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

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

Have we reached the point of comparing Switch 2 with PlayStation 5?
These 2 platforms do not directly compete with each other.
 
Have we reached the point of comparing Switch 2 with PlayStation 5?
These 2 platforms do not directly compete with each other.
The PlayStation 5 is the primary platform for developing video games for most third party developers. So technical comparisons between Nintendo's new hardware and the PlayStation 5 are inevitable.

This is a random thought, but I do wonder if Nintendo's new hardware's devkit spec sheet's going to be leaked by Moore's Law is Dead?
 
The PlayStation 5 is the primary platform for developing video games for most third party developers. So technical comparisons between Nintendo's new hardware and the PlayStation 5 are inevitable.

This is a random thought, but I do wonder if Nintendo's new hardware's devkit spec sheet's going to be leaked by Moore's Law is Dead?
MLID really proved themselves to be an even weirder source than first thought thanks to this leak. He can have stellar information in specific cases, but sometimes it just isn't nearly as clear cut.

We'll have to wait and see for if he gets the holy grail of good sources for the Switch 2, but considering everyone and their mother has been looking for anything about the device for years to no avail, I don't feel confident about that. Maybe if there's a GDC leak somehow, idk.
 
MLID really proved themselves to be an even weirder source than first thought thanks to this leak. He can have stellar information in specific cases, but sometimes it just isn't nearly as clear cut.

We'll have to wait and see for if he gets the holy grail of good sources for the Switch 2, but considering everyone and their mother has been looking for anything about the device for years to no avail, I don't feel confident about that. Maybe if there's a GDC leak somehow, idk.
It's weird cause his Switch 2 leak talk makes utterly no sense from a technical perspective. Its not entirely clear to me what he heard from his "source" and what is his own speculation, but the part about a 5nm soc would be smaller and clocked higher than an 8nm soc literally goes against every principle of semiconductor design. If his source said that, we can immediately discard it.
 
Hardware Unboxed dropped a new video explaining FSR upscaling issues compared to DLSS, as well as 1080p upscaling vs 1080p Native

May seem off topic, but it the Switch 2 does have DLSS, then Nintendo has an advantage


It's the most on topic thing posted here in a while. Thank you!
 
Hardware Unboxed dropped a new video explaining FSR upscaling issues compared to DLSS, as well as 1080p upscaling vs 1080p Native

May seem off topic, but it the Switch 2 does have DLSS, then Nintendo has an advantage


It's hard to think of a more perfect coincidence in the industry than Nintendo pursuing a hybrid home/handheld strategy with Nvidia, just before Nvidia would deliver the perfect piece of technology for upscaling output between home/handheld displays.
 
It's hard to think of a more perfect coincidence in the industry than Nintendo pursuing a hybrid home/handheld strategy with Nvidia, just before Nvidia would deliver the perfect piece of technology for upscaling output between home/handheld displays.
And it took Nintendo 5 years to start supporting it.
 
For reference, AMD has at least three different hardware AI acceleration technologies.

The first is the matrix core, which was first introduced in 2020 on the Instinct MI100 HPC GPU. These are pretty much direct analogs to Nvidia's tensor cores, and can operate on a wide range of data types, from INT4 up to FP64. The latest versions also support structured sparsity acceleration, like tensor cores. So far AMD has only included these in their HPC GPUs, not consumer graphics cards.

The second is WMMA, which is a special instruction type supported on AMD's consumer GPUs from RDNA3 onwards. This can be thought of as a half-way point between running AI code directly on regular shader cores and tensor cores, as these instructions run on shader cores, but at a higher performance than standard shader code does. On RDNA3, this performance increase is limited to FP16, where WMMA performs at 2x the rate of general FP16 shader code. On RDNA4, it looks like INT8 will also benefit from WMMA.

The last is XDNA, which is a stand-alone hardware block for accelerating AI, ie a NPU. It's based on technology from Xilinx (who AMD acquired in 2022) and was first implemented in the Ryzen 7000 series laptop APUs released in early 2023. The latest version claims 16 TOPS of performance at INT8, and although I've seen claims it can also operate on bfloat16 (a variant of FP16), performance with bfloat16 is likely quite a bit lower.

The PS5 Pro is using WMMA instructions as its AI acceleration, which makes complete sense as it's what AMD use in their consumer GPUs. It likely offers a good balance of performance without taking up too much die space.

Despite doing an entire damn paper on AMD's business, I have to admit that I had to look up what WMMA actually does. Grabbed this from AMD GPUOpen:

Apparently the RDNA 4 version is just more efficient and supports 8-bit precision as opposed to the 16-bit implementation of RDNA 3's WMMA.

So TLDR, AI tasks on PS5 go nyoom (if that wasn't already evident from the PSSR's apparent 2ms to upscale time in the document for 1080p to 4K)

If PSSR is anything like DLSS, then it will likely be operating at FP16 precision for any game using a HDR colour buffer (which I would imagine is basically every PS5 game). So the big INT8 TOPS performance increase isn't really relevant there, and it seems like the FP16 WMMA performance is identical to RDNA3 (2x FP32 dual-issue performance).

