• 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.
  • Do you have audio editing experience and want to help out with the Famiboards Discussion Club Podcast? If so, we're looking for help and would love to have you on the team! Just let us know in the Podcast Thread if you are interested!

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

Reader mode
majimazio said:
Click to expand...

Why use this render when the Switch 2 got leaked already

Mega_95_w_Dock_Media_Image_Smaller.jpg
 
I hope nobody minds if I go on an absurd flight-of-fancy about something that's extremely unlikely, but it's been on my list of out-there hypotheticals for literally years, so I felt like I had to go through the process on it.

Does Switch 2 use Samsung's new LLW DRAM memory?
(Spoiler: Betteridge's law applies here)

Samsung just announced a new type of memory called Low-Latency Wide I/O DRAM, or LLW DRAM. It's aimed at portable use-cases, offers high bandwidth (128GB/s per chip) and very low power draw. From what I can tell, it appears to be similar to HBM, which is backed up by their promo material.

The reason I find this so interesting is that I've been saying for years (see here, here, etc.) that a low-power RAM based on HBM would be the ideal, and possibly inevitable, solution for portable gaming devices like the Switch going forward. And here, right at the start of the year of Switch 2, Samsung goes ahead and announces a low-power RAM based on HBM with gaming as the only non-mandatory-AI-buzzword use-case tagged on their tweet.

Firstly, I'd like to go through why HBM-style memory is so useful for a device like the Switch, as briefly as I can (which, let's be honest, usually isn't very brief!). The big issue here is that a portable gaming device like the Switch is always going to be limited by power consumption. There's going to be a maximum size battery they can fit in the device, and a maximum power draw they can sustain while getting decent life out of that battery. Improvements in battery technology are pretty slow, so improvements in performance from generation to generation are mostly going to come from getting more out of that limited power draw, ie improvements in power efficiency.

On the RAM side of things, this presents an issue, because the demand for bandwidth is growing at a greater rate than the power efficiency of that bandwidth. So, for the jump from Switch to Switch 2, we're looking at moving from 25.6GB/s of bandwidth to an expected 102.4GB/s of bandwidth, which is a 4x increase. However, from what I can tell power efficiency of LPDDR ram has only increased by around 2x over that time. So, if Nintendo want 4x the bandwidth, but only have 2x the efficiency, they need to allocate twice as much power to RAM to keep up. On Switch 2 they can likely just about do so, but at some point something's got to give, and they'll need an alternative to LPDDR.

This is where HBM-style memory comes in. HBM is stacked RAM which sits on an interposer next to the GPU/SoC. This gives it two big advantages in terms of power efficiency. The first is that it's a wide and slow standard. Instead of 16 to 64 bit interfaces, HBM modules have 1024-bit interfaces, which would be impractical to implement as traces over a motherboard, but is easily done on an interposer. The second is that, as the connection between memory controller and memory is only over an interposer, not motherboard traces, the connection itself consumes less power. Actual HBM used on products like Hopper still consumes quite a bit of power, but that's at high clocks, delivering massive amounts of bandwidth, and a scaled down "low-power HBM" version designed for mobile devices should manage to beat out LPDDR on both bandwidth and power efficiency by good margins. The downside, of course, being cost, which will be higher by virtue of both the stacked RAM and the need for packaging alongside the SoC on an interposer.

Samsung's LLW DRAM appears to be pretty much exactly the kind of low-power HBM that I was expecting, and it's also almost suspiciously well-suited to Switch 2. It offers 128GB/s of bandwidth, which is higher than the 102GB/s we were expecting, but not by a crazy amount, and at 1.2 pJ/b of claimed power consumption, at peak bandwidth that would come to only 1.2W. For reference, from what I can tell the LPDDR4 in the original Switch consumed about 1.5W for 25.6GB/s, and I've been expecting about 3W for RAM in docked mode on Switch 2 for 102GB/s, so coming in under Switch's LPDDR4 in power while offering 5x the performance would be pretty damn nice. It's in an almost perfect sweet spot for Switch 2's performance and power envelope.

It also might be based on a Nvidia proposal for a low-power variant of HBM, or I might be reaching way too far with my analysis there, but two years ago when I was writing about this, I speculated about Samsung and Nvidia partnering on a low-power HBM-based memory for a future Switch, based on Nvidia's FGDRAM, and LLW DRAM is eerily close to that. I was off by a little bit on power efficiency (1.5 pJ/b vs 1.2 pJ/b), and expected slightly higher total power consumption and bandwidth, but the actual part Samsung have produced seems a better fit for Switch 2 than my speculation (which would have been overkill on bandwidth).

Does the evidence we have for Switch 2 fit LLW DRAM?

Maybe if you want to really stretch things, but not really.

Our first evidence of the memory interface on T239 came from the Nvidia hack, where T239 was listed as having 1 framebuffer partition (ie memory controller), compared to 2 on T234 (Orin). Orin has a 256-bit LPDDR5 memory interface, split into two 128-bit memory controllers, so if T239 has one memory controller, it stands to reason that it has a 128-bit LPDDR5 memory interface.

To play devil's advocate here, LLW DRAM would also very likely involve one memory controller. On Nvidia's HPC GPUs that use HBM, like Hopper, they use two memory controllers per 1024-bit HBM module, with a 512-bit interface per memory controller. As LLW DRAM will have a narrower interface than HBM, likely either 256-bit or 512-bit, it seems reasonable to expect Nvidia would use a single memory controller per chip.

However, what's not that reasonable is to expect this vastly different RAM type to Orin's LPDDR5 seemingly show up nowhere in the hack. It reportedly states that LPDDR5 is used on "T23X and later", and it would be very weird for T239 not to count as "T23X and later". It would be doubly weird for a brand new and completely unique memory type (which would be arguably the most unique feature of the chip) not to get a mention anywhere.

