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

[RennanNT]

Muji doesn't use big.LITTLE. All the cores are big.

Someone may correct me here, because I'm not completely sure on it, but I don't think it's possible to clock individual cores in a cluster differently than the rest of the cluster. All Muji/Oz/Switch 2/REDACTED cores will run at the same clock speed.

It is possible. You can form groups of cores (including a group with just one core) and have individual clocks for each group.

I didn't find the article I read long ago, but here's a video from ARM (see 1:10 in particular, for each core having their own frequency).



They even highlight a 1+7 configuration.

Phones using the X1 core is a good example too, since there's usually just 1 and it's clocked higher than other cores.

 

[Armoured_Bear]

Pretty optimistic to be honest. To recap:

1. Nintendo finally acknowledged the successor to Switch 2 and even gave a window in which it will be shown.
2. Nintendo confirmed a Nintendo Direct in June focusing on first party games for Switch.
3. Leaked specs so far for Switch 2 has been pretty impressive so far, potentially surpassing expectations and leaving us hopeful that the system will be quite capable and a huge step up from the aging Switch console.

So yeah it's been a good month I would say. Excitement Level is at like a 7/10 so far and that could increase as time goes on.

What have I missed?

 

[Darknut85]

It would be funny if a Yamanoi post, of all things, dropped some juicy info hidden behind a post requirement.

* Hidden text: cannot be quoted. *

 

[ItWasMeantToBe19]

What have I missed?

The RAM appears to be ~18% faster than expected.

 

[Hermii]

So few games or game engines this gen tax the CPU that I don't know if it matters very much.

It's like Dragon's Dogma 2 and Baldur's Gate 3 in terms of AAA games 3.5 years in.

If this is like 40-50% of the PS5's CPU, that's great and good enough. Third-party games and ambitious first-party games CPU-wise will run at 30 FPS regardless.

What estimated clock speed would give 40-50% of the PS5's CPU's performance roughly?

UE 5 games? Even though cpu optimization apparently improved a ton in the latest version.

 

[ItWasMeantToBe19]

Very CPU heavy games like DD2 would be interesting and cool for devs to make (though DD2 itself does not sound great), but I wonder if the PC gamer backlash to DD2 will stop devs from pursuing that path.

 

[Gerald]

Very CPU heavy games like DD2 would be interesting and cool for devs to make (though DD2 itself does not sound great), but I wonder if the PC gamer backlash to DD2 will stop devs from pursuing that path.

I’m don’t think any game push cpu too much, DD2 is just bad optimization,Game what will deliver truly alive,immersive,interactive open world of this gen will be GTA VI

 

[gonanzir]

The best part is, some cores may have heavier tasks while others have lighter ones, and DynamIQ can set the appropriate frequency and voltage for each core, eliminating the need for all cores within a cluster to remain at the same frequency and voltage.
In other words, as long as the power requirements are met in the end, you can freely adjust the voltage and frequency(Of course, there is a maximum value.) for each core,This is technically OK.

 

[Steve]

Very CPU heavy games like DD2 would be interesting and cool for devs to make (though DD2 itself does not sound great), but I wonder if the PC gamer backlash to DD2 will stop devs from pursuing that path.

Unlikely since it’s one of Capcom biggest success this year.

 

[kasparov77]

Speaking of the CPU, I don't really expect the CPU to run at >2 GHz, especially if the CPU frequency has to be identical in handheld mode and TV mode, similar with the Nintendo Switch, considering that scaling the CPU frequency is much more difficult than scaling the GPU frequency, and if the CPU has to run for a sustained period of time in handheld mode and TV mode.

i mean, if theres a game that's high cpu demanding but not that intensive gpu-wise they could just clock down the gpu in docked mode and upclock the cpu, no?

 

[Zedark]

So few games or game engines this gen tax the CPU that I don't know if it matters very much.

It's like Dragon's Dogma 2 and Baldur's Gate 3 in terms of AAA games 3.5 years in.

If this is like 40-50% of the PS5's CPU, that's great and good enough. Third-party games and ambitious first-party games CPU-wise will run at 30 FPS regardless.

What estimated clock speed would give 40-50% of the PS5's CPU's performance roughly?

Referring to Thraktor's post, we get a single core score for A78 at 3GHz being roughly the same as PS5's Zen 2 core at 4 GHz. Depending on your expectation (let's say the expected range is 1.5 GHz-2.0 GHz), it would be about 55%-75% of the PS5 CPU core-for-core.

With the necessary caveat that this is a general purpose benchmark, so for specific tasks the ratio can be different.

 

[Dakhil]

Google Tensor vs Google Tensor G2

Google Tensor has a die size of 108.26 mm².

And the Cortex-A78's L2 cache on the Google Tensor G2 has been reduced in half (256 KB) compared to the Cortex-A78's L2 cache on the Exynos 2100 (512 KB).

