don't think they'll ever reveal how they got there. By 'they' I mean Nintendo, maybe Nvidia might.
They don't really do the tech buzzword thing in their marketing; they'll briefly mention 4K output but won't likely make it a front and centre thing. But they DO discuss their hardware spec in depth at places like GDC, where they go on panel discussions and gush the particulars, just like they did with Switch.
 
Last edited:
They don't really do the tech buzzword thing in their marketing; they'll briefly mention 4K output but won't likely make it a front and centre thing. But they DO discuss their hardware spec in depth at places like GDC, where they go on panel discussions and gush the particulars, just like they did with Switch.
if they want a buzzword fest, they should have Nvidia Lightspeed to an "RTX Remix"-esque port of something like Ocarina of Time
 
0
I don’t really think that’s the issue and I have an issue(?) with it and that I find it weird, the ampere white papers it mentions that they actually couple 512 kB of level 2 cache to each 32-bit memory controller. Or it’s equivalent of 32 bit memory controller pairing. Every GPU follows the same schematic except Drake. drake should have 2 MB of level 2 cache following the same set up as the desktop, but it does not. It is the only GPU that is different from the rest. Even ORIN which has a 256-bit interface has 4MB.

Drake has 8x 16-bit memory controllers. 1MB L2

Orin has 16x 16-bit memory controllers. 4MB L2

Even if ORIN doesn’t use the 32-bit, it uses 16-bit and has 256kB coupled to each MC, giving 4096 kB of L2. If you want to look at it like that…

And in case one is curious on this, you can check literally every GA10x chip and count the memory controllers (it’s the bus width bit divided by 32) and multiply that by 512 and you get the kB of L2 cache it has.

GA102 has 12 MCs, has 6144kB
GA104 has 8MCs, has 4096kB. This one is similar in configuration to ORIN, and also has the 4MB of L2 cache.

But Drake has less cache than any other ampere based silicon relative to its MCs. I don’t think 4MB makes sense, but I also don’t think 1MB of L2 makes sense, it should be 2MB.

That or nvidia has a SLC that leverages that… but that’s… an extra layer.
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
 
But it doesn’t have matrix units? that is a misconception. All they did was introduced more instructions, they did not include any physically new hardware. The only GPU that they have they have with actual matrix unit hardware is cDNA. Which isn’t for consumers.

The new unified AMD RDNA ™ 3 features new AI Accelerators
Slide deck showing the AI Accelerator as part of the Dual Compute Unit Architecture
AMD RDNA3 Presentation where they talk about the CU design, and show two AI accelerators per CU pair

Yes, this hardware is exposed via new instructions, the same way that FP64 is exposed as instructions that share the scheduler/cache/register files but are executed on a different block.

No, these are not the same matrix hardware that is in cDNA, this matrix accelerator isn't intended for use in general compute workloads, but strictly in games.
 
The new unified AMD RDNA ™ 3 features new AI Accelerators
Slide deck showing the AI Accelerator as part of the Dual Compute Unit Architecture
AMD RDNA3 Presentation where they talk about the CU design, and show two AI accelerators per CU pair

Yes, this hardware is exposed via new instructions, the same way that FP64 is exposed as instructions that share the scheduler/cache/register files but are executed on a different block.

No, these are not the same matrix hardware that is in cDNA, this matrix accelerator isn't intended for use in general compute workloads, but strictly in games.
it’s really not what you think it is, it’s literally just the shaders that they added more instructions to it. This isn’t anything like tensor cores, it’s straight up, just shaders. Despite the slide, it’s being misread/misinterpreted/misunderstood


The “dedicated AI unit” is really a RDNA3 shader executing the task, it’s not anything like the tensor cores which has it as a completely separate unit in its WGP.

To make it clearer, it’s straight up how AMD has the Texture Mapping Units perform Ray Tracing tasks, while it doesn’t have any exact dedicated RT unit/core like the type in Intel or NVidia GPUs.

AMD calling it “dedicated” is misleading, but I guess it depends on what perspective you look at it. Is there hardware that only does that? No. Is there shader units that are meant to do “double duty” solely for this? Yes. That’s what they are doing.



And I’m saying this, because people have been misrepresenting AMD as the kind to do fixed function hardware, when that is the complete the opposite of what they ever attempt to do. For consumer.

* Hidden text: cannot be quoted. *
I’m aware, I’m only saying that Drake is the only odd one if we use any other Ampere guidance as any reference. I’m not sure why they reduced the Cache but I suppose they have their reasons.
 