Another piece of evidence we have on T239's memory interface is this Nvidia Linux commit from December 2021. This states that "T239 has 8 number of mc channels while orin has 16 number of mc channels". That is, half as many memory controller channels as Orin, therefore reinforcing the info from the hack, that's it's a 128-bit LPDDR5 memory interface. Strictly speaking, this commit relates to pre-silicon testing, and is actually referring to LPDDR4, rather than LPDDR5, presumably for testing sake. Of course it makes sense to do pre-silicon testing with a very similar memory type like LPDDR4. It doesn't make a whole lot of sense to do that if the chip you're testing uses a memory that's nothing like LPDDR4. This file doesn't seem to still exist in Nvidia's public repos (or it's been moved somewhere I can't find), so I don't know if it's been updated to LPDDR5 since silicon hit.

Edit: And for one extra nail in the coffin while I was typing this up, @karmitt's post on the last page seems to be extremely clear on LPDDR5/X.


So, LLW DRAM is almost certainly not being used in Switch 2, and that's kind of a shame. Not because it would allow for a big jump in performance, as it would actually come in at lower bandwidth than LPDDR5X, which is both more plausible and cheaper (although the lower power consumption may allow for a bit more power to be diverted to other components like CPU and GPU in portable mode). The real reason it's a shame is that it's exactly the kind of weird, unique technology we used to get in consoles back in the day, like when Sony would design an entire new CPU architecture from scratch for the PS3, or Nintendo would use some obscure memory tech, like in most of their hardware before the Switch. We don't really get that much any more.

That's not a bad thing overall, as designing completely new CPUs, GPUs or memory standards is way too expensive to do for a gaming console, and there are too many great options available off the shelf to go full Ken Kutaragi on a console, but as console hardware has become more standardised, it's become safer, and more predictable, and a bit more boring. The PS5 and XBSS/X were pretty predictable pieces of hardware all-in-all, with standard AMD CPUs and GPUs, standard GDDR memory, and relatively standard SSDs. The precise amount of GPU performance, or RAM or SSD bandwidth was up for debate, but the general architectures involved were never going to be surprising. The PS6 and next Xbox will almost certainly be the same, with standard AMD architectures and GDDR RAM wrapped up in the usual buzz-words that make it sound more custom than it is.

The Switch 2 is also looking remarkably sensible. After the shock on the GPU size when we learnt it from the hack, when you think about it as a 4N SoC, and with 4N being a pretty sensible choice in 2024, it's really quite a reasonable design. CPU and GPU architectures that are not quite the newest, but with very close performance, a CPU that aligns more closely to other consoles (8 cores) and a GPU that fits well within a small SoC, and can achieve high performance-per-Watt at Nintendo's target power levels. And then an LPDDR5 memory interface that provides bandwidth that's well-matched to the GPU's performance. The FDE is the one unique component, but as both Sony and MS had adopted dedicated decompression hardware it was hardly surprising.

A sensibly designed Switch 2 is the best case for people looking for a capable, reasonably priced piece of hardware, but a part of me wishes they would have gone a little crazy and used LLW RAM (although it was almost certainly developed too late to be an option). Just for a bit of the old mystery you used to get with gaming hardware, like trying to figure out what the hell 1T-SRAM was when Gamecube was announced. I'd say LLW RAM, or LLW2 or whatever becomes of it, is a pretty decent bet for whatever comes after Switch 2, but by that time it'll be the sensible choice, not the unique weird tech it would have been if it were used in Switch 2.
 
Last edited:
karmitt said:
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.

Click to expand...
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.

RedSpring said:
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
Click to expand...
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
 
Scrawnton said:
I feel this way because I feel like every time Nintendo has released a system, we have gone through the hype cycle of what it will be capable of and when the system actually releases the other shoe drops and we have to dial back our expectations and understanding of how powerful the system actually is and what it is really capable of. The reality sets it. This obviously happened with Wii, it happened with Wii U, 3DS was a great upgrade over DS but was a massive step down fro Vita in regards to power, and I remember Switch speculation before launch people were hyped for an Xbox One like device and we had to dial that one back big time. I just feel like we are walking into the same exact situation yet again.
Click to expand...
Thing about Switch 1 is many people incorrectly assumed they'd be getting a TX1 customized to their needs rather than something directly off the shelf. This time, we know they're getting that customized chip, and mostly in what ways.
 
Thraktor said:
I hope nobody minds if I go on an absurd flight-of-fancy about something that's extremely unlikely, but it's been on my list of out-there hypotheticals for literally years, so I felt like I had to go through the process on it.

Does Switch 2 use Samsung's new LLW DRAM memory?
(Spoiler: Betteridge's law applies here)

Samsung just announced a new type of memory called Low-Latency Wide I/O DRAM, or LLW DRAM. It's aimed at portable use-cases, offers high bandwidth (128GB/s per chip) and very low power draw. From what I can tell, it appears to be similar to HBM, which is backed up by their promo material.

The reason I find this so interesting is that I've been saying for years (see here, here, etc.) that a low-power RAM based on HBM would be the ideal, and possibly inevitable, solution for portable gaming devices like the Switch going forward. And here, right at the start of the year of Switch 2, Samsung goes ahead and announces a low-power RAM based on HBM with gaming as the only non-mandatory-AI-buzzword use-case tagged on their tweet.

Firstly, I'd like to go through why HBM-style memory is so useful for a device like the Switch, as briefly as I can (which, let's be honest, usually isn't very brief!). The big issue here is that a portable gaming device like the Switch is always going to be limited by power consumption. There's going to be a maximum size battery they can fit in the device, and a maximum power draw they can sustain while getting decent life out of that battery. Improvements in battery technology are pretty slow, so improvements in performance from generation to generation are mostly going to come from getting more out of that limited power draw, ie improvements in power efficiency.

