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

[Alovon11]

I'm not sure that the disappearance of Orin S was due to Nintendo co-opting it. It probably has been renamed Orin CX. @Thraktor caught it from a GTC session (April 2021):

It seems that Orin CX is intended for Nvidia's Drive IX platform for "in-cabin experiences", including gaming while driving (Elon's bad influence):

The Drive IX platform probably requires less power than Drive AV, which does the heavy lifting of autonomous driving, and thus we may surmise that Orin CX is a less powerful variant than Orin X and Orin regular. The first generation Drive had the same dichotomy between cockpit experience and autonomous driving:

Here are the Orin variants that have been put forth by Nvidia AFAIK:

• 2x Orin (unclear if X) + 2x discrete Ampere GPU
• 2x Orin X
• 2x Orin
• Orin X
• Orin
• Orin CX (possibly supplanted Orin S)
• Windshield ADAS

We are promised "an in-depth view into the NVIDIA Orin product line up" at the latest GTC on Nov 10. Hope it'll be good.

Well if Orin CX shares the same dichotomy from the Drive-Series, then that would put it right on the money for what we've predicted Orin S would be, half of Orin.

So Orin CX would (if being exact), 6 A78AE With 8 "Lovelace" SMs

Which leaves Nintendo room to customize it and use the space to fit 6-8 A78C CPU cores (As the A78AE's are notably bigger)

 

[SiG]

I'm starting to believe this DLSS model will be out by the time Dune Part 2 releases.

 

[chocolate_supra]

I'm starting to believe this DLSS model will be out by the time Dune Part 2 releases.

The DLSSpice must flow

 

[Terrell]

There is no way an 8nm Dane/DLSS/4K Switch launches in 2022 the same price as the TX1+ OLED Switch.

It has to at least be $100 more than the older model.

Even if we assume the OLED Switch gets a price cut to $300 when the 4K Switch launches, that’s still a minimum $400 machine price.

I’m expecting $450-$499 at this point.

It’s ok if it has a premium price tag, Nintendo still plans on selling the older models as the cheaper entry price points for another 4-5 years. The $400+ 4K Switch revision will be for people who want to pick Switcb gaming NOW

The hardware hasn't seen a price drop in 4 and a half years. Even their best selling hardware ever saw price drops in their first 3-4 years. The only dedicated gaming hardware from ANY of the 4 big names (Nintendo, Sega, Microsoft, Sony) I'm aware of that really never did was Game Boy Color, and it was already at a rock-bottom $70 price tag at launch. Even the Wii U saw a single price drop and it took them over a year to sell through their initial production run.
The only reason I can tell for why we haven't seen a price drop yet is Nintendo is using Switch sales to compensate for their prior console failure, which they admitted to back in 2016.

Additionally, the next hardware release doesn't have to be (and likely won't be) released in 2022, considering the SoC it is expected to use isn't even taped out yet. If one does release next year, it's not going to use Dane, but something far closer to the TX1+.

Lastly, the price doesn't "have to be" anything. Game Boy Advance launched for only $30 more than the Game Boy Color that preceded it. Switch launched at the same MSRP as a Wii U (though retailers were selling them below MSRP to ditch them).

 

[Pokemaniac]

The hardware hasn't seen a price drop in 4 and a half years. Even their best selling hardware ever saw price drops in their first 3-4 years. The only dedicated gaming hardware from ANY of the 4 big names (Nintendo, Sega, Microsoft, Sony) I'm aware of that really never did was Game Boy Color, and it was already at a rock-bottom $70 price tag at launch. Even the Wii U saw a single price drop and it took them over a year to sell through their initial production run.
The only reason I can tell for why we haven't seen a price drop yet is Nintendo is using Switch sales to compensate for their prior console failure, which they admitted to back in 2016.

Additionally, the next hardware release doesn't have to be (and likely won't be) released in 2022, considering the SoC it is expected to use isn't even taped out yet. If one does release next year, it's not going to use Dane, but something far closer to the TX1+.

Lastly, the price doesn't "have to be" anything. Game Boy Advance launched for only $30 more than the Game Boy Color that preceded it. Switch launched at the same MSRP as a Wii U (though retailers were selling them below MSRP to ditch them).

IIRC Dane can be taped out as late as Spring and still make 2022. Nvidia's recent chips have had pretty short turnarounds. I think generally we're at the point where it could be taped out any day now, if it hasn't happened already, though.

Also, as far as we know, it is the next chip Nvidia has in the works for Nintendo. There have been no other rumors or other suggestions about anything else, and nothing else viable on Nvidia's roadmaps that isn't a candidate for being Dane itself.

 

[Terrell]

IIRC Dane can be taped out as late as Spring and still make 2022. Nvidia's recent chips have had pretty short turnarounds. I think generally we're at the point where it could be taped out any day now, if it hasn't happened already, though.

Also, as far as we know, it is the next chip Nvidia has in the works for Nintendo. There have been no other rumors or other suggestions about anything else, and nothing else viable on Nvidia's roadmaps that isn't a candidate for being Dane itself.

It's the "as far as we know" that gives me pause.

If it makes 2022, considering the volume Nintendo is likely to need, I'd expect massive shortages at retail. It'd be brutal how few of them they could actually ship.

And we have conflicting tales in this regard as is. A lot of people expected/reported that OLED Switch was to be the model that included DLSS (which would likely mean an Orin SoC), the one that kits allegedly went out for, but based on the timeline for when Orin was to be ready to fabricate, that was never going to happen.
So either OLED was never going to be more that it was and everyone conflated it with whatever Dane was intended for, or there was something else that was supposed to go in it. I personally lean to the former explanation, but I can't totally rule out the latter, either, which would indicate to me that, if OLED Switch was meant to carry a different SoC, there was another SoC in play, because it couldn't have been Dane.