Last edited:
I fully agree that launching a new console with a remarkable game is critical, and only 3 games fit the bill in my opinion: Zelda, Mario, and Mario Kart.
We've seen nothing of the latter two, though we can be sure that they've been actively worked on for possibly several years. I'm personally betting that the reason why we haven't seen a thing about those is because Nintendo wants to showcase them in the best condition possible, that is running on a better hardware than the Tegra X1 switch. I'm also betting that both games will come within the first 6 months of the release of that new console, and one will be the launch title.

While we haven't seen much about the new Zelda, we have seen it running several times now, and definitely on the current Switch. While I absolutely do not doubt that the game will receive a patch on the new hardware, and will look gorgeous, I don't believe that both release dates are tied. However, I wouldn't be surprised if the new console released concomitantly with a Zelda DLC, in addition to Mario or Mario Kart.

What I currently see is either the new console releasing in the first half of 2024 alongside Mario kart and a Zelda DLC, and the new Mario releasing at the end of 2024. Or alternatively, though less likely in my opinion, the console releasing at the end of 2023 (say, October) alongside the new Mario and maybe Metroid Prime 4, with Mario Kart and the Zelda DLC coming during spring 2024.
 
I fully agree that launching a new console with a remarkable game is critical, and only 3 games fit the bill in my opinion: Zelda, Mario, and Mario Kart.
We've seen nothing of the latter two, though we can be sure that they've been actively worked on for possibly several years. I'm personally betting that the reason why we haven't seen a thing about those is because Nintendo wants to showcase them in the best condition possible, that is running on a better hardware than the Tegra X1 switch. I'm also betting that both games will come within the first 6 months of the release of that new console, and one will be the launch title.

While we haven't seen much about the new Zelda, we have seen it running several times now, and definitely on the current Switch. While I absolutely do not doubt that the game will receive a patch on the new hardware, and will look gorgeous, I don't believe that both release dates are tied. However, I wouldn't be surprised if the new console released concomitantly with a Zelda DLC, in addition to Mario or Mario Kart.

What I currently see is either the new console releasing in the first half of 2024 alongside Mario kart and a Zelda DLC, and the new Mario releasing at the end of 2024. Or alternatively, though less likely in my opinion, the console releasing at the end of 2023 (say, October) alongside the new Mario and maybe Metroid Prime 4, with Mario Kart and the Zelda DLC coming during spring 2024.
please show me when Nintendo has ever launched a system with a Mario kart as the launch title. I’ll wait.
 
8 deluxe was pretty close. Which was a new game to the vast majority.
“Close” isn’t march 3rd 2017, therefore my point still stands.

Mario Kart has never, ever, ever, ever been a launch title for any Nintendo system. It has been a launch window title at the earliest.

Point blank, period.
 
Quite frankly, it doesn’t really make any sense to launch a whole new game in the same series so soon right after the final wave of the Mario kart 8 DLC. That ends in December, and to launch one right after in three months is a very bizarre move especially from Nintendo. The likeliest and earliest is the end of 2024 or the following year, a.k.a. 2025.

Nintendo for reasons that we don’t know, has selected to never launch a system with Mario kart, despite it being a very strong selling title for their systems and can eclipse any title on their consoles, they haven’t selected it for selling the new console. Perhaps it is because Mario kart does not really try to reinvent self other than being Mario kart unlike the likes of Mario or Zelda which do try to be very different.

Like for example, Nintendo can easily launch a Nintendo system with a new animal crossing but they don’t do it. Simply because something sells a lot doesn’t mean they will use anything that sells a lot as a launch title for a system.



Splatoon 3 already happened and that DLC ends in late 2024. Pokémon is lol. Mario Kart is supporting the main title of the system but it isn’t treated as the launch title. Plus… the team (EPD 9) is currently doing work for the current title. Pikmin sells… but not systems and depends on other titles to sell. Metroid Prime is Retro who love to take their time… sigh…. anyway. Tomodachi life isn’t that common to rely upon to make any assessment at the moment so I won’t say it is or isn’t. A new IP is a given to launch the system with, or an IP revisited for the new system to be given new life but it isn’t the star of the show.


2D Mario and 3D Mario are absent, longer than expected, and their teams haven’t released anything in ages. Well, actually the team for 2D is set to release Pikmin 4 next year (maybe).


But, EPD8 is nowhere to be seen suspiciously 🤔. Last title they did was Bowser’s Fury and that was with NST assisting them on that.