On the RAM side of things, this presents an issue, because the demand for bandwidth is growing at a greater rate than the power efficiency of that bandwidth. So, for the jump from Switch to Switch 2, we're looking at moving from 25.6GB/s of bandwidth to an expected 102.4GB/s of bandwidth, which is a 4x increase. However, from what I can tell power efficiency of LPDDR ram has only increased by around 2x over that time. So, if Nintendo want 4x the bandwidth, but only have 2x the efficiency, they need to allocate twice as much power to RAM to keep up. On Switch 2 they can likely just about do so, but at some point something's got to give, and they'll need an alternative to LPDDR.

This is where HBM-style memory comes in. HBM is stacked RAM which sits on an interposer next to the GPU/SoC. This gives it two big advantages in terms of power efficiency. The first is that it's a wide and slow standard. Instead of 16 to 64 bit interfaces, HBM modules have 1024-bit interfaces, which would be impractical to implement as traces over a motherboard, but is easily done on an interposer. The second is that, as the connection between memory controller and memory is only over an interposer, not motherboard traces, the connection itself consumes less power. Actual HBM used on products like Hopper still consumes quite a bit of power, but that's at high clocks, delivering massive amounts of bandwidth, and a scaled down "low-power HBM" version designed for mobile devices should manage to beat out LPDDR on both bandwidth and power efficiency by good margins. The downside, of course, being cost, which will be higher by virtue of both the stacked RAM and the need for packaging alongside the SoC on an interposer.

Samsung's LLW DRAM appears to be pretty much exactly the kind of low-power HBM that I was expecting, and it's also almost suspiciously well-suited to Switch 2. It offers 128GB/s of bandwidth, which is higher than the 102GB/s we were expecting, but not by a crazy amount, and at 1.2 pJ/b of claimed power consumption, at peak bandwidth that would come to only 1.2W. For reference, from what I can tell the LPDDR4 in the original Switch consumed about 1.5W for 25.6GB/s, and I've been expecting about 3W for RAM in docked mode on Switch 2 for 102GB/s, so coming in under Switch's LPDDR4 in power while offering 5x the performance would be pretty damn nice. It's in an almost perfect sweet spot for Switch 2's performance and power envelope.

It also might be based on a Nvidia proposal for a low-power variant of HBM, or I might be reaching way too far with my analysis there, but two years ago when I was writing about this, I speculated about Samsung and Nvidia partnering on a low-power HBM-based memory for a future Switch, based on Nvidia's FGDRAM, and LLW DRAM is eerily close to that. I was off by a little bit on power efficiency (1.5 pJ/b vs 1.2 pJ/b), and expected slightly higher total power consumption and bandwidth, but the actual part Samsung have produced seems a better fit for Switch 2 than my speculation (which would have been overkill on bandwidth).

Does the evidence we have for Switch 2 fit LLW DRAM?

Maybe if you want to really stretch things, but not really.

Our first evidence of the memory interface on T239 came from the Nvidia hack, where T239 was listed as having 1 framebuffer partition (ie memory controller), compared to 2 on T234 (Orin). Orin has a 256-bit LPDDR5 memory interface, split into two 128-bit memory controllers, so if T239 has one memory controller, it stands to reason that it has a 128-bit LPDDR5 memory interface.

To play devil's advocate here, LLW DRAM would also very likely involve one memory controller. On Nvidia's HPC GPUs that use HBM, like Hopper, they use two memory controllers per 1024-bit HBM module, with a 512-bit interface per memory controller. As LLW DRAM will have a narrower interface than HBM, likely either 256-bit or 512-bit, it seems reasonable to expect Nvidia would use a single memory controller per chip.

However, what's not that reasonable is to expect this vastly different RAM type to Orin's LPDDR5 seemingly show up nowhere in the hack. It reportedly states that LPDDR5 is used on "T23X and later", and it would be very weird for T239 not to count as "T23X and later". It would be doubly weird for a brand new and completely unique memory type (which would be arguably the most unique feature of the chip) not to get a mention anywhere.

Another piece of evidence we have on T239's memory interface is this Nvidia Linux commit from December 2021. This states that "T239 has 8 number of mc channels while orin has 16 number of mc channels". That is, half as many memory controller channels as Orin, therefore reinforcing the info from the hack, that's it's a 128-bit LPDDR5 memory interface. Strictly speaking, this commit relates to pre-silicon testing, and is actually referring to LPDDR4, rather than LPDDR5, presumably for testing sake. Of course it makes sense to do pre-silicon testing with a very similar memory type like LPDDR4. It doesn't make a whole lot of sense to do that if the chip you're testing uses a memory that's nothing like LPDDR4. This file doesn't seem to still exist in Nvidia's public repos (or it's been moved somewhere I can't find), so I don't know if it's been updated to LPDDR5 since silicon hit.

Edit: And for one extra nail in the coffin while I was typing this up, @karmitt's post on the last page seems to be extremely clear on LPDDR5/X.


So, LLW DRAM is almost certainly not being used in Switch 2, and that's kind of a shame. Not because it would allow for a big jump in performance, as it would actually come in at lower bandwidth than LPDDR5X, which is both more plausible and cheaper (although the lower power consumption may allow for a bit more power to be diverted to other components like CPU and GPU in portable mode). The real reason it's a shame is that it's exactly the kind of weird, unique technology we used to get in consoles back in the day, like when Sony would design an entire new CPU architecture from scratch for the PS3, or Nintendo would use some obscure memory tech, like in most of their hardware before the Switch. We don't really get that much any more.

That's not a bad thing overall, as designing completely new CPUs, GPUs or memory standards is way too expensive to do for a gaming console, and there are too many great options available off the shelf to go full Ken Kutaragi on a console, but as console hardware has become more standardised, it's become safer, and more predictable, and a bit more boring. The PS5 and XBSS/X were pretty predictable pieces of hardware all-in-all, with standard AMD CPUs and GPUs, standard GDDR memory, and relatively standard SSDs. The precise amount of GPU performance, or RAM or SSD bandwidth was up for debate, but the general architectures involved were never going to be surprising. The PS6 and next Xbox will almost certainly be the same, with standard AMD architectures and GDDR RAM wrapped up in the usual buzz-words that make it sound more custom than it is.