Like I said, the second option doesn't seem likely to me and I think OLED is exactly what it was designed to be, but if we're going to take the second option as a possibility, there'd have to be another SoC in play.

 

[Dakhil]

If it makes 2022, considering the volume Nintendo is likely to need, I'd expect massive shortages at retail. It'd be brutal how few of them they could actually ship.

I believe that even if Nintendo and Nvidia tape out Dane before the end of 2021, product shortages are unfortunately inevitable, especially if demand for the DLSS model* turns out to be as strong as demand for the Nintendo Switch. Saying that, I don't think Dane would be the primary reason for product shortages, going by Toshiba's statements about power regulating chip supplies. Not even Apple is immune to product shortages nowadays.

 

[Pokemaniac]

It's the "as far as we know" that gives me pause.

If it makes 2022, considering the volume Nintendo is likely to need, I'd expect massive shortages at retail. It'd be brutal how few of them they could actually ship.

And we have conflicting tales in this regard as is. A lot of people expected/reported that OLED Switch was to be the model that included DLSS (which would likely mean an Orin SoC), the one that kits allegedly went out for, but based on the timeline for when Orin was to be ready to fabricate, that was never going to happen.
So either OLED was never going to be more that it was and everyone conflated it with whatever Dane was intended for, or there was something else that was supposed to go in it. I personally lean to the former explanation, but I can't totally rule out the latter, either, which would indicate to me that, if OLED Switch was meant to carry a different SoC, there was another SoC in play, because it couldn't have been Dane.

Like I said, the second option doesn't seem likely to me and I think OLED is exactly what it was designed to be, but if we're going to take the second option as a possibility, there'd have to be another SoC in play.

As you say, the most reasonable explanation for the confusion around the Switch OLED is that two models were conflated. It is very easy to see how this could have happened, as there would have been nearly no overlap in which one sources would know about until basically right up until Switch OLED was announced. We also have hard evidence that Switch OLED was always Mariko (TX1+), as it initially showed up that way in the Switch firmware in spring 2020.

There are a couple details that don't entirely line up, like the reports of games being shown at E3 this year, but, as was posted here earlier, there is some circumstantial evidence that Dane may have been intended to release in 2021 at some point, as a new Jetson Nano was originally targeted for this year (for context, the current Jetson Nano is built from binned Switch chips). Alternatively, that part could have just been wrong/misinterpreted.

 

[Dakhil]

So TSMC talked more about TSMC's N3 process node during the TSMC Open Innovation Platform Ecosystem Forum yesterday. And although I don't see Nintendo and Nvidia using TSMC in the foreseeable future, especially when Dane's concerned, TSMC showed a chart comparing the Cortex-A78 between different process nodes from TSMC in terms of performance and power efficiency, which I find interesting. (I'm sure @Look over there will also find the chart interesting.)

 

[My Tulpa]

The hardware hasn't seen a price drop in 4 and a half years. Even their best selling hardware ever saw price drops in their first 3-4 years. The only dedicated gaming hardware from ANY of the 4 big names (Nintendo, Sega, Microsoft, Sony) I'm aware of that really never did was Game Boy Color, and it was already at a rock-bottom $70 price tag at launch. Even the Wii U saw a single price drop and it took them over a year to sell through their initial production run.
The only reason I can tell for why we haven't seen a price drop yet is Nintendo is using Switch sales to compensate for their prior console failure, which they admitted to back in 2016.

Well yea, Switch consoles still, after 4.5 years can barely stay in stock for very long. No point in a price cut under these conditions. Might as well make that profit on hardware sales while you can.

And just want to point out, Nintendo’s money loss from 2011-2016 was due to the 3ds, not the Wii U.

Additionally, the next hardware release doesn't have to be (and likely won't be) released in 2022, considering the SoC it is expected to use isn't even taped out yet. If one does release next year, it's not going to use Dane, but something far closer to the TX1+.

From what I understand, the Dane SoC will have been in production plenty in time for a 2022 launch. Even if I was going to pessimistic and say a March 2023 launch, the new Switch is going to be announced in 2022 at the very least. So the price point is a 2022 thing either way

Lastly, the price doesn't "have to be" anything. Game Boy Advance launched for only $30 more than the Game Boy Color that preceded it.

You are talking about systems $100 or less lol. That’s a 44% increase in price from the GBC to GBA.

A 44% increase here would mean OLED $350 and 4K Switch $500. Or, if there is a price cut, OLED $300 and the 4K Switch $440

Makes sense to me!

And your examples are true successors, the price point of completely new consoles needs to be aggressive if you want the userbase to cross over quicker.

This new model is a revision, not a successor. So it doesn’t have to be priced so aggressively. Nintendo will want a bigger % difference to delineate models. Nintendo is fine if people buy Tx1+ Switches the next 5 years. Nintendo games will sell either way. The 4K switch is for people who really want those enhanced Switch gaming graphics/performance. That migration can be slow, it’s not really a problem

 

[Look over there]

Hmm, so for N5->N3, A78 getting -26% power@iso-freq or +10% freq@iso-power is at the lower end of the ranges given at their roadmap update on the 18th (-25-30% power or +10-15% perf).
N4P should also be really close to N3 in power or perf, according to this. So the claim there is that N5->N4P offers -22% power@iso-perf or +11% perf@iso-power. N3's real advantage here would be the generation's worth of logic density improvement.
Seeing as how N4P is actually the second refinement from N5 (N5->N4->N4P), N3->N3 DTCO isn't bad at all. They certainly are trying to wring everything out of FinFET.