Honestly, I'm a lot more interested in seeing what Switch 2 devs do with tensor cores than PS5 Pro devs do with WMMA instructions, even though the latter offers much higher performance. PSSR (if it's good) is a good use of the hardware, but the fact that the base PS5 doesn't have any dedicated ML acceleration hardware means any really interesting use of it will be limited by having to have a non-ML fallback on the base PS5. Switch 2 devs will be able to develop games knowing that tensor cores will always be there, so I wouldn't be surprised if we see much more inventive use of ML on the console compared to PS5 Pro.
 
It's weird cause his Switch 2 leak talk makes utterly no sense from a technical perspective. Its not entirely clear to me what he heard from his "source" and what is his own speculation, but the part about a 5nm soc would be smaller and clocked higher than an 8nm soc literally goes against every principle of semiconductor design. If his source said that, we can immediately discard it.
I think that's Moore's Law is Dead's speculation, especially since I don't believe he mentioned his sources when making that claim.

Anyway, I want to be clear that I don't disagree with the point about Nintendo reducing the number of CUDA cores and increasing the GPU frequency if fabricated Drake using TSMC's 4N process node not making much sense.

But saying that, I don't think going smaller and faster is necessarily contrary to every principle regarding designing semiconductors.

In fact, going smaller and faster is Sony's philosophy when designing the GPU on the PlayStation 5's APU. The GPU on the PlayStation 5's APU has 36 CUs running at up to 2.23 GHz (dynamic frequency), whereas the GPU on the Xbox Series X's APU has 52 CUs running at 1.825 GHz. And the PlayStation 5's APU's die sizes (~300 mm² for Oberon and <260 mm² for Oberon Plus) are smaller than the Xbox Series X's die size of 360.45 mm².

And smaller chips are generally cheaper than bigger chips, especially considering that Sony mentioned a little more than a year after the PlayStation 5 launched that the PlayStation 5 (with a disc drive)'s no longer selling at a loss.

Of course, there are caveats to going smaller and faster. Considering the relationship between frequency and power consumption is generally exponential (e.g. Ryzen 9 6900HS's frequency vs CPU voltage graph with respect to the APU power courtesy of Dr Ian Cutress, who at the time, worked for Anandtech), the probability of lower chip yields is much higher when the frequency's higher.

In fact, Bloomberg reported back in 15 September 2020 that the PlayStation 5's APU's yields were as low as 50%. (Sony only denied the portion of Bloomberg's report about Sony reducing the production number of the PlayStation 5 by 4 million units.)

And since the power consumption increases as the frequency increases, the temperature increases as the power consumption increases as well (e.g. Ryzen 9 6900HS's APU power vs CPU voltage/temperature graph), which is probably why Sony opted to use liquid metal to cool the PlayStation 5's APU.

So I do believe Nintendo has no intention of reducing the number of CUDA cores and increasing the GPU frequency if fabricating Drake using TSMC's 4N process node from a chip yields and a thermal standpoint.
 
MLID really proved themselves to be an even weirder source than first thought thanks to this leak. He can have stellar information in specific cases, but sometimes it just isn't nearly as clear cut.

We'll have to wait and see for if he gets the holy grail of good sources for the Switch 2, but considering everyone and their mother has been looking for anything about the device for years to no avail, I don't feel confident about that. Maybe if there's a GDC leak somehow, idk.

I feel like MLID is generally correct about AMD based products. I believe his track record on Nvidia leaks is very spotty...he has been wrong on most Nvidia leaks in the past. Kopite makes fun of MLID on Twitter all the time in regards of his Nvidia leaks. Which for that matter Kopite is generally right about Nvidia when it comes to desktop cards and not on the tegra line. So MLID is a weird source when it comes to Nintendo and Nvidia in general as he is 85% about Nvidia based products.

As far as the switch 2. Nintendo silence and secrecy screams pretty loud that they have something cooking with the switch 2. Just wished there was more leaks to get us riled up. I believe the PS5 pro discussion is getting a bit out of hand.
 
I think that's Moore's Law is Dead's speculation, especially since I don't believe he mentioned his sources when making that claim.

Anyway, I want to be clear that I don't disagree with the point about Nintendo reducing the number of CUDA cores and increasing the GPU frequency if fabricated Drake using TSMC's 4N process node not making much sense.

But saying that, I don't think going smaller and faster is necessarily contrary to every principle regarding designing semiconductors.

In fact, going smaller and faster is Sony's philosophy when designing the GPU on the PlayStation 5's APU. The GPU on the PlayStation 5's APU has 36 CUs running at up to 2.23 GHz (dynamic frequency), whereas the GPU on the Xbox Series X's APU has 52 CUs running at 1.825 GHz.

And smaller chips are generally cheaper than bigger chips, especially considering that Sony mentioned a little more than a year after the PlayStation 5 launched that the PlayStation 5 (with a disc drive)'s no longer selling at a loss.