Google Tensor G2's probably fabricated using Samsung's 5LPP process node, considering Samsung's 5LPP process node is only 5% better than Samsung's 5LPE process node, which was used to fabricate Google Tensor.

 

[Ghostsonplanets]

Muji doesn't use big.LITTLE. All the cores are big.

Someone may correct me here, because I'm not completely sure on it, but I don't think it's possible to clock individual cores in a cluster differently than the rest of the cluster. All Muji/Oz/Switch 2/REDACTED cores will run at the same clock speed.

You just need individual power/voltage rails for each core. It's done in Mobile, particurlarly high-end SoCs with Cortex X or favored cores. Using a single voltage rail/power domain for the entire cluster is cheaper, however.

For an application like Switch 2, where at least 7 out of the 8 cores will be in use all the time and clocked at a fixed maximum value while in operation, I'd argue doing such extra work doesn't make sense.

 

[bmfrosty]

Asking again after thinking a bit more about the situation:

* Hidden text: cannot be quoted. *

Putting it at >2000 is a bit much. Some of us do shut up once in a while.

 

[Darknut85]

Putting it at >2000 is a bit much. Some of us do shut up once in a while.

If you look at it positively, it‘s a goal to work towards. You could maybe even reach it before Switch 2 launch.

 

[gonanzir]

You just need individual power/voltage rails for each core. It's done in Mobile, particurlarly high-end SoCs with Cortex X or favored cores. Using a single voltage rail/power domain for the entire cluster is cheaper, however.

For an application like Switch 2, where at least 7 out of the 8 cores will be in use all the time and clocked at a fixed maximum value while in operation, I'd argue doing such extra work doesn't make sense.

Even when running TotK, not every single core of the Switch is constantly at 100% utilization. So, I think Switch2 could be like this.

 

[ItWasMeantToBe19]

Unlikely since it’s one of Capcom biggest success this year.

Ehhhhhh, it did OK, but it only has 60% positive reviews on Steam and had pretty bad legs.

 

[Gerald]

Referring to Thraktor's post, we get a single core score for A78 at 3GHz being roughly the same as PS5's Zen 2 core at 4 GHz. Depending on your expectation (let's say the expected range is 1.5 GHz-2.0 GHz), it would be about 55%-75% of the PS5 CPU core-for-core.

With the necessary caveat that this is a general purpose benchmark, so for specific tasks the ratio can be different.

Also a78c is better than normal a78

 

[Armoured_Bear]

The RAM appears to be ~18% faster than expected.

Cool!
Thanks.

 

[darthdiablo]

Putting it at >2000 is a bit much. Some of us do shut up once in a while.

Just for your current message count, blink and you might miss it

 

[oldpuck]

Phones using the X1 core is a good example too, since there's usually just 1 and it's clocked higher than other cores.

Yes, but in those phones, the X1 core is the only X1 core. The DynamIQ documentation implies that all cores of the same type are in a "shared unit" which have the same clock speed. The A78C documentation is unclear about how it can be broken up in DSUs, as it contradicts itself several times, possibly because the documentation is not entirely updated from the base A78 variant.

 

[Gerald]

In devices where a78c was used if it make more power than a78?

 

[darthdiablo]

Cool!
Thanks.

Also just in case you weren't around for the spec reveal by LiC. Other example that is a possible "expectation surpassed" (which you asked about) - although probably not as much around here - is that there is 12GB RAM, when there were some (bad) leakers out there saying 8GB can be reasonably expected.

 

[Dakhil]

In devices where a78c was used if it make more power than a78?

Going by the Snapdragon 8cx Gen 3, laptops.

 

[Gerald]

Going by the Snapdragon 8cx Gen 3, laptops.

And the power consumption between a78c and a78 is the same?

 

[Armoured_Bear]

'Also just in case you weren't around for the spec reveal by LiC. Other example that is a possible "expectation surpassed" (which you asked about) - although probably not as much around here - is that there is 12GB RAM, when there were some (bad) leakers out there saying 8GB can be reasonably expected.

Thanks, I had gotten that, very relieved to hear, it's a great realistic amount.

 

[GuardianStalker]

Hello everyone. I have another technical question. If Nintendo and Nvidia continue their cooperation, what are the chances that a theoretical “Switch 3”, which may be released sometime in 203x, will be backwards compatible with the Switch and Switch 2?

The background to this question is my current thoughts on switching to digital games. The market share of digital players is getting bigger and bigger, and there is no real guarantee that a “Switch 3” will still use physical games. If this is not the case, then digital games might be the only way to use your Switch and Switch 2 games on the Switch 3.