 

[Golden Darkness]

Not that it'll play directly with the upcoming hardware, but the EU commission is going to dig deeper into the Nvidia/ARM merger.

Nvidia's $40 billion takeover of Arm faces in-depth investigation from EU

The European Commission, the executive arm of the EU, has launched an "in-depth investigation" into Nvidia's acquisition of U.K. chip designer Arm.

www.cnbc.com

 

[Dakhil]

Not that it'll play directly with the upcoming hardware, but the EU commission is going to dig deeper into the Nvidia/ARM merger.

Nvidia's $40 billion takeover of Arm faces in-depth investigation from EU

The European Commission, the executive arm of the EU, has launched an "in-depth investigation" into Nvidia's acquisition of U.K. chip designer Arm.

www.cnbc.com

Here's an interesting tidbit from the European Commission's press release, with the potentially relevant parts highlighted in yellow.

The preliminary investigation suggests that the merged entity would also have the economic incentive to engage in such foreclosure strategies which could reduce competition in the market for the supply of processor products across different fields of application:

• datacentre CPUs;
• smart network interconnects ('SmartNICs') used in datacentres to offload network, storage, and security processing from the CPU to reduce its workload and accelerate its processing;
• semiconductors used for automotive advanced driver-assistance systems ('ADAS'), which encompass a broad range of technical features enabling vehicles to assist the driver;
• semiconductors used in infotainment applications, which refers to in-vehicle information and entertainment for drivers and passengers and includes various features, such as audio and video playback, automotive navigation systems, USB and Bluetooth connectivity, internet access, and Wi-Fi;
• SoCs equipping high-performance IoT devices;
• SoCs used in gaming consoles;
• SoCs used in general-purpose PCs.

 

[Alovon11]

Here's an interesting tidbit from the European Commission's press release, with the potentially relevant parts highlighted in yellow.

The EU, going and leaking Dane /s

 

[ReddDreadtheLead]

I was looking on wikipedia, at all the Tegra and they don't really seem to skimp on the L2 Cache for them, increasing per generation of Tegra, so the likelihood of at least 8MB of L2 maybe seems likely?

Tegra - Wikipedia

en.wikipedia.org

 

[Dakhil]

I was looking on wikipedia, at all the Tegra and they don't really seem to skimp on the L2 Cache for them, increasing per generation of Tegra, so the likelihood of at least 8MB of L2 maybe seems likely?

Tegra - Wikipedia

en.wikipedia.org

Arm mentions that the Cortex-A78C has 256 - 512 KB of L2 cache. So assuming the amount of L2 cache is per CPU core, which is the case for the Snapdragon 888, Exynos 2100, etc., the hexa-core (6) configuration of the Cortex-A78C has up to 3 MB of L2 cache, and the octa-core (8) configuration of the Cortex-A78C has up to 4 MB of L2 cache.

I don't know if I can see Nvidia customising the amount of L2 cache for the Cortex-A78C, considering that Nvidia didn't customise the amount of L2 cache for the Cortex-A57, since Arm mentioned that the Cortex-A57 has a single shared L2 unified cache of 512 KB - 2 MB. But I can see Nvidia using 8 MB of L3 cache for the Cortex-A78C, since Arm mentioned that the Cortex-A78C allows up to 8 MB of shared L3 cache.

 

[Look over there]

For the A78C, it's 256-512 KB of L2 cache per core, yes.

Older Tegras had climbing L2 cache because the L2 was the last level cache; they didn't have a L3.
Though yes, Xavier does look odd next to them; it's 2 MB of L2 for each pair of Carmel cores and finally there's 4 MB of L3 cache (exclusive).

I would expect Nvidia to stick with the specifications for L2/L3 cache. There's no reason for L2 to be as big as L3 anyway, I think? Too much SRAM cranks up the area taken up by said cache which increases design difficulty to keep latency within the desired range, and we want the levels closer to the CPU to have less latency. Or at least, that's how I understand things.

 

[NineTailSage]

For the A78C, it's 256-512 KB of L2 cache per core, yes.

Older Tegras had climbing L2 cache because the L2 was the last level cache; they didn't have a L3.
Though yes, Xavier does look odd next to them; it's 2 MB of L2 for each pair of Carmel cores and finally there's 4 MB of L3 cache (exclusive).

I would expect Nvidia to stick with the specifications for L2/L3 cache. There's no reason for L2 to be as big as L3 anyway, I think? Too much SRAM cranks up the area taken up by said cache which increases design difficulty to keep latency within the desired range, and we want the levels closer to the CPU to have less latency. Or at least, that's how I understand things.

I think all of that being said the M1 has 16MB of L2 cache and it has to be on every competitive chip makers mind of what is making Apple's chip so performant in comparison.

 

[Look over there]

Hmm, so it does, interesting.
Although strictly speaking, for the M1, 12 MB of L2 is for the cluster of performance cores and 4 MB is for the cluster of efficiency cores. The fact that the system level cache is shared with the GPU makes me guess that there's a stronger need for higher L2 cache dedicated to CPU cores. Like, I have this passing thought that the L2 cache is half functionally-the-last-level and half intermediate level.