Of course, there are caveats to going smaller and faster. Considering the relationship between frequency and power consumption is generally exponential (e.g. Ryzen 9 6900HS's frequency vs CPU voltage graph with respect to the APU power courtesy of Dr Ian Cutress, who at the time, worked for Anandtech), the probability of lower chip yields is much higher when the frequency's higher.

In fact, Bloomberg reported back in 15 September 2020 that the PlayStation 5's APU's yields were as low as 50%. (Sony only denied the portion of Bloomberg's report about Sony reducing the production number of the PlayStation 5 by 4 million units.)

And since the power consumption increases as the frequency increases, the temperature increases as the power consumption increases as well (e.g. Ryzen 9 6900HS's APU power vs CPU voltage/temperature graph), which is probably why Sony opted to use liquid metal to cool the PlayStation 5's APU.

So I do believe Nintendo has no intention of reducing the number of CUDA cores and increasing the GPU frequency if fabricating Drake using TSMC's 4N process node from a chip yields and a thermal standpoint.
Yea that was badly formulated of me, and you addressed the reason in your post. If you're going for power efficiency, which is the case very much with Nintendo it is the opposite of sound design for the bolded reason.
 
0
MLID really proved themselves to be an even weirder source than first thought thanks to this leak. He can have stellar information in specific cases, but sometimes it just isn't nearly as clear cut.

We'll have to wait and see for if he gets the holy grail of good sources for the Switch 2, but considering everyone and their mother has been looking for anything about the device for years to no avail, I don't feel confident about that. Maybe if there's a GDC leak somehow, idk.
Why are we still bringing up MLID? His track record is very spotty at best.

He also made several claims of Nintendo switched vendors from Nvidia to AMD...
 
RDNA 2's ray tracing performance is inferior to Ampere's ray tracing performance. (The PlayStation 5 uses RDNA 2's ray tracing technology whereas Nintendo's next-gen console uses Ampere's ray tracing technology.)


I'm talking about the console equipped with Drake since that hasn't been officially launched by Nintendo yet.

I was only talking about how I expect Nintendo's next-gen console's ray tracing performance to not be too far off from the PlayStation 5 Pro's rumoured ray tracing performance, theoretically speaking.

I expect the PlayStation 5 Pro's ray tracing performance to be on par with Ada Lovelace's ray tracing performance in the absolute best case scenario, and on par with Ampere's ray tracing performance in the absolute worst case scenario.

But yeah, I don't really expect that to really help third party developers, since as mentioned, third party developers still have to develop for the Xbox Series S in mind as well if third party developers want to develop for the Xbox Series X.

I remember this post below from Raploz that the result is a better Ray Tracing on the Ps5-like hardware than in the ampere hardware.

We can extrapolate the ray-tracing performance we can get on Switch 2 and PS5/SX/SS by using existing PC parts, and then we can compare how they perform relative to each other. This is by no means fully accurate, but it can give us some idea of what to expect as long as we don't focus on specific FPS numbers, but instead we focus on the relative performance to each other.

The closest GPU to the PS5 is a 6700xt. It has slightly more cores (2304 vs 2560), and in the video I'll link below it's running at 2.6GHz vs 2.2 on the PS5.

The 3050 mobile is (much) faster than the t239 on Switch 2, it has more cores (2048 vs 1536 shader cores), more ray-tracing cores (16 vs 12) and also runs at much higher clocks. In the second video below, the 3050 is consuming 40-45w and running at 1.8GHz, Switch 2 will probably run at 1.1GHz while docked.

The closest GPU to Series S is the 6500xt. It has fewer cores (1024 vs 1280 on Series S) but it usually runs at higher clocks on PC, so we can expect similar performance.

Minecraft RTX on a 6700xt (closest match to PS5):


Minecraft RTX on a 3050 mobile (closest match to Switch 2):


Minecraft RTX on a 6500xt (closest match to Series S):


If we compare the performance of each card relative to each other at ray-tracing, we get 6700xt (PS5) > 3050m (Switch 2) > 6500xt (Series S).

Let's keep in mind that the 6500xt is more similar to the Series S performance-wise than the other two cards are to their respective consoles, so the actual ray-tracing performance of the t239 should be closer to Series S than the difference between the 6500xt and the 3050 mobile in the videos. The gap between PS5 and the 6700xt is not that big, but the gap between the 3050 mobile and the t239 is more substantial, so take that into consideration while comparing.

If we take all that into consideration, and extrapolate those PC numbers to consoles, Switch 2 will probably perform as well as - or slightly better than - a Series S at pure ray-tracing, but it won't match the PS5, even with the advantage of using dedicated ray-tracing cores.


Ok, that test is not made on DLSS 3.5. Can someone that has a laptop with a RTX on a 3050 mobile redo that test to see if it runs better now?

But that test shows that Switch 2 probably will not have a ray tracing that good as you hope. Or am I missing something?
 