The Switch 2 is also looking remarkably sensible. After the shock on the GPU size when we learnt it from the hack, when you think about it as a 4N SoC, and with 4N being a pretty sensible choice in 2024, it's really quite a reasonable design. CPU and GPU architectures that are not quite the newest, but with very close performance, a CPU that aligns more closely to other consoles (8 cores) and a GPU that fits well within a small SoC, and can achieve high performance-per-Watt at Nintendo's target power levels. And then an LPDDR5 memory interface that provides bandwidth that's well-matched to the GPU's performance. The FDE is the one unique component, but as both Sony and MS had adopted dedicated decompression hardware it was hardly surprising.

A sensibly designed Switch 2 is the best case for people looking for a capable, reasonably priced piece of hardware, but a part of me wishes they would have gone a little crazy and used LLW RAM (although it was almost certainly developed too late to be an option). Just for a bit of the old mystery you used to get with gaming hardware, like trying to figure out what the hell 1T-SRAM was when Gamecube was announced. I'd say LLW RAM, or LLW2 or whatever becomes of it, is a pretty decent bet for whatever comes after Switch 2, but by that time it'll be the sensible choice, not the unique weird tech it would have been if it were used in Switch 2.
Click to expand...
This is really interesting! I wonder if Nintendo might do a Switch 2 Pro in 2027/2028?
 
LiC said:
* Hidden text: cannot be quoted. *


* Hidden text: cannot be quoted. *
Click to expand...

Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
 
Last edited:
Thraktor said:
Yup, it was barely mentioned by Microsoft in their PR for the console, and includes support for a custom compression algorithm called BCPACK which is specifically designed to compress texture data, and as such likely achieves much better compression ratios than PS5's Kraken, which is a general-purpose compression algorithm.
Click to expand...
A lot of developers must not support it as I see PS5 games comes out smaller than the Series counterpart.
 
oldpuck said:
TL;DR: I'd take the RAM. Which I was going to say when you asked, but then I had to check a bunch of benchmarks to be sure, because I Have A Problem(tm). :ROFLMAO:

This is another place where Steam Deck changed my mind. The OLED I bought has a nice bandwidth bump, and yes, it does improve performance, and it doesn't matter. It smooths out dropped frames in some games, in some places, but it doesn't make them go away. And it doesn't give a high enough performance bump for you to up the frame cap or increase visual quality.

The extra RAM means you that regardless of the frame rate, or even the resolution you're looking at highest quality textures. That's a definite win. But AMD and Nvidia are obviously different architectures, which is why I had to check benchmarks.

The 3070 is one of the more bandwidth starved cards in the RTX 30 line up. The 3070 Ti has only 6% more compute power, but it's got a whopping 35% more memory bandwidth, putting it on the high end for bandwidth. Digital Foundry has a bunch of benchmarks for these cards, and... it's 6% faster. Slightly higher on 4k games, but lower on 1080p games. Same on the RT benchmarks.

The 3080 has a 12GB version that is much the same. Only 3% different in TFLOPS, but a 20% increase in memory bandwidth... and a 6% improvements in actual games, disappearing quickly as you drop to 1080p.

I'm sure that software developers optimizing for the hardware could do amazing things with the extra bandwidth. But it doesn't look like existing engines are really hitting bandwidth limits on Ampere hardware, so staying in line with the rest of the RTX 30 series seems like 3rd parties will be in great shape. And as for first party stuff, Nintendo has the most bandwidth optimized engine on the market. What they're doing with Tears of the Kingdom and 25GB/s of bandwidth is insane.

Side Note: Since I have all these benchmarks and specs in a spreadsheet (I SAID I HAVE A PROBLEM AND THE FIRST STEP IS RECOGNIZING IT) I decided to look at the less sexy parts of the architecture. ROPS, TMUS, and the L2 cache. All of these systems interact in various ways to create the final efficiency of the system.

There are folks hoping for 4 MB of L2 cache, and not the 1 MB (as the leak is ambiguous). If you look at cache as a proportion of the memory bandwidth available, 1 MB is already more than any desktop card. 4 MB would be beyond generous, and likely pretty expensive. 1MB is already luxurious.

Texture mapping units are part of the SM design, so the ratio there always matches. Sufficient or insufficient, there is no way to tell, because it's locked into the Ampere/Lovelace design.

ROPs are a little different. They're by GPC, so sometimes you get extra ROPS relative to SMs after binning. With just one GPC, that isn't happening on T239, but it also doesn't seem to matter, Performance doesn't seem to track with ROPS in a way that indicates it would be a problem.
Click to expand...

During the whole crypto craze, I overpaid for an RTX 3060, which at the time I felt I wanted the 3060 ti, or perhaps the 3070.

Now though, after hearing how starved those other GPUs are, even in some cases that Hardware Unboxed reported, the 3060 would pull ahead in performance because of the higher amount of Ram, despite the bandwidth being much lower.

I now no longer have regrets in buying the 3060.


Thraktor said:
I hope nobody minds if I go on an absurd flight-of-fancy about something that's extremely unlikely, but it's been on my list of out-there hypotheticals for literally years, so I felt like I had to go through the process on it.

Does Switch 2 use Samsung's new LLW DRAM memory?
(Spoiler: Betteridge's law applies here)

Samsung just announced a new type of memory called Low-Latency Wide I/O DRAM, or LLW DRAM. It's aimed at portable use-cases, offers high bandwidth (128GB/s per chip) and very low power draw. From what I can tell, it appears to be similar to HBM, which is backed up by their promo material.

The reason I find this so interesting is that I've been saying for years (see here, here, etc.) that a low-power RAM based on HBM would be the ideal, and possibly inevitable, solution for portable gaming devices like the Switch going forward. And here, right at the start of the year of Switch 2, Samsung goes ahead and announces a low-power RAM based on HBM with gaming as the only non-mandatory-AI-buzzword use-case tagged on their tweet.