Grezzo has also been absent since Link’s Awakening, and they are working on a new title that hasn’t been revealed. Maybe it was LA? Maybe it wasn’t and it’s a new IP? Maybe it’s an old IP? Who knows?


But it’s possible that they will be prevalent in the first year of the new system’s launch, and if not then, the second year.

But it’s geez so so expect it the 2-3rd year.


Chibi Robo is dead. But Bandai Namco can be heading this.


Nintendo is depending on Bandai Namco for help with titles as is, wouldn’t even be shocked.

But whatever the launch title is, it is made special by Nintendo. To showcase not only the new gimmick, but also what the system can do. It isn’t a tech demo, but it takes advantage of the hardware uniqueness.

Sometimes it’s more subtle… sometimes it’s really in your face. All the same it incorporates a feature of the system.
 
Ah yeah patterns.
It's weird that you're singling out Mario Kart in particular when 1.) the only console launches that Nintendo did with titles that even sort of resembled that were the bad ones, and 2.) that series is sort of occupied elsewhere right now.

The typical Nintendo launch day lineup consists of a mainline Zelda or Mario, which is frequently accompanied by a tech demo for the hardware. TotK obviously fits the bill for Zelda, but we're kind of due a Mario, as well, so it really could be either, but TotK also generally fits the expected timing, so it's no surprise most have attached to that one. They could do something different, as this won't exactly be a typical hardware launch for them, but, as mentioned above, Mario Kart doesn't really fit at all in a world where the Booster Course Pass exists.
 
0
Quite frankly, it doesn’t really make any sense to launch a whole new game in the same series so soon right after the final wave of the Mario kart 8 DLC. That ends in December, and to launch one right after in three months is a very bizarre move especially from Nintendo. The likeliest and earliest is the end of 2024 or the following year, a.k.a. 2025.

Nintendo for reasons that we don’t know, has selected to never launch a system with Mario kart, despite it being a very strong selling title for their systems and can eclipse any title on their consoles, they haven’t selected it for selling the new console. Perhaps it is because Mario kart does not really try to reinvent self other than being Mario kart unlike the likes of Mario or Zelda which do try to be very different.

Like for example, Nintendo can easily launch a Nintendo system with a new animal crossing but they don’t do it. Simply because something sells a lot doesn’t mean they will use anything that sells a lot as a launch title for a system.



Splatoon 3 already happened and that DLC ends in late 2024. Pokémon is lol. Mario Kart is supporting the main title of the system but it isn’t treated as the launch title. Plus… the team (EPD 9) is currently doing work for the current title. Pikmin sells… but not systems and depends on other titles to sell. Metroid Prime is Retro who love to take their time… sigh…. anyway. Tomodachi life isn’t that common to rely upon to make any assessment at the moment so I won’t say it is or isn’t. A new IP is a given to launch the system with, or an IP revisited for the new system to be given new life but it isn’t the star of the show.


2D Mario and 3D Mario are absent, longer than expected, and their teams haven’t released anything in ages. Well, actually the team for 2D is set to release Pikmin 4 next year (maybe).


But, EPD8 is nowhere to be seen suspiciously 🤔. Last title they did was Bowser’s Fury and that was with NST assisting them on that.


Grezzo has also been absent since Link’s Awakening, and they are working on a new title that hasn’t been revealed. Maybe it was LA? Maybe it wasn’t and it’s a new IP? Maybe it’s an old IP? Who knows?


But it’s possible that they will be prevalent in the first year of the new system’s launch, and if not then, the second year.

But it’s geez so so expect it the 2-3rd year.


Chibi Robo is dead. But Bandai Namco can be heading this.


Nintendo is depending on Bandai Namco for help with titles as is, wouldn’t even be shocked.

But whatever the launch title is, it is made special by Nintendo. To showcase not only the new gimmick, but also what the system can do. It isn’t a tech demo, but it takes advantage of the hardware uniqueness.

Sometimes it’s more subtle… sometimes it’s really in your face. All the same it incorporates a feature of the system.
Nintendo hasn’t really done cross gen before (with a couple of exceptions).

I think assuming MK 9 (or any other game) is cross gen, will change the equation on release timing.
 
The last time Nintendo was so successful was during the DS and Wii era. Their respective life cycle lasted seven and six years.

I can't understand the impatience, which has existed for several years now, about a new generation of Switch.

If a new generation comes out in 2023, it will just be in line with normal Nintendo practices when a product is successful. The same amount of time also passed between the SNES and the N64.