I remember this post below from Raploz that the result is a better Ray Tracing on the Ps5-like hardware than in the ampere hardware.
We can extrapolate the ray-tracing performance we can get on Switch 2 and PS5/SX/SS by using existing PC parts, and then we can compare how they perform relative to each other. This is by no means fully accurate, but it can give us some idea of what to expect as long as we don't focus on specific FPS numbers, but instead we focus on the relative performance to each other.

The closest GPU to the PS5 is a 6700xt. It has slightly more cores (2304 vs 2560), and in the video I'll link below it's running at 2.6GHz vs 2.2 on the PS5.

The 3050 mobile is (much) faster than the t239 on Switch 2, it has more cores (2048 vs 1536 shader cores), more ray-tracing cores (16 vs 12) and also runs at much higher clocks. In the second video below, the 3050 is consuming 40-45w and running at 1.8GHz, Switch 2 will probably run at 1.1GHz while docked.

The closest GPU to Series S is the 6500xt. It has fewer cores (1024 vs 1280 on Series S) but it usually runs at higher clocks on PC, so we can expect similar performance.

Minecraft RTX on a 6700xt (closest match to PS5):


Minecraft RTX on a 3050 mobile (closest match to Switch 2):


Minecraft RTX on a 6500xt (closest match to Series S):


If we compare the performance of each card relative to each other at ray-tracing, we get 6700xt (PS5) > 3050m (Switch 2) > 6500xt (Series S).

Let's keep in mind that the 6500xt is more similar to the Series S performance-wise than the other two cards are to their respective consoles, so the actual ray-tracing performance of the t239 should be closer to Series S than the difference between the 6500xt and the 3050 mobile in the videos. The gap between PS5 and the 6700xt is not that big, but the gap between the 3050 mobile and the t239 is more substantial, so take that into consideration while comparing.

If we take all that into consideration, and extrapolate those PC numbers to consoles, Switch 2 will probably perform as well as - or slightly better than - a Series S at pure ray-tracing, but it won't match the PS5, even with the advantage of using dedicated ray-tracing cores.

Ok, that test is not made on DLSS 3.5. Can someone that has a laptop with a RTX on a 3050 mobile redo that test to see if it runs better now?

But that test shows that Switch 2 probably will not have a ray tracing that good as you hope. Or am I missing something?

In terms of raw performance, Nintendo's next-gen console probably won't come near the PlayStation 5 and the Xbox Series X|S in terms of ray tracing. But I wasn't talking solely about raw performance. Nintendo's next-gen console's rumoured to support ray reconstruction. And I do think ray reconstruction should allow Nintendo's next-gen console to be comparable to the PlayStation 5 and the Xbox Series X|S in terms of ray tracing.
 
The PS5 may be the main platform for third-party developers, but as of today, the minimum requirement for this generation still remains the Xbox Series S.

Personally, I'm optimistic about Switch 2. It will probably be the new minimum requirement for this generation, but the gaming industry needs a new platform on which to distribute games. These are difficult times for the gaming industry. Third party support will certainly be stronger than on Switch. So developers and publishers will do everything to publish as many games as possible. Furthermore, I think this generation is destined to last longer.

As for performance, I'm even more relaxed. If the console can guarantee 1080p or even 2K reconstructed via DLSS at 30 fps or 60 fps, depending on the games, I think it will be fine. Whether internally the games are rendered at 540p, 720p, 900p is irrelevant. The important thing is that the view is good on TV. Nobody expects graphic detail from PS5. It is clear that on PS5 if a certain game will have high detail and density, the same game on Switch 2 will probably have medium detail. The important thing is that the glance remains of a good level and visually satisfying on TV. 🙂

For me, I mostly play on TVs, specifically a 4K OLED TV. My current consoles are still Switch and PS4. If I consider PS4, I can say that the majority of games run at 1080p with HDR. In the case of more graphically complex games the frame rate is always set at 30 fps. The truth is that it plays very well and visually the performance of the games ranges from excellent to superb. Modern TVs have also improved in the management of upscaling, so an average high resolution such as 1080p spread across 4K TVs still performs very well.

In short, the important thing is that the Tegra T239 has the right modern technology to deal with the games of this generation. If some game is too complicated to port (e.g. GTA 6), personally I wouldn't find it dramatic. I think we'll still get the vast majority of AAA games. I expect full support from Ubisoft, Capcom, EA, Square Enix, CDR Project, etc. and with more graphically pleasing and satisfying results. 🙂
 
the thing we have to remember about these ray tracing benchmarks is that the floor is still lowering. with UE5's Lumen coming to Android and IOS, and other RT effects already running on mobile devices, we shouldn't just look at what's out now and say "yea, Drake can't do that"
 
The PS5 may be the main platform for third-party developers, but as of today, the minimum requirement for this generation still remains the Xbox Series S.