 

[NineTailSage]

Hmm, so it does, interesting.
Although strictly speaking, for the M1, 12 MB of L2 is for the cluster of performance cores and 4 MB is for the cluster of efficiency cores. The fact that the system level cache is shared with the GPU makes me guess that there's a stronger need for higher L2 cache dedicated to CPU cores. Like, I have this passing thought that the L2 cache is half functionally-the-last-level and half intermediate level.

I could be wrong but I believe that there is the 12MB L2 for performance cores and 4MB of L2 for efficiency cores(16MB total)
plus the 16MB of SLC (System Level Cache). Not sure if that's what you were attesting to but on any front it's a large amount of die space dedicated for just cache in an SoC design...

 

[Look over there]

That's certainly a notable chunk out of a 120 mm^2 chip on N5.

Anyway, to organize my impression of this:
A core first checks its private L1 cache
If it's a miss, the core then moves onto the L2 cache shared by the cluster of performance or efficiency cores it's in
If it's a miss there, you then move onto the system level cache
(and of course if you miss there, it's RAM time, then following a miss there, it's storage time)
All of that is typical memory hierarchy stuff so far, but the diversion here is the system level cache also being accessible by GPU.

My layman gut instinct is that because the system level cache would be populated by stuff that the CPU cores want and stuff that the GPU wants, the hit rate will be lowered (from the perspective of specifically just one of the two) relative to a last level cache dedicated to just the CPU or the GPU individually. Thus, I have a thought that a larger L2 cache serves as compensation for the decreased hit rate at the last level.
But that's just a wild guess. I haven't heard of anything to support that.

 

[NineTailSage]

That's certainly a notable chunk out of a 120 mm^2 chip on N5.

Anyway, to organize my impression of this:
A core first checks its private L1 cache
If it's a miss, the core then moves onto the L2 cache shared by the cluster of performance or efficiency cores it's in
If it's a miss there, you then move onto the system level cache
(and of course if you miss there, it's RAM time, then following a miss there, it's storage time)
All of that is typical memory hierarchy stuff so far, but the diversion here is the system level cache also being accessible by GPU.

My layman gut instinct is that because the system level cache would be populated by stuff that the CPU cores want and stuff that the GPU wants, the hit rate will be lowered (from the perspective of specifically just one of the two) relative to a last level cache dedicated to just the CPU or the GPU individually. Thus, I have a thought that a larger L2 cache serves as compensation for the decreased hit rate at the last level.
But that's just a wild guess. I haven't heard of anything to support that.

I agree and everything about the way Apple has designed the M1 family of chips is about low latency access to memory allotments...
Down to the Lpddr4x being located on the same substrate of the M1 is just pure efficiency and minimizing bottlenecks as much as possible (without using an HBM solution) for bandwidth.

I'm not saying that Apple are reinventing the wheel here as both Nintendo and Microsoft used expensive eDRAM to try and eliminate the memory bandwidth bottleneck for both WiiU and XboxOne, but there just wasn't enough of that memory available to have a meaningful performance upgrade. Apple has just doubled down at this point on fast access to memory and even to AMD's credit the shining feature of RDNA2 has been Infinity Cache...

 

[NineTailSage]

So another topic I wanted to bring up since many of these early rumors for potential nex-gen Switch hardware have been incorporated in one form or another in the OLED model. What are the chances by say Switch 2 or maybe the refresh of that model gets a mini-LED panel?
Apple has already rolled out their line of mini-LED products and Nintendo now not only using OLED but opting for a decent panel (not just going with the cheapest one available) signifies the company putting an emphasis on at least a premium handheld gaming option.

Rumour: 'Switch Pro' Could Feature A Mini-LED Display, According To A New Report

Nintendo's looking to improve screen performance and battery life

www.nintendolife.com

I also don't see Nintendo wanting to be limited by the amount of next-gen Switch's they can sell based on what Samsung can manufacture (this is the same reason Apple were looking for something away from OLED). At least with mini-LED the promise of a wider playing field and vast reduction in prices once the market ramps up production is certain to be the better option for a hybrid gaming device.

I'm not sure if this will have any barring on the Dane Switch as OLED is probably the smarter choice to go with (currently for Nintendo) as many manufacturers are looking into what is next display tech wise. But it will bring about some questions if what's limiting OLED Switch's getting out the door ends up being the output from Samsung on displays supplied, will Nintendo then go the Apple route of seeking a different avenue down the road of mini-LED...

Edit: This whole question about the cost of mini-LED vs OLED was from a year ago, but it will be interesting to monitor this topic as more TV's are coming out now using mini-LED technology along with Apple's line-up of products.

Can we expect mini LED displays in phones and tablets?

Answer (1 of 2): Yes, only if its manufacturing cost plummets which is less than OLED’s. 1. At present it is certain that mini LED displays won’t be featured in smartphone displays because of its manufacturing cost and manufacturing complexity. 2. If the manufacturing process and cost goes down ...

www.quora.com

 

[Dakhil]

So another topic I wanted to bring up since many of these early rumors for potential nex-gen Switch hardware have been incorporated in one form or another in the OLED model. What are the chances by say Switch 2 or maybe the refresh of that model gets a mini-LED panel?
Apple has already rolled out their line of mini-LED products and Nintendo now not only using OLED but opting for a decent panel (not just going with the cheapest one available) signifies the company putting an emphasis on at least a premium handheld gaming option.

Rumour: 'Switch Pro' Could Feature A Mini-LED Display, According To A New Report

Nintendo's looking to improve screen performance and battery life

www.nintendolife.com

Probably a 0% chance if the DLSS model* is an iterative successor*** scheduled for release in holiday 2022 at the earliest. As for the DLSS model* refresh, I think it depends on a couple factors:

• Does Nintendo wants to release the DLSS model* refresh at the same price as the DLSS model*, similar to how the price of the Nintendo Switch (2019) is exactly the same price as the price of the Nintendo Switch (2017)?
• Are there any smartphones that are already using a similarly branded Mini LED displays that are relatively inexpensive by the time the DLSS model* refresh is planned for release?

