According to a famous Japanese gaming journalist, the Nintendo Switch will likely the NVIDIA Pascal ‘Parker’ architecture for its GPU, and the Tegra chip inside the platform could help the Switch evolve just like the PS4 and the PS4 Pro.

The news comes from famous Japanese journalist, Nishikawa Zenji, who has written an article about the Nintendo Switch and its possible specs on Japanese website 4Gamer. Zenji also predicted that Sony would introduce an enhanced version of the base PS4 several years ago.

While Nintendo and NVIDIA have confirmed that the Switch will use a custom NVIDIA Tegra chip, the exact specs for the platform weren’t revealed. Nintendo will likely reveal these during its special Nintendo Switch Presentation in January 2017, but for now, we’re left guessing. The article from Zenji is written in Japanese, but we have done our best to translate his post. There’s quite some interesting stuff in the article on 4gamer, and I highly suggest to read it by running it through a translator.

According to Zenji however, it’s likely that the custom Tegra chip inside the Nintendo Switch is based on NVIDIA Pascal ‘Parker’ technology. This due to the fact that Nintendo plans to launch the Switch in March 2017 and the latest Tegra generation in present time is the ‘Parker’ technology. Some might argue that the Switch could use NVIDIA Volta tech as well, but Nintendo’s isn’t known for using the latest technology. Of course, the Tegra ‘Maxwell’ architecture is also a possibility, but this seems
far-fetched due to its 20nm SOC. Zenji notes that there are no indications that Nintendo will opt for the ‘Maxwell’ architecture within the Switch.

Zenji writes that the custom NVIDIA chip will  feature a floating-point performance around 1 TFLOPS, and according to the Japanese journalist, there is almost “no possibility” that the Switch will perform above 1.5 TFLOPS (single-precision floating-point performance) due to the battery drive inside the Switch.

Japanese journalist Nishikawa Zenj has produced a report on 4Gamer about the upcoming Nintendo Switch and its likely specifications. Zenji reports that the Nintendo Switch will likely use the Nvidia Pascal ‘Parker’ architecture for the system GPU. He also claims that the special Tegra chip could allow Nintendo to enhance the system much like Sony has done with the PlayStation 4 Pro. Zenji is the same reporter responsible for predicting that Sony would released an technically enhanced version of the PlayStation 4 which we now know as the PlayStation 4 Pro.

• Switch will likely use the NVIDIA Pascal ‘Parker’ architecture for its GPU
• the Tegra chip inside the platform could help the Switch evolve just like the PS4 and the PS4 Pro
• There are no indications that Nintendo will opt for the ‘Maxwell’ architecture within the Switch
• The custom NVIDIA chip will feature a floating-point performance around 1 TFLOPS
• There is almost “no possibility” that the Switch will perform above 1.5 TFLOPS due to the battery drive inside the Switch

http://wccftech.com/nintendo-switch-nvidia-tegra-pascal/

Thanks to N-Dub Nation for the news tip!

Makes sense for Pascal but 1 TF Seems pretty high... Even according to Nvidia's own blog post back when they released Parker, it is 750Gflops in FP32 or 1.5TF in FP16.

https://blogs.nvidia.com/blog/2016/08/22/parker-for-self-driving-cars/

Granted they could have certainly changed it but that would mean that it is using even more power than it was before which would be pretty essh for the battery life.

But if that is the case, then it could be on par or very close to x1's performance levels due to it being Nvidia which generally means TF is more efficient than AMD's one. We have seen quite a few times where AMD had higher memory bandwidth and higher TF but Nvidia still beat them.

With that being said though, it still needs to be that high and there are also a lot more things that goes into making a GPU. I will remain skeptical.

Hopefully the Switch is using a Tegra X2. If it ends up being 1TFlops it's really close to Xbox One power level at least.

All the bulletpoints are plausible.  The Tegra X2 could explain the delay if Nvidia wanted to refine the Switch chip further given how recent that tech is (and they don't want the first consumer impresion of X2 to be bad).  1TFLOPS is...tricky.  Not impossible mind, the Tegra X2 "Parker" already hits 750 GLFOPS in 32fpp.  Augmentation could push it to 1 to 1.1 TFLOPS, if Nvidia and Nintendo so desire (also heard some talk on a techy forum that the 16fpp performance of the Tegra X2 could be used to help get a performane boost for games that go out of their way to use 16fpp for whatever is possible (incidentally, Unreal Engine 4 makes this very easy)).  Not surprised they don't want Maxwell.  And yeah, 1.5 TFLOPS is impossible though also not overly necessary as Nvidia has always been markedly better with a performance to power ratio than AMD, the company behind other console tech. 

Encouraging if true, but now the rumors seem to pull against each other.  We'll have to wait and see who is right.

If thats true, and that lets say it has around 1 TF for handheld mode and close to the 1.5 TF in console mode(with it docked, it dosent have to worry about battery and thuis actives more GPU components/no need to underclock), then this will be a very good home console, and a beast of a handheld.It should put to rest, or at least ease, the fear of not being powerful enough for third parties.

Good news if true.

The issue is the 1.5 would be hard to get out of the Parker in general.  That's double the GLFOPS and will also result in more heat which will require more cooling for which the Switch's fairly compact active cooling may not be able to properly supply.  It's much more likely they will just shoot for 1 to 1.1 TFLOPS and wring out every last drop of performance.  That's why you would go with Nvidia, efficiency.  

Basically, I could see augmenting the Parker to get a 35% performance boost to get real world performance parity or something close with Xbone.  I don't see how they could get a 100% performance boost without driving the price way too  high and within the design constraints of the Switch in general.

If true that would be plenty powerful. It should be enough that there wouldn't be any major hurdles to porting.

Given that it supposedly utilizes a "custom chip" the released specs of the Tegra X2 don't necessarily apply to the Switch. Although, as I understand it, just comparing TFLOPs without additional context doesn't mean much.

Still, this does sound like a very powerful chip, which would probably explain the poor battery life.

well if true, then developers will have to configure their games for docked and undocked. sounds like trouble to me.