I'm not at all pronouncing on the need for new hardware, obviously there needs to be one and that's not my point, but I find this general turmoil quite astonishing and I don't think it has existed in the past in such proportions.
 
The last time Nintendo was so successful was during the DS and Wii era. Their respective life cycle lasted seven and six years.

I can't understand the impatience, which has existed for several years now, about a new generation of Switch.

If a new generation comes out in 2023, it will just be in line with normal Nintendo practices when a product is successful. The same amount of time also passed between the SNES and the N64.

I'm not at all pronouncing on the need for new hardware, obviously there needs to be one and that's not my point, but I find this general turmoil quite astonishing and I don't think it has existed in the past in such proportions.
I would bet on this being a combination of some factors:

1. The popularity of the Switch, keeping in mind that before it came the Wii U (not popular), and before it, the Wii (popular, but from a time where the average gamer wasn't as educated in hardware nitty-gritties as today);

2. How fast mobile technology has advanced these past few years - which greatly exaggerates how "outdated" the current Switch specs are, compared to, say, modern top of the line phones; and

3. AI-Upscaling, like DLSS, FSR and XeSS, which is a match made in heaven for the Switch form factor (personally, this right here is the #1 reason I'm following Drake's development so eagerly).
 
The last time Nintendo was so successful was during the DS and Wii era. Their respective life cycle lasted seven and six years.

I can't understand the impatience, which has existed for several years now, about a new generation of Switch.

If a new generation comes out in 2023, it will just be in line with normal Nintendo practices when a product is successful. The same amount of time also passed between the SNES and the N64.

I'm not at all pronouncing on the need for new hardware, obviously there needs to be one and that's not my point, but I find this general turmoil quite astonishing and I don't think it has existed in the past in such proportions.
You forget the elephant in the room: the Switch is underpowered compared to its current-gen siblings (and it shows). We are now also used to mid-gen refreshes, a model that would have helped increase the visual fidelity of the games on the system, and Nintendo titles in particular.
 
At IEDM this year, TSMC's SRAM cell is only 5% smaller at 3nm than 5nm (0.0210 => 0.0199 um2)!
1670928681668
 
What is the consensus on backwards compatibility between OG Switch and Drake, specifically regarding an improved experience for existing Switch games on the new platform?

Nintendo has always offered some form of backwards compatibility between its console generations, from the Wii for home consoles and the GBC for handhelds. But, perhaps with the exception of the GBC, the new platform has never offered a better experience for existing games. Gamecube games don't run faster on the Wii, nor do Wii games run faster on the Wii U. It's the same for handhelds, even between revisions of the same generation. If I'm not mistaken, 3DS games don't run faster on the New 3DS / New 2DS XL despite a much more powerful hardware.

Beyond the obvious hardware reasons (incompatible architectures and consoles using previous generation hardware), it might be desirable for Nintendo to ensure and identical experience across generations for several reasons.

First, to ensure optimal compatibility. There is no guarantee that without a patch, a game designed to run at maximum speed on its original platform will run correctly on a more powerful platform. And it doesn't seem to me to be relevant for most publishers/developers to patch an old game to improve its performance on a new platform. It can be a big effort, if it is even possible (disbanded development teams, legal issues) and it can cause legal problems. And it involves maintaining an older game for an additional platform. No clear financial incentive there without Nintendo sweetening the deal. So it might be a good rule of thumb to throttle the hardware down to the original target platform envelope.

Then, I imagine that Nintendo will still be selling original Switch consoles for months or even years. A game made for the original Switch offering a much better experience on the new machine would put the original Switch in a very unfavorable light, give an unfair advantage in online games, etc.

Finally, Nintendo is the world champion at reselling the same games to players multiple times. Whether it's through simple ports or HD or remastered versions that often do the bare minimum.

I am well aware that the Xbox and Playstation ecosystems are rather generous in this respect. They often offer a greatly increased game experience on XBox Series X or PS5 for games of the previous generation (and even older) in a totally transparent way. And it does seem like the way to go to promote the new platform, in a most consumer-friendly way.

But Nintendo doesn't work this way. It's probably among the most consumer-hostile entertainment company there is. Maybe for traditionalist reasons or maybe because they don't have the same core market as their competitor (core-gamer US vs family-friendy Japan), not really sure about that.

But long story short, is there any reason to believe Drake will be any different?
 
What is the consensus on backwards compatibility between OG Switch and Drake, specifically regarding an improved experience for existing Switch games on the new platform?