Before I explain why I ask the second question, it should be noted that the model name for the OLED model's display is AMS699VC01. The reason why I mentioned the model name for the OLED model's display is because the OLED model's display is branded as the V series of OLED displays from Samsung, which were probably considered the mainstream OLED displays in 2020, but are probably no longer considered as such in 2021. And most of the smartphones using V series of OLED displays from Samsung are relatively inexpensive, so the OLED display on the OLED model is probably relatively inexpensive, although not the least expensive OLED display Nintendo could have chosen.

So far, I haven't heard any reliable rumours about smartphones using Mini LED displays at around 2023 or 2024, so for now, I think the chance for the DLSS model* refresh using a Mini LED display is very low.

*** → a successor, but with a longer than expected cross-gen period

 

[Terrell]

Well yea, Switch consoles still, after 4.5 years can barely stay in stock for very long. No point in a price cut under these conditions. Might as well make that profit on hardware sales while you can.

And just want to point out, Nintendo’s money loss from 2011-2016 was due to the 3ds, not the Wii U.

I think you're over-stating a lack of availability. Non-OLED Switch is readily available on every regional Amazon site and at every brick-and-mortar retailer in a 200km radius of me. There's also evidence from Famitsu sales figures and market analysis that Switch sales are steadily waning. In addition, Nintendo's share price fell 20% so far this year, dropping another 1.5% after the OLED launch, which has its own financial effect. So current market conditions aren't what you think they are. Either expect a price drop or for sales to continue their steady decline.

To the second point here, likely both 3DS and Wii U contributed to Nintendo's losses in that time period, but given the differences between hardware and software sold for 3DS and Wii U in the time period of 2012-2016 and how 3DS hardware sales were profit generators by mid-2012, between the 2, one of those things is most definitely not like the other. It's not even remotely debatable to suggest which contributed substantially more to that.

You are talking about systems $100 or less lol. That’s a 44% increase in price from the GBC to GBA.

A 44% increase here would mean OLED $350 and 4K Switch $500. Or, if there is a price cut, OLED $300 and the 4K Switch $440

Makes sense to me!

And your examples are true successors, the price point of completely new consoles needs to be aggressive if you want the userbase to cross over quicker.

This new model is a revision, not a successor. So it doesn’t have to be priced so aggressively. Nintendo will want a bigger % difference to delineate models. Nintendo is fine if people buy Tx1+ Switches the next 5 years. Nintendo games will sell either way. The 4K switch is for people who really want those enhanced Switch gaming graphics/performance. That migration can be slow, it’s not really a problem

First, reducing dollar value to a percentage is a rather crude way of looking at things. A $30 price jump is a much easier pill to swallow than a $150 price bump, as is a $100 MSRP compared to $450-500. To argue differently would be a fool's errand.

And this position of Nintendo not caring if a new hardware release sells slowly is a contradiction to your earlier claim.
Nintendo allegedly won't do a price cut because they want to make the maximum amount of money from their hardware... but they'll invest a whole bunch of their R&D budget into a brand-new (and more expensive to produce) SoC, among other potential component changes, for it to languish with slow sales at retail? You see the problem there, don't you?
By this logic, Switch Lite should have launched at $249 to maximize profits, since that still delineates it from the original Switch model. And it's unlikely they saved $100 per unit in component costs over the standard Switch to make that Switch Lite price as profitable as the original was at that time (R&D for TX1+ Mariko would have eaten into a huge chunk of any money saved by using a smaller LCD, lighter battery and non-detachable controls).
Money in this business relies on software sales, large amounts of revenue on hardware is typically just a nice extra, so Nintendo will price however they feel they must to keep software sales as high as possible for every piece of hardware they release.

Additionally, to that last point about software being the most important revenue source, if you insist on the "revision" idea, every enhanced hardware revision Nintendo has ever released has exclusive software, both from Nintendo and from 3rd-parties (2 out of every 3 Game Boy Color games was only playable on GBC, even the modest seller that was New 3DS had a Xenoblade and Fire Emblem Warriors), but that requires an install base to sell such titles to, so slow-selling hardware of ANY sort is not a smart play. There's no universe in which Nintendo is fine with hardware meeting with slow sales.

 

[Terrell]

So another topic I wanted to bring up since many of these early rumors for potential nex-gen Switch hardware have been incorporated in one form or another in the OLED model. What are the chances by say Switch 2 or maybe the refresh of that model gets a mini-LED panel?
Apple has already rolled out their line of mini-LED products and Nintendo now not only using OLED but opting for a decent panel (not just going with the cheapest one available) signifies the company putting an emphasis on at least a premium handheld gaming option.

Rumour: 'Switch Pro' Could Feature A Mini-LED Display, According To A New Report

Nintendo's looking to improve screen performance and battery life

www.nintendolife.com

I also don't see Nintendo wanting to be limited by the amount of next-gen Switch's they can sell based on what Samsung can manufacture (this is the same reason Apple were looking for something away from OLED). At least with mini-LED the promise of a wider playing field and vast reduction in prices once the market ramps up production is certain to be the better option for a hybrid gaming device.