Firstly, I'd like to go through why HBM-style memory is so useful for a device like the Switch, as briefly as I can (which, let's be honest, usually isn't very brief!). The big issue here is that a portable gaming device like the Switch is always going to be limited by power consumption. There's going to be a maximum size battery they can fit in the device, and a maximum power draw they can sustain while getting decent life out of that battery. Improvements in battery technology are pretty slow, so improvements in performance from generation to generation are mostly going to come from getting more out of that limited power draw, ie improvements in power efficiency.

On the RAM side of things, this presents an issue, because the demand for bandwidth is growing at a greater rate than the power efficiency of that bandwidth. So, for the jump from Switch to Switch 2, we're looking at moving from 25.6GB/s of bandwidth to an expected 102.4GB/s of bandwidth, which is a 4x increase. However, from what I can tell power efficiency of LPDDR ram has only increased by around 2x over that time. So, if Nintendo want 4x the bandwidth, but only have 2x the efficiency, they need to allocate twice as much power to RAM to keep up. On Switch 2 they can likely just about do so, but at some point something's got to give, and they'll need an alternative to LPDDR.

This is where HBM-style memory comes in. HBM is stacked RAM which sits on an interposer next to the GPU/SoC. This gives it two big advantages in terms of power efficiency. The first is that it's a wide and slow standard. Instead of 16 to 64 bit interfaces, HBM modules have 1024-bit interfaces, which would be impractical to implement as traces over a motherboard, but is easily done on an interposer. The second is that, as the connection between memory controller and memory is only over an interposer, not motherboard traces, the connection itself consumes less power. Actual HBM used on products like Hopper still consumes quite a bit of power, but that's at high clocks, delivering massive amounts of bandwidth, and a scaled down "low-power HBM" version designed for mobile devices should manage to beat out LPDDR on both bandwidth and power efficiency by good margins. The downside, of course, being cost, which will be higher by virtue of both the stacked RAM and the need for packaging alongside the SoC on an interposer.

Samsung's LLW DRAM appears to be pretty much exactly the kind of low-power HBM that I was expecting, and it's also almost suspiciously well-suited to Switch 2. It offers 128GB/s of bandwidth, which is higher than the 102GB/s we were expecting, but not by a crazy amount, and at 1.2 pJ/b of claimed power consumption, at peak bandwidth that would come to only 1.2W. For reference, from what I can tell the LPDDR4 in the original Switch consumed about 1.5W for 25.6GB/s, and I've been expecting about 3W for RAM in docked mode on Switch 2 for 102GB/s, so coming in under Switch's LPDDR4 in power while offering 5x the performance would be pretty damn nice. It's in an almost perfect sweet spot for Switch 2's performance and power envelope.

It also might be based on a Nvidia proposal for a low-power variant of HBM, or I might be reaching way too far with my analysis there, but two years ago when I was writing about this, I speculated about Samsung and Nvidia partnering on a low-power HBM-based memory for a future Switch, based on Nvidia's FGDRAM, and LLW DRAM is eerily close to that. I was off by a little bit on power efficiency (1.5 pJ/b vs 1.2 pJ/b), and expected slightly higher total power consumption and bandwidth, but the actual part Samsung have produced seems a better fit for Switch 2 than my speculation (which would have been overkill on bandwidth).

Does the evidence we have for Switch 2 fit LLW DRAM?

Maybe if you want to really stretch things, but not really.

Our first evidence of the memory interface on T239 came from the Nvidia hack, where T239 was listed as having 1 framebuffer partition (ie memory controller), compared to 2 on T234 (Orin). Orin has a 256-bit LPDDR5 memory interface, split into two 128-bit memory controllers, so if T239 has one memory controller, it stands to reason that it has a 128-bit LPDDR5 memory interface.

To play devil's advocate here, LLW DRAM would also very likely involve one memory controller. On Nvidia's HPC GPUs that use HBM, like Hopper, they use two memory controllers per 1024-bit HBM module, with a 512-bit interface per memory controller. As LLW DRAM will have a narrower interface than HBM, likely either 256-bit or 512-bit, it seems reasonable to expect Nvidia would use a single memory controller per chip.

However, what's not that reasonable is to expect this vastly different RAM type to Orin's LPDDR5 seemingly show up nowhere in the hack. It reportedly states that LPDDR5 is used on "T23X and later", and it would be very weird for T239 not to count as "T23X and later". It would be doubly weird for a brand new and completely unique memory type (which would be arguably the most unique feature of the chip) not to get a mention anywhere.

Another piece of evidence we have on T239's memory interface is this Nvidia Linux commit from December 2021. This states that "T239 has 8 number of mc channels while orin has 16 number of mc channels". That is, half as many memory controller channels as Orin, therefore reinforcing the info from the hack, that's it's a 128-bit LPDDR5 memory interface. Strictly speaking, this commit relates to pre-silicon testing, and is actually referring to LPDDR4, rather than LPDDR5, presumably for testing sake. Of course it makes sense to do pre-silicon testing with a very similar memory type like LPDDR4. It doesn't make a whole lot of sense to do that if the chip you're testing uses a memory that's nothing like LPDDR4. This file doesn't seem to still exist in Nvidia's public repos (or it's been moved somewhere I can't find), so I don't know if it's been updated to LPDDR5 since silicon hit.

Edit: And for one extra nail in the coffin while I was typing this up, @karmitt's post on the last page seems to be extremely clear on LPDDR5/X.


So, LLW DRAM is almost certainly not being used in Switch 2, and that's kind of a shame. Not because it would allow for a big jump in performance, as it would actually come in at lower bandwidth than LPDDR5X, which is both more plausible and cheaper (although the lower power consumption may allow for a bit more power to be diverted to other components like CPU and GPU in portable mode). The real reason it's a shame is that it's exactly the kind of weird, unique technology we used to get in consoles back in the day, like when Sony would design an entire new CPU architecture from scratch for the PS3, or Nintendo would use some obscure memory tech, like in most of their hardware before the Switch. We don't really get that much any more.