But long story short, is there any reason to believe Drake will be any different?
1: the BC solution on Drake will not be handled by Nintendo, it will be handled by Nvidia.

2. Nintendo is not immune to trends in the industry.
 
1: the BC solution on Drake will not be handled by Nintendo, it will be handled by Nvidia.
Would Nvidia handle things that differently from Nintendo? It would not prevent Nintendo from enforcing lower clocks and/or disable hardware features in some kind of "compatibility profile" anyway. In an ideal world, Nintendo could try to maximize user experience and compatibility on a game-by-game basis (like they probably do for Virtual Console and NSO games) but that doesn't seem to scale very well. Maybe for select games?
OTOH, it's not the 90s anymore. I suppose most games can scale to stronger hardware natively without breaking.
2. Nintendo is not immune to trends in the industry.
They do take their sweet time to adopt them though. The very obvious lack of online features on Switch demonstrates they are lagging years behind the competition and that they are not ready to take much risks.

Note: not trying to be confrontational in any way here. Just playing Devil's Advocate for a bit. Of course I'd love 4K 60 fps Age of Calamity on Drake :)
 
Would Nvidia handle things that differently from Nintendo? It would not prevent Nintendo from enforcing lower clocks and/or disable hardware features in some kind of "compatibility profile" anyway. In an ideal world, Nintendo could try to maximize user experience and compatibility on a game-by-game basis (like they probably do for Virtual Console and NSO games) but that doesn't seem to scale very well. Maybe for select games?
OTOH, it's not the 90s anymore. I suppose most games can scale to stronger hardware natively without breaking.

They do take their sweet time to adopt them though. The very obvious lack of online features on Switch demonstrates they are lagging years behind the competition and that they are not ready to take much risks.

Note: not trying to be confrontational in any way here. Just playing Devil's Advocate for a bit. Of course I'd love 4K 60 fps Age of Calamity on Drake :)

1. You are right, Nintendo will ultimately oversee all the work that Nvidia does for them, and make the final decisions. But technically its hard to see how it would be even technically possible to make an A78 core run as slow as an A57 core at 1ghz. Same goes for the gpu.
Also I do think Nvidia being in charge makes a difference in that we can expect a more technically impressive BC solution than what Nintendo could have done on their own.

2. Agree with you on this point.
 
0
I would bet on this being a combination of some factors:

1. The popularity of the Switch, keeping in mind that before it came the Wii U (not popular), and before it, the Wii (popular, but from a time where the average gamer wasn't as educated in hardware nitty-gritties as today);

2. How fast mobile technology has advanced these past few years - which greatly exaggerates how "outdated" the current Switch specs are, compared to, say, modern top of the line phones; and

3. AI-Upscaling, like DLSS, FSR and XeSS, which is a match made in heaven for the Switch form factor (personally, this right here is the #1 reason I'm following Drake's development so eagerly).
I'm super excited about DLSS and all the technical advancements that have been talked about for months. I want them too.

However, I am surprised that we have managed to transform this enthusiasm into collective depression. It seems to me that this impatience is difficult to explain.

I mean, yes, sure, I want to play Xenoblade more comfortably, like I already wanted on the Wii, but I really don't feel like the Switch has become incapable of delivering viable games, which would be a valid reason for not being able to wait any longer.
You forget the elephant in the room: the Switch is underpowered compared to its current-gen siblings (and it shows). We are now also used to mid-gen refreshes, a model that would have helped increase the visual fidelity of the games on the system, and Nintendo titles in particular.

What I find interesting is that this elephant is not particularly new: the Nintendo Wii was far exceeded in terms of power by the Playstation 3 and the Xbox 360. It was even less powerful than the previous Xbox.

We have known for a while that Nintendo does not intend to compete with other consoles on the technological level. And it had never sparked such heated controversy before.

What amazes me the most is that there's literally absolutely no reason to think that we won't have a new generation, as always, but a lot of people still seem very concerned.

You are right, maybe consumer habits have changed. Maybe also the Nintendo Switch audience is a bit different and has different expectations.
 
0
How likely it is for the Drake SOC to include a full Tegra X1 or for the new console to embark both chips much like the 3DS and Wii U did before?
 
What is the consensus on backwards compatibility between OG Switch and Drake, specifically regarding an improved experience for existing Switch games on the new platform?