I'm not sure if this will have any barring on the Dane Switch as OLED is probably the smarter choice to go with (currently for Nintendo) as many manufacturers are looking into what is next display tech wise. But it will bring about some questions if what's limiting OLED Switch's getting out the door ends up being the output from Samsung on displays supplied, will Nintendo then go the Apple route of seeking a different avenue down the road of mini-LED...

Edit: This whole question about the cost of mini-LED vs OLED was from a year ago, but it will be interesting to monitor this topic as more TV's are coming out now using mini-LED technology along with Apple's line-up of products.

Can we expect mini LED displays in phones and tablets?

Answer (1 of 2): Yes, only if its manufacturing cost plummets which is less than OLED’s. 1. At present it is certain that mini LED displays won’t be featured in smartphone displays because of its manufacturing cost and manufacturing complexity. 2. If the manufacturing process and cost goes down ...

www.quora.com

My feeling about OLED has been that Nintendo knew that OLED screens were going into their next hardware but, due to production constraints, needed to make space for themselves in a production schedule earlier than usual if they wanted enough for their next major hardware (among other components) by "motivating" (read: shaming) their suppliers to have more coming off the line, so they started putting in some component orders wicked early and decided to do something with them to illustrate demand via a guaranteed sellthrough rate. As of this summer, Samsung's small OLED (smartphones-to-laptops) production is just shy of 200K/month for ALL potential clients, including their own electronics divisions. If we see that ramp up again in really short order, it seems to me like Nintendo made a smart play having something to concretely illustrate component demand... at least, if my thinking on the subject is correct, anyways.

 

[Deleted member 540]

I think you're over-stating a lack of availability. Non-OLED Switch is readily available on every regional Amazon site and at every brick-and-mortar retailer in a 200km radius of me. There's also evidence from Famitsu sales figures and market analysis that Switch sales are steadily waning. In addition, Nintendo's share price fell 20% so far this year, dropping another 1.5% after the OLED launch, which has its own financial effect. So current market conditions aren't what you think they are. Either expect a price drop or for sales to continue their steady decline.

Oh no. Is it serious?

 

[Ookaze]

That's certainly a notable chunk out of a 120 mm^2 chip on N5.

Anyway, to organize my impression of this:
A core first checks its private L1 cache
If it's a miss, the core then moves onto the L2 cache shared by the cluster of performance or efficiency cores it's in
If it's a miss there, you then move onto the system level cache
(and of course if you miss there, it's RAM time, then following a miss there, it's storage time)

That's not how it works, a core can never access anything other than its L1 cache, or else it would grind to a halt, aka the IPC would plummet in most workloads as even the fastest L1 cache memory is too slow for a CPU core. The L1 cache accesses data from the slower L2 cache, the L2 from the next slower (and less expensive) memory. Expert level optimizations in code involve moving the data in cache the less.

All of that is typical memory hierarchy stuff so far, but the diversion here is the system level cache also being accessible by GPU.

My layman gut instinct is that because the system level cache would be populated by stuff that the CPU cores want and stuff that the GPU wants, the hit rate will be lowered (from the perspective of specifically just one of the two) relative to a last level cache dedicated to just the CPU or the GPU individually. Thus, I have a thought that a larger L2 cache serves as compensation for the decreased hit rate at the last level.
But that's just a wild guess. I haven't heard of anything to support that.

It can be done perhaps by having an API that will always stick motion vectors in a designated address space section of this SLC to help DLSS. When the workload is known, these can of things can be easier to do. Cache optimization is a very complex topic IMO.

 

[Ookaze]

Oh no. Is it serious?

If you read the article, it's only nonsense trying to paint an alternate reality. I don't know what is up with M. Mochizuki against Nintendo, but even before he was at Bloomberg, he was trying to negatively impact Nintendo's stocks. He seems settled on a hit job like Pachter was, with a bear tactic. I don't take him seriously for 2+ years now and I end up always correct when doing that, like with Pachter.

 

[ILikeFeet]

If you read the article, it's only nonsense trying to paint an alternate reality. I don't know what is up with M. Mochizuki against Nintendo, but even before he was at Bloomberg, he was trying to negatively impact Nintendo's stocks. He seems settled on a hit job like Pachter was, with a bear tactic. I don't take him seriously for 2+ years now and I end up always correct when doing that, like with Pachter.

Mochizuki isn't completely against Nintendo like Pachter was. I think Mochizuki is trying to force Nintendo to better acknowledge his findings

 

[uncleoptimus]

If you read the article, it's only nonsense trying to paint an alternate reality. I don't know what is up with M. Mochizuki against Nintendo, but even before he was at Bloomberg, he was trying to negatively impact Nintendo's stocks. He seems settled on a hit job like Pachter was, with a bear tactic. I don't take him seriously for 2+ years now and I end up always correct when doing that, like with Pachter.

I really disagree with this take and I don't find it grounded in reality based on the reporter's articles I have skimmed. Since most Nintendo sales has been quite positive of recent years, if anything, it makes him look like "pro-Nintendo" over this time. (Which I guess carries some irony since the way the whole "Pro" model thing shook out really rankled a lot of folks in the end!)

(Note btw Mochizuki is also the reporter that supposedly has a Sony hate boner and tried to tank the good vibes of the PS5 launch. Dude swinging around a lot of hate boners. Guess that makes him a sooper-secret Microsoft operative or maybe he works for Google Stadia division. )

Meanwhile dude's twitter is all about the tech he enjoys and the games he is diggin on his switch and PS5. Thin line btwn love and hate I suppose.

Oh no. Is it serious?