 

[Hermii]

Also just in case you weren't around for the spec reveal by LiC. Other example that is a possible "expectation surpassed" (which you asked about) - although probably not as much around here - is that there is 12GB RAM, when there were some (bad) leakers out there saying 8GB can be reasonably expected.

After the recent revelation that 8gb Apple devices are really 12 gb, I'm not sure anything less than 12 was ever even a possibility.

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

Asking again after thinking a bit more about the situation: *** Hidden text: You do not have sufficient rights to view the hidden text. Visit the forum thread! ***

famiboards.com

 

[MetalLord]

wonder how nintendo artists are adapting to a higher power system ? a sneak peak :

and from another artist

 

[gonanzir]

Yes, but in those phones, the X1 core is the only X1 core. The DynamIQ documentation implies that all cores of the same type are in a "shared unit" which have the same clock speed. The A78C documentation is unclear about how it can be broken up in DSUs, as it contradicts itself several times, possibly because the documentation is not entirely updated from the base A78 variant.

https://www.mediatek.com/products/smartphone-2/mediatek-dimensity-8200

This supports 1* [email protected] ghz and 3*[email protected]

 

[gonanzir]

,

 

[carbvan]

wonder how nintendo artists are adapting to a higher power system ? a sneak peak :

Most of these are taken from an artstation account of someone going by miuraman. What's the point of this? Random creature designs with no environments or game context does not show the power of a system at all. You've basically just taken someone else's art and posted it for no reason. At least credit the guy: https://www.artstation.com/miuraman

 

[Hartmann]

It seems that many Korean leakers are saying that Nintendo is going to use Samsung.

Nintendo requested parts from Samsung. These parts appear to be for the second generation of the Nintendo Switch.

We are collaborating with Samsung Electronics to develop Nintendo Switch AP smoothly without any issues.

This is more interesting.

가변 120Hz는 비밀
Variable 120Hz is a secret

 

[Dakhil]

After the recent revelation that 8gb Apple devices are really 12 gb, I'm not sure anything less than 12 was ever even a possibility.

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

Asking again after thinking a bit more about the situation: *** Hidden text: You do not have sufficient rights to view the hidden text. Visit the forum thread! ***

famiboards.com

64-bit 4 GB (32 Gb) LPDDR5X modules do exist from Micron.

So I think Apple using two 64-bit 6 GB (48 Gb) LPDDR5X modules is primarily a cost saving measure.

Edit: Actually, all the 4 GB (32 Gb) LPDDR5X modules from Micron and Samsung turn out to be 32-bit only. So yeah, 128-bit 8 GB LPDDR5X with 4 GB (32 Gb) modules is never a realistic possibility.

 

[Mendinso]

Hello everyone. I have another technical question. If Nintendo and Nvidia continue their cooperation, what are the chances that a theoretical “Switch 3”, which may be released sometime in 203x, will be backwards compatible with the Switch and Switch 2?

The background to this question is my current thoughts on switching to digital games. The market share of digital players is getting bigger and bigger, and there is no real guarantee that a “Switch 3” will still use physical games. If this is not the case, then digital games might be the only way to use your Switch and Switch 2 games on the Switch 3.

CPU will be the biggest factor. If the ARM cores continue to maintain compatibility, then it won't be an issue. GPU will continue to need some level of emulation layer for shader cache but other than that, if the os retains all the appropriate calls, emulate whatever cpu or os calls are missing, and so on, compatibility should be feasible to maintain. As with anything, likely won't be 100% compatible with everything, due to cpu and or rendering bugs being possible but it'll be fairly close.

Cartridge slot could be a concern too but I think Nintendo will maintain a slot as long as the cartridges are consistent.

 

[gonanzir]

The Qualcomm Snapdragon 855 Pre-Dive: Going Into Detail on 2019's Flagship Android SoC

www.anandtech.com

When I looked up some information, I found that although some chips use DynamIQ, they indeed don't support individual core voltage and frequency adjustments, which is somewhat of an exception.

The highest performance core, “Kryo 485 Gold Prime” as Qualcomm calls it, is clocked in at 2.84GHz – putting it on its own clock domain – and is seemingly configured with a 512KB L2 cache. The other three cores are clocked at 2.42GHz and retain smaller 256KB L2 caches. This configuration is quite odd – you also would expect Qualcomm to take advantage of the new DynamIQ cluster design, which is able to support different frequency and voltage planes, however things get even odder. The prime core actually doesn’t have its own voltage plane, and thus it has to share its voltage plane with the other three big coresThis revelation of the prime core not having its own power domain is quite shocking and it invalidates a lot of the benefits of actually having a separate clock plane for a core. In effect the real-world benefit here isn’t any different than simply clock-gating the core.