That's not a bad thing overall, as designing completely new CPUs, GPUs or memory standards is way too expensive to do for a gaming console, and there are too many great options available off the shelf to go full Ken Kutaragi on a console, but as console hardware has become more standardised, it's become safer, and more predictable, and a bit more boring. The PS5 and XBSS/X were pretty predictable pieces of hardware all-in-all, with standard AMD CPUs and GPUs, standard GDDR memory, and relatively standard SSDs. The precise amount of GPU performance, or RAM or SSD bandwidth was up for debate, but the general architectures involved were never going to be surprising. The PS6 and next Xbox will almost certainly be the same, with standard AMD architectures and GDDR RAM wrapped up in the usual buzz-words that make it sound more custom than it is.

The Switch 2 is also looking remarkably sensible. After the shock on the GPU size when we learnt it from the hack, when you think about it as a 4N SoC, and with 4N being a pretty sensible choice in 2024, it's really quite a reasonable design. CPU and GPU architectures that are not quite the newest, but with very close performance, a CPU that aligns more closely to other consoles (8 cores) and a GPU that fits well within a small SoC, and can achieve high performance-per-Watt at Nintendo's target power levels. And then an LPDDR5 memory interface that provides bandwidth that's well-matched to the GPU's performance. The FDE is the one unique component, but as both Sony and MS had adopted dedicated decompression hardware it was hardly surprising.

A sensibly designed Switch 2 is the best case for people looking for a capable, reasonably priced piece of hardware, but a part of me wishes they would have gone a little crazy and used LLW RAM (although it was almost certainly developed too late to be an option). Just for a bit of the old mystery you used to get with gaming hardware, like trying to figure out what the hell 1T-SRAM was when Gamecube was announced. I'd say LLW RAM, or LLW2 or whatever becomes of it, is a pretty decent bet for whatever comes after Switch 2, but by that time it'll be the sensible choice, not the unique weird tech it would have been if it were used in Switch 2.
Click to expand...

Sorry for sending down such a rabbit hole, but I’d be lying if I said I don’t enjoy the kinds of deep dives you provide on such nerdy topics.

Maybe you’re correct that this particular kind of memory is far too new to be used, but would be useful for a future Nintendo system.
 
Derachi said:
Cloud gaming is good as an optional way to interact with an existing game/ecosystem. Like it rules I can play my Xbox Series X games on my phone remotely/through the cloud. It would never work as the only way to interact with a game/ecosystem.
Click to expand...
Yeah just recently set up my ps5 and Xbox to remote play to on my steamdeck. It's pretty great
 
Shubshub said:
Is someone able to better elaborate on what T239 really means though?

I feel like we've been hearing that code name thrown around for a few years now even back to the time of the alleged Switch Pro

Is T239 supposed to be a suped up Tegra X1? Is it a successor chip to the X1? Is it an entirely new line of Tegra chips? Where does it fall on the list of Tegra devices?
Click to expand...
Rich's article was linked to and it's excellent, but quick summary: it's a totally custom chip for Nintendo, developed by the Tegra team.

Nvidia essentially had two branches of the Tegra line, the chips for consumer devices (TX1) and chips for robotics (Xavier). The consumer chips never really panned out for any customer but Nintendo, and the robotics chips were only really a hit in the automotive sector. Orin, the "current" Tegra chip is a spiritual successor to Xavier, now very specifically customized to the auto industry.

T239, also called Drake, is the spiritual successor to the TX1, but it's totally new technology, heavily customized for Nintendo's use. It has custom hardware decompressing game assets quickly, It has a CPU cluster designed for high performance "always on" devices, instead of the "short bursts" of activity like phones and tablets. It has a very large memory bus to support heavy graphics applications, and it has a very large GPU, about twice the size of the Xbox One's GPU, but with all the modern features of Nvidia's RTX 30 cards. That means ray tracing and DLSS.

Codename_Steam said:
Did the GPU of the INDY have any eDRAM? Was it literally a Wii U GPU cut in half?
Click to expand...
It did not, IIRC. It wasn't a Wii U GPU cut in half, more like 1/3. The Wii U GPU was ... strange. It was, essentially, two GPUs slapped together. One was the Wii's GPU, and the second was a more modern GPU by the same team. The GPU module all combined was called "Latte"

The INDY GPU was called "Decaf" and it removed the Wii/GameCube backwards compatibility GPU, and was targeting about half the performance of the Wii U's "modern" GPU component.

Shubshub said:
I dont really think Nintendo needs to be too worried about competing directly with Steam Deck in terms of raw specs
Click to expand...
I would just add that there is really now way Nintendo doesn't blow the Steam Deck out of the water... when plugged into the TV. I wouldn't be surprised if it slightly underperforms the Steam Deck as a purely handheld device, but makes up for it with it's modern architecture and bespoke ports.

Shubshub said:
Maybe they are assuming that NVIDIA had to offer a deal to Nintendo to supply the chip for the console because other console manufacturers didnt want to touch NVIDIA after PS3
Click to expand...
It's a little unclear, but there is evidence that Nintendo got a deal from Nvidia, but it wasn't on the chips. It was on software. Nvidia's Tegra team needed a win, and TX1's big product was supposed to be the Google Pixel C. The Pixel C came out, but it was a boondoggle for Google, who abandoned the original plans. The Tegra team made a huge pitch to Nintendo.

Nintendo didn't want to leave their custom GPU, because they had a huge software investment in it. In fact, that's what Indy was all about - rebuilding the handhelds around the same GPU as the TV consoles, so they could reuse the highly optimized software stack they'd been building since the N64 (the N64, the GameCube, the Wii and the Wii U all had their GPUs built by the same team).