"Serious", Nah. It reports observed figures and identifies three major points:

a) supply of the OLED
b) a downward sales trend coming off extraordinary circumstances with the pandemic
c) the switch is over 4.5 years old

With point c), to me that underlines how strong the switch platform is; I can't think of another major system that gets to year 5 without a major price drop! (The article does mention the Europe price drop, but attributes that to currency fluctuation adjustment)

 

[Vanillalite]

March 2023 really wouldn’t be that bad, and we’d probably get a Switch like reveal in fall of 2022.

 

[Dakhil]

So Samsung announced HDR10+ Gaming during Samsung Developer Conference 2021 (SDC 2021), which is considered to be an alternative to Dolby Vision gaming, which Microsoft supports on the Xbox Series X|S. HDR10+ Gaming supports VRR up to 4K 120 Hz, automated HDR calibration, which FlatpanelsHD believes is derived from HDR Gaming Interest Group (HGiG), which Samsung's a member of, and low latency source tone mappings. (And interestingly enough, Nvidia's also a member of HGiG.)

Assuming Nintendo wants to support VRR for handheld mode and TV mode for the DLSS model*, could Samsung approach Nintendo to support HDR10+ Gaming? HDR10+ is an open standard and is royalty free, unlike Dolby Vision, which is proprietary and requires a licence fee payment.

 

[ILikeFeet]

So Samsung announced HDR10+ Gaming during Samsung Developer Conference 2021 (SDC 2021), which is considered to be an alternative to Dolby Vision gaming, which Microsoft supports on the Xbox Series X|S. HDR10+ Gaming supports VRR up to 4K 120 Hz, automated HDR calibration, which FlatpanelsHD believes is derived from HDR Gaming Interest Group (HGiG), which Samsung's a member of, and low latency source tone mappings. (And interestingly enough, Nvidia's also a member of HGiG.)

Assuming Nintendo wants to support VRR for handheld mode and TV mode for the DLSS model*, could Samsung approach Nintendo to support HDR10+ Gaming? HDR10+ is an open standard and is royalty free, unlike Dolby Vision, which is proprietary and requires a licence fee payment.

if Nintendo supported HDR, you can just look to how they handle audio to see which way they would swing

 

[Herb Alpert]

If you read the article, it's only nonsense trying to paint an alternate reality. I don't know what is up with M. Mochizuki against Nintendo, but even before he was at Bloomberg, he was trying to negatively impact Nintendo's stocks. He seems settled on a hit job like Pachter was, with a bear tactic. I don't take him seriously for 2+ years now and I end up always correct when doing that, like with Pachter.

Come on, don't transfer console war agendas bullshit onto someone like mochizuki.

 

[Terrell]

Oh no. Is it serious?

This response is correct:

"Serious", Nah. It reports observed figures and identifies three major points:

a) supply of the OLED
b) a downward sales trend coming off extraordinary circumstances with the pandemic
c) the switch is over 4.5 years old

With point c), to me that underlines how strong the switch platform is; I can't think of another major system that gets to year 5 without a major price drop! (The article does mention the Europe price drop, but attributes that to currency fluctuation adjustment)

In any case, any persistent downward trend in sales of hardware, regardless of cause, is always a concern to platform holders, one that they always intend to correct at the earliest opportunity before things take a worse turn. And the easiest way to correct that is either pack-in games to augment the value of the hardware at that price or price drops. This downward trend has been going on all year, but it had been relatively slight at first and has temporarily levelled off. Nintendo is probably waiting to see how November in the West and New Year in Japan plays out year-over-year before a decision is made.

But saying that supply isn't meeting demand at the current price point is not an accurate picture of the situation in 2021, except for the OLED model, that was my ultimate point.

 

[BlackTangMaster]

One of the first 2022 chipsets has been taped out. Dane is still yet to be seen on the node used for the 2020 chipsets.

 

[YolkFolk]

I think you're over-stating a lack of availability. Non-OLED Switch is readily available on every regional Amazon site and at every brick-and-mortar retailer in a 200km radius of me. There's also evidence from Famitsu sales figures and market analysis that Switch sales are steadily waning. In addition, Nintendo's share price fell 20% so far this year, dropping another 1.5% after the OLED launch, which has its own financial effect. So current market conditions aren't what you think they are. Either expect a price drop or for sales to continue their steady decline.

It’s just natural sales are going to decline. The system is almost 5 years old. It’s not because Nintendo are doing anything bad. There just comes a point in a systems life when sales peak and they begin to slow down compared to what they once were.

 

[Ookaze]

Come on, don't transfer console war agendas bullshit onto someone like mochizuki.

Where do you see console war agenda in what I wrote? I talk only about the Switch, Nintendo and stocks. You must have mixed what I said with someone responding to me that introduced other consoles, I don't know why. Read the article and it will be obvious that what he paints to its unaware audience (which are investors related to Bloomberg, not gamers) is an alternate reality.
For example, he paints the perfectly normal decline in 2021 (since new years big numbers) as an outlier and concerning phenomenom, he cites another analysts as saying the Switch may (he's not even sure) have peaked in FY 2021 because of no more powerful model released (his famous non existent Switch Pro), putting a greater onus on software to drive profit, like that hasn't always been Nintendo's way of making profit. Mochizuki's wording is careful and he lets another analyst say these words that are still in his hit piece as these are clueless statements that I'm sure he knows are nonsense. He puts the concern about no Switch Pro into investors' mouth when he is the one that manufactured this vision for years now.
He frames his words about Switch being on shelves in 2021 as an abnormal thing and that a price cut is needed to clear "unsold inventory".
This article has countless other nonsense like that that paint IMO an unhealthy alternate vision of reality, and looks like a bear tactic like everything I've seen him write for years.