Personally, I'm optimistic about Switch 2. It will probably be the new minimum requirement for this generation, but the gaming industry needs a new platform on which to distribute games. These are difficult times for the gaming industry. Third party support will certainly be stronger than on Switch. So developers and publishers will do everything to publish as many games as possible. Furthermore, I think this generation is destined to last longer.

As for performance, I'm even more relaxed. If the console can guarantee 1080p or even 2K reconstructed via DLSS at 30 fps or 60 fps, depending on the games, I think it will be fine. Whether internally the games are rendered at 540p, 720p, 900p is irrelevant. The important thing is that the view is good on TV. Nobody expects graphic detail from PS5. It is clear that on PS5 if a certain game will have high detail and density, the same game on Switch 2 will probably have medium detail. The important thing is that the glance remains of a good level and visually satisfying on TV. 🙂

For me, I mostly play on TVs, specifically a 4K OLED TV. My current consoles are still Switch and PS4. If I consider PS4, I can say that the majority of games run at 1080p with HDR. In the case of more graphically complex games the frame rate is always set at 30 fps. The truth is that it plays very well and visually the performance of the games ranges from excellent to superb. Modern TVs have also improved in the management of upscaling, so an average high resolution such as 1080p spread across 4K TVs still performs very well.

In short, the important thing is that the Tegra T239 has the right modern technology to deal with the games of this generation. If some game is too complicated to port (e.g. GTA 6), personally I wouldn't find it dramatic. I think we'll still get the vast majority of AAA games. I expect full support from Ubisoft, Capcom, EA, Square Enix, CDR Project, etc. and with more graphically pleasing and satisfying results. 🙂
To this day, my TV is 720p, which ended up working great for the Switch, hoping to find out the real target resolution of the NG so I can upgrade my TV lol.
 
The PS5 may be the main platform for third-party developers, but as of today, the minimum requirement for this generation still remains the Xbox Series S.

Personally, I'm optimistic about Switch 2. It will probably be the new minimum requirement for this generation, but the gaming industry needs a new platform on which to distribute games. These are difficult times for the gaming industry. Third party support will certainly be stronger than on Switch. So developers and publishers will do everything to publish as many games as possible. Furthermore, I think this generation is destined to last longer.

As for performance, I'm even more relaxed. If the console can guarantee 1080p or even 2K reconstructed via DLSS at 30 fps or 60 fps, depending on the games, I think it will be fine. Whether internally the games are rendered at 540p, 720p, 900p is irrelevant. The important thing is that the view is good on TV. Nobody expects graphic detail from PS5. It is clear that on PS5 if a certain game will have high detail and density, the same game on Switch 2 will probably have medium detail. The important thing is that the glance remains of a good level and visually satisfying on TV. 🙂

For me, I mostly play on TVs, specifically a 4K OLED TV. My current consoles are still Switch and PS4. If I consider PS4, I can say that the majority of games run at 1080p with HDR. In the case of more graphically complex games the frame rate is always set at 30 fps. The truth is that it plays very well and visually the performance of the games ranges from excellent to superb. Modern TVs have also improved in the management of upscaling, so an average high resolution such as 1080p spread across 4K TVs still performs very well.

In short, the important thing is that the Tegra T239 has the right modern technology to deal with the games of this generation. If some game is too complicated to port (e.g. GTA 6), personally I wouldn't find it dramatic. I think we'll still get the vast majority of AAA games. I expect full support from Ubisoft, Capcom, EA, Square Enix, CDR Project, etc. and with more graphically pleasing and satisfying results. 🙂
You can also lump in SD and other portable PCs with Switch 2, cause they have roughly similar specs. If you target one, you might as well target them all.

I know in terms of install base, all of them added up together the numbers are fairly insignificant compared to a Nintendo console, but in terms of game sales I assume owners of these on average buy a lot more games than the average Switch 2 owner.
 
0
It's silly to look at the Switch 2 and compare it to the PS5's raw grunt. Switch 2's strengths is that it will utilize lower rendering resolutions, nothing like the Switch mind you, but lower than the PS5's. It can also utilize the DLSS for a much clearer upscale image.

Who cares if the fps is 30fps in graphical showpieces anyways? Most people streaming rebirth have gone with the 30fps graphics fidelity setting in their Action RPG. Switch 2 will have half the performance of the PS5, with a clearer final image than you'd expect from that performance, thanks to DLSS. That's a great result compared to the Switch. Hell we are talking about 540p on the Switch, Switch 2 might have to deal with 1080p resolutions in heavy games that need 60fps.

RT performance will be great, but not at the same resolution as the PS5, though DLSS upscaling should give it an advantage beyond this ~50% performance I'd place it at.

As for the 50% performance, this is because tensor cores do fp16, this is a huge advantage over the Switch's use of fp16 through its Cuda Cores, or RDNA's mixed precision either. See these older GPUs, every clock cycle, they have to decide to do fp32 or fp16, and switch back and forth as needed. With the Tensor cores doing FP16 and at the same speed per second as the Cuda cores do FP32, the GPU will never be a bottleneck for Switch 2. Cache, RAM, Bandwidth, and CPU will all max out before any optimized game (for mixed precision) on Switch 2.