It is true that there’s a large amount of scenarios where there’s predominantly a single larger thread active, this is particularly true in web browsing workloads. Such a 1+3 configuration would achieve better performance and possible better efficiency than a 2+2 configuration, but because the cores aren’t running on separate voltage planes it means the actual benefits here in real-world applications are just going to be quite minor. The net result is that the setup is leaving a lot of power efficiency on the table: the voltage supplied to both core groups is always going to be the greater of whatever is being asked for, even if one of the two groups could operate on (much) less voltage.

 

[Gerald]

It seems that many Korean leakers are saying that Nintendo is going to use Samsung.




This is more interesting.

It must be fake, since all parts about what we have info is not from samsung, screen is Innolux

 

[Hermii]

64-bit 4 GB (32 Gb) LPDDR5X modules do exist from Micron.

So I think Apple using two 64-bit 6 GB (48 Gb) LPDDR5X modules is primarily a cost saving measure.

it woudnt change my point. If Apple gets 12gb cheaper than 8gb, so would probably Nintendo. Hence 8gb was never a real possibility.

 

[MetalLord]

Most of these are taken from an artstation account of someone . What's the point of this? Random creature designs with no environments or game context does not show the power of a system at all. You've basically just taken someone else's art and posted it for no reason.

its someone who works at nintendo,using a pseudonym online, so dont direct link people here, they probably want to keep low profile,it had nothing to do with switch 2 power its just an example of what a nintendo artist can do with more power

 

[Dakhil]

it woudnt change my point. If Apple gets 12gb cheaper than 8gb, so would probably Nintendo. Hence 8gb was never a real possibility.

I've edited my original post. But I found out that all the 4 GB (32 Gb) LPDDR5X modules from Micron and Samsung are only 32-bit. So 128-bit 8 GB LPDDR5X with 4 GB (32 Gb) LPDDR5X modules is never a realistic possibility to begin with.

 

[darthdiablo]

It seems that many Korean leakers are saying that Nintendo is going to use Samsung.




This is more interesting.

Doesn’t look like anything here can be verified. Vague, non specific mention of parts, impossible to confirm anything here so I am not sure there is any value in those tweets to bother vetting for

Also the last tweet about variable 120, is that even about Nintendo? Looks like the twitter profile tweets stuff about Samsung in general

 

[Gerald]

Doesn’t look like anything here can be verified. Vague, non specific mention of parts, impossible to confirm anything here so I am not sure there is any value from those tweets

Also we know that screen is from innolux not samsung, is big bs

 

[Dakhil]

Also the last tweet about variable 120, is that even about Nintendo? Looks like the twitter profile tweets stuff about Samsung in general

The link the last post was replying to was specifically about Nintendo.

But as Gerald said, probably bullshit, especially since Samsung no longer produces LCD displays.

 

[Hermii]

Also we know that screen is from innolux not samsung, is big bs

We know the early devkits had innolux panels. We don't know that was the final panel.

 

[darthdiablo]

The link the last post was replying to was specifically about Nintendo.

That link just shows “deleted” and when I checked the last tweet it didn’t seem to be in response to anything - maybe it’s showing differently on desktop (I am on mobile currently)

We know the early devkits had innolux panels. We don't know that was the final panel.

As far as I know that is still unconfirmed. 7.9" Innolux TFT-LCD panels are visible in shipment data going to nvidia India around the same time T239 validation testing occurred, but nothing in those entries specifically ties those to T239 or anything Nintendo.

Personally I choose to believe it's what devkits are using yeah.

 

[Dakhil]

That link just shows “deleted” and when I checked the last tweet it didn’t seem to be in response to anything - maybe it’s showing differently on desktop (I am on mobile currently)

I copied and pasted the wrong link. The link should work now.

---

We don't know that was the final panel.

Yes. But Samsung's definitely not the supplier, especially since Samsung no longer produces LCD displays.

 

[ItWasMeantToBe19]

Samsung provides parts in a ton of different areas that this "leak" doesn't seem very informative or brave, lol.

 

[darthdiablo]

I copied and pasted the wrong link. The link should work now.

---

Yes. But Samsung's definitely not the supplier, especially since Samsung no longer produces LCD displays.

Ah I see it now.

Yeah, I think this thread have seen that article before and most are dismissing the article, because it mentions Switch 2 is scheduled to be released as early as 2H of this year, and also mentions Samsung 7LPH as the node process

 

[Hermii]

Yes. But Samsung's definitely not the supplier, especially since Samsung no longer produces LCD displays.

For all we know the source of the lcd reporting are those early devkits with a non final panel.

Maybe reaching here, but I would say there's a non zero chance.

 

[Danny]

This is just a repeat from a similar rumor from 1 month ago that Nintendo supposedly will be using Samsung for almost everything when it comes to Switch 2. Doesn't make sense, no console hardware maker would just use 1 supplier but would use many different suppliers to reduce excessive dependance on 1 supplier.