 

[Terrell]

It’s just natural sales are going to decline. The system is almost 5 years old. It’s not because Nintendo are doing anything bad. There just comes a point in a systems life when sales peak and they begin to slow down compared to what they once were.

I agree completely with your assessment. That being said, price cuts rejuvenate slumping hardware sales and always have, because it widens the number of people who will buy it at its new price that either wouldn't or couldn't before.

 

[Herb Alpert]

Where do you see console war agenda in what I wrote? I talk only about the Switch, Nintendo and stocks. You must have mixed what I said with someone responding to me that introduced other consoles, I don't know why. Read the article and it will be obvious that what he paints to its unaware audience (which are investors related to Bloomberg, not gamers) is an alternate reality.
For example, he paints the perfectly normal decline in 2021 (since new years big numbers) as an outlier and concerning phenomenom, he cites another analysts as saying the Switch may (he's not even sure) have peaked in FY 2021 because of no more powerful model released (his famous non existent Switch Pro), putting a greater onus on software to drive profit, like that hasn't always been Nintendo's way of making profit. Mochizuki's wording is careful and he lets another analyst say these words that are still in his hit piece as these are clueless statements that I'm sure he knows are nonsense. He puts the concern about no Switch Pro into investors' mouth when he is the one that manufactured this vision for years now.
He frames his words about Switch being on shelves in 2021 as an abnormal thing and that a price cut is needed to clear "unsold inventory".
This article has countless other nonsense like that that paint IMO an unhealthy alternate vision of reality, and looks like a bear tactic like everything I've seen him write for years.

Playstation fans already think he hates Sony. Now you say that he hates Nintendo.
Maybe he's secretly working for Microsoft. Or ouya.

 

[ILikeFeet]

Playstation fans already think he hates Sony. Now you say that he hates Nintendo.
Maybe he's secretly working for Microsoft. Or ouya.

 

[Dakhil]

One of the first 2022 chipsets have been taped out. Dane is still yet to be seen on the node used for the 2020 chip
sets.

I think April 2022 is the absolute latest time for Dane to be taped out, assuming Nintendo's targeting a holiday 2022 release, based on how kopite7kimi mentioned on 11 March 2020 that GA102's taped out and the products using GA102 were released on September 2020. As for when is the earliest time for Dane to be taped out, I don't know. I think around November 2021 is definitely a possibility, especially with Nvidia talking about the SoCs in the Orin family on 10 November 2021 during GTC 2021 (November 2021).

 

[Sorbete]

Sorry if this is not the right thread to ask.

I looking to upgrade my 10 years old 32'' 720p LG TV. The thing is that I would prefer to avoid 4K TVs, sadly most of the offers are indeed 4K TVs. Is there's anything that I need to look or beware if I end up buying a 4k?

 

[ILikeFeet]

Sorry if this is not the right thread to ask.

I looking to upgrade my 10 years old 32'' 720p LG TV. The thing is that I would prefer to avoid 4K TVs, sadly most of the offers are indeed 4K TVs. Is there's anything that I need to look or beware if I end up buying a 4k?

Not unless you want to spend for a great tv. I got a $250 TCL that's good enough

 

[Alovon11]

Sorry if this is not the right thread to ask.

I looking to upgrade my 10 years old 32'' 720p LG TV. The thing is that I would prefer to avoid 4K TVs, sadly most of the offers are indeed 4K TVs. Is there's anything that I need to look or beware if I end up buying a 4k?

Honestly, if you are looking for a new TV I'd try to see if it has Freesync or VRR support tbh.

Even if not 4k (although most if not all do have it), it would make things a lot better experience-wise.

 

[Alovon11]

I think April 2022 is the absolute latest time for Dane to be taped out, assuming Nintendo's targeting a holiday 2022 release, based on how kopite7kimi mentioned on 11 March 2020 that GA102's taped out and the products using GA102 were released on September 2020. As for when is the earliest time for Dane to be taped out, I don't know. I think around November 2021 is definitely a possibility, especially with Nvidia talking about the SoCs in the Orin family on 10 November 2021 during GTC 2021 (November 2021).

I am inclined to agree.

Although if they do get a November 2021 Tape Out for Dane, then honestly that puts an August Release into possibility properly IMHO.

(Although technically if April 2022 is enough to hit a November 2022 launch, that would mean they could launch it literally anywhere in the Summer if they wanted to if they tape it out in November 2021)

 

[MetalLord]

Job Openings - Careers at Nintendo of America

Senior Engineer, Shader Compiler (NTD), Software Development

careers.nintendo.com

 

[Falcatops]

Sorry if this is not the right thread to ask.

I looking to upgrade my 10 years old 32'' 720p LG TV. The thing is that I would prefer to avoid 4K TVs, sadly most of the offers are indeed 4K TVs. Is there's anything that I need to look or beware if I end up buying a 4k?

Any reason in particular as to why you don’t want a 4K tv?

 

[NineTailSage]

Sorry if this is not the right thread to ask.

I looking to upgrade my 10 years old 32'' 720p LG TV. The thing is that I would prefer to avoid 4K TVs, sadly most of the offers are indeed 4K TVs. Is there's anything that I need to look or beware if I end up buying a 4k?

As others have already mentioned that because most manufacturers are off of making 1080p televisions, your choices will be very limited from super cheap brands or super expensive name brands that are in short supply of those specific resolutions.

In other words you can't go wrong looking at a 4k TV because prices are extremely low compared to where they started.
The real question is what do you want out of the TV (what's the primary usage for)?