This issue that these other GPUs have, make using mixed precision more complicated, and limit it's usefulness. Switch also used mixed precision in a lot go games because ports have to find performance.
if Switch sucessor can play Xbox series X/S e PS5 without major downgrades, as we saw with some games on Switch(like the Witcher 3, that would be excelent)
 
Why are we still bringing up MLID? His track record is very spotty at best.

He also made several claims of Nintendo switched vendors from Nvidia to AMD...
... because he was the subject of a pretty big leak that basically dropped the details of the upcoming PlayStation 5 Pro?
Idk, I believe it's fair to admit that, when there's an instance of a leaker being obviously correct, someone can get something correct and say "Good Job" at such efforts.

Also, note that none of this was to lend him credit for the buck-wild Switch 2 leaks he provided. A broken clock getting something right doesn't automatically vindicate when they were wrong on many occasions.

To avoid this becoming a leaker-insult event horizon, I'm moving on from there.
In terms of raw performance, Nintendo's next-gen console probably won't come near the PlayStation 5 and the Xbox Series X|S in terms of ray tracing. But I wasn't talking solely about raw performance. Nintendo's next-gen console's rumoured to support ray reconstruction. And I do think ray reconstruction should allow Nintendo's next-gen console to be comparable to the PlayStation 5 and the Xbox Series X|S in terms of ray tracing.
I was pondering about this for a fair bit, and I think this will be a key aspect of the Switch 2. The only feature that's a bit questionable on the Switch 2 is Frame Generation. Basically all other Nvidia technologies are good to go, and Ray Reconstruction is a massive boon for that reason.
 
The PlayStation 5 is the primary platform for developing video games for most third party developers. So technical comparisons between Nintendo's new hardware and the PlayStation 5 are inevitable.

This is a random thought, but I do wonder if Nintendo's new hardware's devkit spec sheet's going to be leaked by Moore's Law is Dead?
I don't know if the overlap is that close, the kind of third party games Playstation gets is pretty different from the third party games Nintendo gets. Playstation is all about AAA third party games while Switch has its biggest support from indie developers.
 
0
Why are we still bringing up MLID? His track record is very spotty at best.

He also made several claims of Nintendo switched vendors from Nvidia to AMD...

I think because many sites are running with his leaked specs as gospel, then makes it relevant to discuss the content of whether any of it makes sense...
 
I think because many sites are running with his leaked specs as gospel, then makes it relevant to discuss the content of whether any of it makes sense...
But we already agree that it doesn't make sense. And that's also with including "because Nintendo" arguments, too.
... because he was the subject of a pretty big leak that basically dropped the details of the upcoming PlayStation 5 Pro?
Idk, I believe it's fair to admit that, when there's an instance of a leaker being obviously correct, someone can get something correct and say "Good Job" at such efforts.

Also, note that none of this was to lend him credit for the buck-wild Switch 2 leaks he provided. A broken clock getting something right doesn't automatically vindicate when they were wrong on many occasions.

To avoid this becoming a leaker-insult event horizon, I'm moving on from there.
It can still be construed as spreading more misinformation around. But yeah, maybe we should steer away from his claims for the time being.
 
I think it’s important to note that MLID got the leak from a leaked document, meanwhile the switch 2 report he made were sources who said they barley worked on the project, so if some reason the sources are right then there’s likely more happened when the sources left the tegra 239 project.
 
Building incredibly small BVHs seems pretty important for the Switch 2 considering how slow mobile RAM is. Hopefully Nintendo can build an algorithm optimizes BVHs to allow ray tracing to just cover up for the worst aspects of screen space reflections and screen space GI (if the Switch 2 is powerful enough for those...)
 
Y’all think the LCD screen might be something similar to the PlayStation portal because it’s a premium LCD? Or just use modern LCD.
Bloomberg mentioned on May 2023 that the Sharp CEO Robert Wu mentioned that Sharp was working with a company on a video game console during the R&D phase in an analyst call after releasing Sharp's quarterly earnings. And Sharp plans to launch pilot LCD panel production lines that fiscal year for the new video game console.
 
Bloomberg mentioned on May 2023 that the Sharp CEO Robert Wu mentioned that Sharp was working with a company on a video game console during the R&D phase in an analyst call after releasing Sharp's quarterly earnings. And Sharp plans to launch pilot LCD panel production lines that fiscal year for the new video game console.
So it’ll be using modern LCD screen for the switch 2 which are brand new. Also what’s a pilot LCD is it the name of the brand?

Also when can we expect the switch 2 to start mass production, because I think buying a lot of parts and letting it be in some warehouse to rot (not literally) is probably not the best idea?
 
Y’all think the LCD screen might be something similar to the PlayStation portal because it’s a premium LCD? Or just use modern LCD.
If this rumor is for the Switch 2, I could see Nintendo adding some custom features to it.