So Nvidia built a prototype version of a replacement for that software stack as part of their pitch to Nintendo. Obviously Nvidia knows their own hardware very well. We don't have the actual details of the deal that ultimately got signed, but it seems like Nintendo paid good money for the chips, which was very profitable for Nvidia - but Nintendo got a massive software development, on a super rushed time scale, for free.
 
oldpuck said:
Rich's article was linked to and it's excellent, but quick summary: it's a totally custom chip for Nintendo, developed by the Tegra team.

Nvidia essentially had two branches of the Tegra line, the chips for consumer devices (TX1) and chips for robotics (Xavier). The consumer chips never really panned out for any customer but Nintendo, and the robotics chips were only really a hit in the automotive sector. Orin, the "current" Tegra chip is a spiritual successor to Xavier, now very specifically customized to the auto industry.

T239, also called Drake, is the spiritual successor to the TX1, but it's totally new technology, heavily customized for Nintendo's use. It has custom hardware decompressing game assets quickly, It has a CPU cluster designed for high performance "always on" devices, instead of the "short bursts" of activity like phones and tablets. It has a very large memory bus to support heavy graphics applications, and it has a very large GPU, about twice the size of the Xbox One's GPU, but with all the modern features of Nvidia's RTX 30 cards. That means ray tracing and DLSS.


It did not, IIRC. It wasn't a Wii U GPU cut in half, more like 1/3. The Wii U GPU was ... strange. It was, essentially, two GPUs slapped together. One was the Wii's GPU, and the second was a more modern GPU by the same team. The GPU module all combined was called "Latte"

The INDY GPU was called "Decaf" and it removed the Wii/GameCube backwards compatibility GPU, and was targeting about half the performance of the Wii U's "modern" GPU component.


I would just add that there is really now way Nintendo doesn't blow the Steam Deck out of the water... when plugged into the TV. I wouldn't be surprised if it slightly underperforms the Steam Deck as a purely handheld device, but makes up for it with it's modern architecture and bespoke ports.


It's a little unclear, but there is evidence that Nintendo got a deal from Nvidia, but it wasn't on the chips. It was on software. Nvidia's Tegra team needed a win, and TX1's big product was supposed to be the Google Pixel C. The Pixel C came out, but it was a boondoggle for Google, who abandoned the original plans. The Tegra team made a huge pitch to Nintendo.

Nintendo didn't want to leave their custom GPU, because they had a huge software investment in it. In fact, that's what Indy was all about - rebuilding the handhelds around the same GPU as the TV consoles, so they could reuse the highly optimized software stack they'd been building since the N64 (the N64, the GameCube, the Wii and the Wii U all had their GPUs built by the same team).

So Nvidia built a prototype version of a replacement for that software stack as part of their pitch to Nintendo. Obviously Nvidia knows their own hardware very well. We don't have the actual details of the deal that ultimately got signed, but it seems like Nintendo paid good money for the chips, which was very profitable for Nvidia - but Nintendo got a massive software development, on a super rushed time scale, for free.
Click to expand...

Do you think Backwards Compatibility will be Natively achieved? Or do you think it will require some level of emulation?
 
Staff Post Regarding Dubious Insiders And Better Sourcing
  • Index
    • Hey everyone, there is an ongoing conversation regarding where certain widely regarded dubious “insiders” fit on Famiboards. We are always listening to the reports we receive as well as posts in Site Feedback and Suggestions even if we do not respond to everyone. While we aren’t making broad policy changes just yet as conversations are ongoing, we would like to remind everyone at the very least Future Hardware has higher standards for sourcing. HotGirlVideos69 and Zippo do not reach reach that standard and should not be shared here. Thank you.

    -xghost777, Tangerine Cookie, meatbag, Barely Able
     
  • Index
    • Shoulder said:
    Intelligence doesn’t necessarily increase with wallet size…
    Click to expand...
    Intelligence Doesn’t Increase With Wallet Size: The Story of Twitter

    Now streaming on YouTube Red

    oldpuck said:
    It's a little unclear, but there is evidence that Nintendo got a deal from Nvidia, but it wasn't on the chips. It was on software. Nvidia's Tegra team needed a win, and TX1's big product was supposed to be the Google Pixel C. The Pixel C came out, but it was a boondoggle for Google, who abandoned the original plans. The Tegra team made a huge pitch to Nintendo.

    Nintendo didn't want to leave their custom GPU, because they had a huge software investment in it. In fact, that's what Indy was all about - rebuilding the handhelds around the same GPU as the TV consoles, so they could reuse the highly optimized software stack they'd been building since the N64 (the N64, the GameCube, the Wii and the Wii U all had their GPUs built by the same team).

    So Nvidia built a prototype version of a replacement for that software stack as part of their pitch to Nintendo. Obviously Nvidia knows their own hardware very well. We don't have the actual details of the deal that ultimately got signed, but it seems like Nintendo paid good money for the chips, which was very profitable for Nvidia - but Nintendo got a massive software development, on a super rushed time scale, for free.
    Click to expand...
    This story is wild. What an incredible hail mario by the Nvidia teams, wow.
     
    Concernt said:
    I could see Nintendo bring out a WiFi 6E headset device for a reasonable price, but there's concerns with that like if NG Switch even supports WiFi 6E, and would it have to drop its internet connection to connect to the headset.
    Click to expand...
    Forgive me for asking, but what is a WiFi 6E headset, aside it probably connecting to the internet? Is there something special within the 6E in there?
     
    Lugia667 said:
    sorry if I'm repeating others here but all this 8GB is complete speculation at this point and we don't actually have any inside info on it right?
    Click to expand...
    The 8gig talk comes from a Taiwanese economic journal which has a history of not being right and HotGirlVideos69. Nate still thinks it's higher, and Nerco put up an article earlier today predicting 16 gigs.
     
    LiC said:
    * Hidden text: cannot be quoted. *
    Click to expand...

    Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.