Nintendo has always offered some form of backwards compatibility between its console generations, from the Wii for home consoles and the GBC for handhelds. But, perhaps with the exception of the GBC, the new platform has never offered a better experience for existing games. Gamecube games don't run faster on the Wii, nor do Wii games run faster on the Wii U. It's the same for handhelds, even between revisions of the same generation. If I'm not mistaken, 3DS games don't run faster on the New 3DS / New 2DS XL despite a much more powerful hardware.

Beyond the obvious hardware reasons (incompatible architectures and consoles using previous generation hardware), it might be desirable for Nintendo to ensure and identical experience across generations for several reasons.

First, to ensure optimal compatibility. There is no guarantee that without a patch, a game designed to run at maximum speed on its original platform will run correctly on a more powerful platform. And it doesn't seem to me to be relevant for most publishers/developers to patch an old game to improve its performance on a new platform. It can be a big effort, if it is even possible (disbanded development teams, legal issues) and it can cause legal problems. And it involves maintaining an older game for an additional platform. No clear financial incentive there without Nintendo sweetening the deal. So it might be a good rule of thumb to throttle the hardware down to the original target platform envelope.

Then, I imagine that Nintendo will still be selling original Switch consoles for months or even years. A game made for the original Switch offering a much better experience on the new machine would put the original Switch in a very unfavorable light, give an unfair advantage in online games, etc.

Finally, Nintendo is the world champion at reselling the same games to players multiple times. Whether it's through simple ports or HD or remastered versions that often do the bare minimum.

I am well aware that the Xbox and Playstation ecosystems are rather generous in this respect. They often offer a greatly increased game experience on XBox Series X or PS5 for games of the previous generation (and even older) in a totally transparent way. And it does seem like the way to go to promote the new platform, in a most consumer-friendly way.

But Nintendo doesn't work this way. It's probably among the most consumer-hostile entertainment company there is. Maybe for traditionalist reasons or maybe because they don't have the same core market as their competitor (core-gamer US vs family-friendy Japan), not really sure about that.

But long story short, is there any reason to believe Drake will be any different?

Nintendo is usually consistent with at least 1 gen BC except in a handful of cases. They are probably the most consistent with handheld BC.
 
If older Switch games won't run any better on Drake, that will only promote PC emulation and hacking the new device, if it's possible in the future.
 
You forget the elephant in the room: the Switch is underpowered compared to its current-gen siblings (and it shows). We are now also used to mid-gen refreshes, a model that would have helped increase the visual fidelity of the games on the system, and Nintendo titles in particular.
How is that an elephant in the room. The switch was outdated from the get-go. Drake will still be underpowered compared to Series X and PS5, and Nintendo consoles have been underpowered -and the hybrid model will ensure it stays that way. We will probably get new hardware soon, if only because Switch sales will start to decline and to ensure third party ports are feasible, but high horsepower is not Nintendo's goal and hasn't been since, at least, the Wii.
 
How is that an elephant in the room. The switch was outdated from the get-go. Drake will still be underpowered compared to Series X and PS5, and Nintendo consoles have been underpowered -and the hybrid model will ensure it stays that way. We will probably get new hardware soon, if only because Switch sales will start to decline and to ensure third party ports are feasible, but high horsepower is not Nintendo's goal and hasn't been since, at least, the Wii.
Maybe it was a poor choice of words, I simply wanted to give a good explanation about why the market wanted a beefier Switch model so early in its lifecycle (compared to past generations).
 
How is that an elephant in the room. The switch was outdated from the get-go. Drake will still be underpowered compared to Series X and PS5, and Nintendo consoles have been underpowered -and the hybrid model will ensure it stays that way. We will probably get new hardware soon, if only because Switch sales will start to decline and to ensure third party ports are feasible, but high horsepower is not Nintendo's goal and hasn't been since, at least, the Wii.
How was it outdated from the get go? I thought the Switch punched above it’s weight class in some aspects compared to other tech at the time.
 
How is that an elephant in the room. The switch was outdated from the get-go. Drake will still be underpowered compared to Series X and PS5, and Nintendo consoles have been underpowered -and the hybrid model will ensure it stays that way. We will probably get new hardware soon, if only because Switch sales will start to decline and to ensure third party ports are feasible, but high horsepower is not Nintendo's goal and hasn't been since, at least, the Wii.
Outdated ? In what world ? What systems allowed u to play BotW, MK8D, Skyrim type of games on the go and in Full HD on the big screen in 2016/2017 ?

Having less raw power than stationary systems doesnt mean a system is underpowered, let alone that its outdated.
Thats like calling the SteamDeck outdated or underpowered because it doesnt outperform PS5/XSX.
 
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