"ccording to Bloomberg (thanks, VCG), CEO Robert Wu confirmed that Sharp is working with an unnamed company on the development of a new gaming console since the R&D stage and that it plans to launch pilot LCD-panel production lines for the device this fiscal year."

Considering the screen supplier wouldn't have been needed to be involved since the R&D stage if it was a bog standard panel.

Someone suggested they could add a reflective panel for outdoor play, which imo would have been very cool, see second link.


 
Last edited:
Hardware Unboxed dropped a new video explaining FSR upscaling issues compared to DLSS, as well as 1080p upscaling vs 1080p Native

May seem off topic, but it the Switch 2 does have DLSS, then Nintendo has an advantage


tl;dw: For docked mode / home console purposes at resolutions of 1440p+, DLSS has a minor advantage vs FSR. For portable devices with resolutions 1080p-, DLSS has a big advantage.

For comparison, here's his similar video from about a year back comparing 1440p and 4K rather than 1080p.
 
Also what’s a pilot LCD is it the name of the brand?
The "pilot" bit could refer to "pilot plan". A pilot plan is when you test a sample of the universe, before applying it on the general population. In the case of Sharp: they're testing part of their production lines before committing to utilizing the 100%. Pilot plans are usually done to see how things flow, and if there's anything wrong during the process, or can be made better, corrections are made early in the production phase, versus going full speed and have to deal with headaches throughout the entire process.
 
The PlayStation 5 is the primary platform for developing video games for most third party developers. So technical comparisons between Nintendo's new hardware and the PlayStation 5 are inevitable.

This is a random thought, but I do wonder if Nintendo's new hardware's devkit spec sheet's going to be leaked by Moore's Law is Dead?
In the past it certainly was, but nowadays it isn't.
Third-party developers can't just treat Switch 2 like a PlayStation 5 but weaker and second-rate,
especially in these times when Nintendo treats developers well.
 
If this rumor is for the Switch 2, I could see Nintendo adding some custom features to it.

"ccording to Bloomberg (thanks, VCG), CEO Robert Wu confirmed that Sharp is working with an unnamed company on the development of a new gaming console since the R&D stage and that it plans to launch pilot LCD-panel production lines for the device this fiscal year."

Considering the screen supplier wouldn't have been needed to be involved since the R&D stage if it was a bog standard panel.

Someone suggested they could add a reflective panel for outdoor play, which imo would have been very cool, see second link.


the Sharp rumours go quite a way back so potentially this new screen has been in the pipeline for years.
It wouldn't shock me. Nintendo is taking a moderate risk jumping from OLED back to LCD, so the screen likely has to be something special to justify it. Having it be a reasonably great, thin LCD screen that doesn't add too much weight while surpassing the visual quality of the Switch 1's models would be a best case scenario for them.
 
Also what’s a pilot LCD is it the name of the brand?
No. A pilot LCD panels production line is a short production run designed to demonstrate that Sharp's bespoke LCD panels can be produced repeatedly and consistently as specified.
 
I do wonder what they will even sell with it, considering they flat out said they have no big games this year
Grab a third-party game
Market it with the PS5 Pro
Add another smaller title as a showcase for some reason
???
Profit(?)


I think Sony was banking on things going a fair bit smoother than they did for this year...
 
It will probably be the same screen as the Portal, but it would be nice if it had HDR as an option.
I don't think that's very likely, considering the PlayStation Portal very likely uses BOE's LCD panels, courtesy of Hartmann.
In my opinion, I am convinced that the PS Portal display is made by BOE.
The part number for the PS Portal display is "80FHM6613".
ZfKinYy.jpg


Generally, the first two digits of the display part number are the manufacturer's code.
However, in the PS Portal, that digit is hidden.
w8wSZWe.jpg


So what do the other digits in the part number mean?
Naturally, the following interpretations are possible
80: 8 inch
FHM: Full-HD Module
6613: Unique code

If you look up "FHM" in this section, only the BOE uses it.
In other words, we can determine that the PS Portal display is made by BOE.
 
I do wonder what they will even sell with it, considering they flat out said they have no big games this year
Since there will be first-party games this year (just maybe not a Spider-Man, TLOU, or Horizon), I'd guess one of those, a third-party game they get the marketing rights for that hasn't been shown yet, or perhaps they'll get a team to work on a big PS5 Pro patch on an older game and market it as new again. If the system has a good lineup of existing games that will benefit, they can just run with that, too. It is a shame GTA VI isn't until next year, since that'd be the obvious choice to me.
 
It would make sense if Switch 2 was 4nm but even if it is 8, there are still some pretty basic assumptions we can make that describe the performance.

I can't see the CPU being lower than 1.4 Ghz. I can't see the GPU clocked any lower than the OG Switch and likely a bit higher. The memory bandwidth may come in a bit lower. But basically, it should still more or less be capable enough for Nintendo's needs. Unless 8nm ends up being so bad that the other power saving enhancements we expect this system to have do nothing of note, it should be fine.
 