    Thraktor said:
    Yup, it was barely mentioned by Microsoft in their PR for the console, and includes support for a custom compression algorithm called BCPACK which is specifically designed to compress texture data, and as such likely achieves much better compression ratios than PS5's Kraken, which is a general-purpose compression algorithm.
    Click to expand...

    Wrong comparison. Both consoles have a general-purpose compression algorithm and a texture compression algorithm, which are used together. PS5's Oodle Kraken GPCA (combined with Oodle-optimised hardware decompression hardware, apparently) is better than Xbox's Zlib (the same algorithm as PS4 and Xbox One) by about 10%. PS5's TCA is Oodle Texture - I don't know how that compares to BCPACK (I do know it came to PS5 a bit after launch while BCPACK was there day), but PS5 games are routinely about 10% smaller, matching the GPCA difference, so I'd bet they're about equal.
     
    Eclissi said:
    RAM - 12x16 GB in LPDDR5/X
    Click to expand...
    Goddamn, 196GB of RAM? This is gonna be a beast

    Lugia667 said:
    sorry if I'm repeating others here but all this 8GB is complete speculation at this point and we don't actually have any inside info on it right?
    Click to expand...
    Completely baseless speculation that's pessimistic for the sake of being pessimistic with a sprinkling of condescension.

    Course remains the same on 12-16GB RAM.

    Apenknul said:
    Forgive me for asking, but what is a WiFi 6E headset, aside it probably connecting to the internet? Is there something special within the 6E in there?
    Click to expand...
    • WiFi 5 offers faster speeds of up to 1 Gbps on the 5GHz band but has congestion issues in dense environments.
    • WiFi 6 boosts speed up to 4 Gbps and efficiency with OFDMA and TWT but requires compatible devices.
    • WiFi 6E unleashes 6GHz band for reduced congestion, lower latency, and speeds over 5 Gbps.
    (source)

    Honestly if the antenna doesn't cost all that much, no reason not to splurge and futureproof on WiFi 6e for the new console.
     
    oldpuck said:
    Rich's article was linked to and it's excellent, but quick summary: it's a totally custom chip for Nintendo, developed by the Tegra team.

    Nvidia essentially had two branches of the Tegra line, the chips for consumer devices (TX1) and chips for robotics (Xavier). The consumer chips never really panned out for any customer but Nintendo, and the robotics chips were only really a hit in the automotive sector. Orin, the "current" Tegra chip is a spiritual successor to Xavier, now very specifically customized to the auto industry.

    T239, also called Drake, is the spiritual successor to the TX1, but it's totally new technology, heavily customized for Nintendo's use. It has custom hardware decompressing game assets quickly, It has a CPU cluster designed for high performance "always on" devices, instead of the "short bursts" of activity like phones and tablets. It has a very large memory bus to support heavy graphics applications, and it has a very large GPU, about twice the size of the Xbox One's GPU, but with all the modern features of Nvidia's RTX 30 cards. That means ray tracing and DLSS.


    It did not, IIRC. It wasn't a Wii U GPU cut in half, more like 1/3. The Wii U GPU was ... strange. It was, essentially, two GPUs slapped together. One was the Wii's GPU, and the second was a more modern GPU by the same team. The GPU module all combined was called "Latte"

    The INDY GPU was called "Decaf" and it removed the Wii/GameCube backwards compatibility GPU, and was targeting about half the performance of the Wii U's "modern" GPU component.


    I would just add that there is really now way Nintendo doesn't blow the Steam Deck out of the water... when plugged into the TV. I wouldn't be surprised if it slightly underperforms the Steam Deck as a purely handheld device, but makes up for it with it's modern architecture and bespoke ports.


    It's a little unclear, but there is evidence that Nintendo got a deal from Nvidia, but it wasn't on the chips. It was on software. Nvidia's Tegra team needed a win, and TX1's big product was supposed to be the Google Pixel C. The Pixel C came out, but it was a boondoggle for Google, who abandoned the original plans. The Tegra team made a huge pitch to Nintendo.

    Nintendo didn't want to leave their custom GPU, because they had a huge software investment in it. In fact, that's what Indy was all about - rebuilding the handhelds around the same GPU as the TV consoles, so they could reuse the highly optimized software stack they'd been building since the N64 (the N64, the GameCube, the Wii and the Wii U all had their GPUs built by the same team).

    So Nvidia built a prototype version of a replacement for that software stack as part of their pitch to Nintendo. Obviously Nvidia knows their own hardware very well. We don't have the actual details of the deal that ultimately got signed, but it seems like Nintendo paid good money for the chips, which was very profitable for Nvidia - but Nintendo got a massive software development, on a super rushed time scale, for free.
    Click to expand...

    Incredible.

    Just to confirm (apologies if this was clear previously and I think I’m correct reading between the lines) that the INDY GPU is what Nintendo was originally thinking for the NX/Switch before Tegra X1 came into the picture?
     
    carbvan said:
    The 8gig talk comes from a Taiwanese economic journal which has a history of not being right and HotGirlVideos69. Nate still thinks it's higher, and Nerco put up an article earlier today predicting 16 gigs.
    Click to expand...
    And merch! Don't forget the merch.
     
    Please try to base your arguments on credible sources and reasoning. “Because Nintendo” is not a sufficient basis. -mariodk18, Dardan Sandiego, Dragoncaine, meatbag, big lantern ghost
    Kartina said:
    I find it funny how the predictions that are more optimistic that come from leaked data are derided and criticized for setting themselves for disappointment while extremely pessimistic predictions (i’ve heard stuff like the next switch being xbox one level but weaker than a base ps4) based on vibes and “Nintendo is always cheap” are accepted as realistic.
    This is what I call the “Because Nintendo difference”
    Click to expand...
    considering how low the ram & storage are based on that leak, it's much more believable than any other leak that says otherwise mainly cuz again Nintendo isn't really into making cutting edge tech
     
    Reader mode
    You must log in or register to reply here.
    Please read this staff post before posting.

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


    Back
    Top Bottom