If this rumor is for the Switch 2, I could see Nintendo adding some custom features to it.

"ccording to Bloomberg (thanks, VCG), CEO Robert Wu confirmed that Sharp is working with an unnamed company on the development of a new gaming console since the R&D stage and that it plans to launch pilot LCD-panel production lines for the device this fiscal year."

Considering the screen supplier wouldn't have been needed to be involved since the R&D stage if it was a bog standard panel.

Someone suggested they could add a reflective panel for outdoor play, which imo would have been very cool, see second link.



Yeah I posted this awhile back and it's far more intriguing if Nintendo does go this route to make their device more unique to other handheld gaming systems by improving functionality and usability...
 
I do wonder what they will even sell with it, considering they flat out said they have no big games this year
We know what they're going to sell it with, and it's not GTA VI for the time being. They still got Concord unnannounced for this year, Helldivers 2 definitely didn't hold back graphically... That's my bet.
 
Of course, there are caveats to going smaller and faster. Considering the relationship between frequency and power consumption is generally exponential (e.g. Ryzen 9 6900HS's frequency vs CPU voltage graph with respect to the APU power courtesy of Dr Ian Cutress, who at the time, worked for Anandtech), the probability of lower chip yields is much higher when the frequency's higher.

In fact, Bloomberg reported back in 15 September 2020 that the PlayStation 5's APU's yields were as low as 50%. (Sony only denied the portion of Bloomberg's report about Sony reducing the production number of the PlayStation 5 by 4 million units.)

And since the power consumption increases as the frequency increases, the temperature increases as the power consumption increases as well (e.g. Ryzen 9 6900HS's APU power vs CPU voltage/temperature graph), which is probably why Sony opted to use liquid metal to cool the PlayStation 5's APU.

So I do believe Nintendo has no intention of reducing the number of CUDA cores and increasing the GPU frequency if fabricating Drake using TSMC's 4N process node from a chip yields and a thermal standpoint.
This got me thinking just now. What if a big reason why the Switch is clocked for the CPU/GPU the way it is was to ensure that they'd have an ample supply among all the other devices using the Tegra X1 from chip binning? The PS5 APU was very specific to it, and not really used anywhere else, but the spec set for it meant that any chip that did not meet it basically got tossed out. For the Tegra X1, Nintendo wasn't the only one using it, nor were they the first. Nvidia obviously made it for use in their Shield TV as well as Jetsons/Drives, and there was Google with the Pixel C. Those are certainly clocked higher than what Switch is set to, but it would also mean that for yields that weren't up to spec for those could possibly be used for Switch. Nintendo does have their "boost mode" meant for loading by clocking the CPU to 1.7Ghz, but that's done with dropping the GPU clock to almost nothing at 76.8Mhz. Various people on the net have overclocked and tested their Switch systems being stable at higher clocks, but not everyone got equal results. The TX1+ gets even better results from overclocking, but that's because it was designed to go higher than what TX1 could in the first place. IMO, if the TX1 was not used as much for the other devices, granting Nintendo more higher-capable chips, it may have been possible that the Switch could have been assigned to be clocked higher.

With the custom T239 used in the Switch 2, it's believed that no one other than Nintendo is using it, so they get the full range rather than limiting themselves to the low-end for yields. It doesn't mean we're going to get a Switch 2 that will be top-clocked (like 2.6Ghz CPU / 1.38Ghz GPU from the Linkedin profile and Nvidia hack) because that would mean low yields like what PS5 dealt with, but it doesn't necessarily mean it's going to be specced low either. Having high spec and high yields is the obvious choice if it's available, which is why I feel that give more reason for them to go with TSMC 4N rather than Samsung 8nm.
 
It would make sense if Switch 2 was 4nm but even if it is 8, there are still some pretty basic assumptions we can make that describe the performance.

I can't see the CPU being lower than 1.4 Ghz. I can't see the GPU clocked any lower than the OG Switch and likely a bit higher. The memory bandwidth may come in a bit lower. But basically, it should still more or less be capable enough for Nintendo's needs. Unless 8nm ends up being so bad that the other power saving enhancements we expect this system to have do nothing of note, it should be fine.
The problem is basically that GPUs have a performance per watt sweet spot, which Thraktor estimated to be around 500mhz for Ampere. If you go under or over this, performance per vat drops exponentially. You never really see devices go under their sweet spot, because then you rather use a smaller chip. You do see devices go over it, if performance takes priority over efficiency.

Every other 8nm ampere card that exists would consume far too much power at 12SM/ 500mhz to be used in handheld mode for a portable device. Drake got some Lovelace efficiency enhancements, but thats very likely not enough for a chip of that size to make sense. All the estimates before we knew the actual specs, was between 4 and 8 SM.
 


I might've missed posting this in here, but this is interesting consideration which Nintendo and other Handheld gaming makers have to go through to get a good device.
 
Please read this new, consolidated staff post before posting.

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


Back
Top Bottom