Tegra X1 prod line will be stopped I guess. We will get infos on January from NVIDIA.
Last edited:
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
Based on how tech has progressed since the Switch, how powerful could a Switch Pro in early 2021 actually be?
- Thread starter ItWasMeantToBe19
- Start date
- Status
Tegra X1 prod line will be stopped I guess. We will get infos on January from NVIDIA.
What do you mean stopped? TX1+ is what all newer Switchs run on. How would it be "stopped"
He was probably referring to 12 nm leaving mass production for Nvidia.
Yeah honestly even though we're talking about much more substantial upgrades being possible I do agree there's still the potential for this. All of the speculation is just about discussing what's possible and maybe what's likely at times but there's no real reason to discount any possibilities at this stage.
Production line will be stopped at the begining of jan.
Wouldn't this necessitate either moving to 8nm/7nm process by default then?
The issue with supporting DLSS is that the SoC would inherently be more powerful than the base switch. Using deep learning super sampling also requires a minimum amount of performance from the gpu in order to scale frames on time.
Can you elaborate on this? What are your expectations for a Pro?
The point point most people are making is that it's probably just as cost efficient (or even better) to bolt on DLSS/Ampere to a few ARM 7x core than trying to modify the existing X1/GPU to add DLSS.
Hence it's easy to imagine a big jump across the board. Said differently, a X1 w/DLSS is the least likely scenario.
We either get
1) nothing, or
2) die shrink X1, or
3) pretty huge in performance jump with new a SOC bearing DLSS - as a number of people have speculated.
So would they continue manufacturing base Switches at 8nm/7nm using the same TX1 layout? That would signal another huge boost in battery life, no?
And if this is happening in January how to will they have stock to ship in early 2021?
If the standard Switch is updated again, I don't think they'll increase the battery life. It is more than enough.
It would be an inherent energy savings from shrinking the node. I guess they could offset it by reducing the battery size?
Oh what? What are they shipping from Feb till the launch of the new model?
If they're stopping production, then Nintendo will be riding on whatever stock they have built up. I assume chips ain't the bottleneck they have right now, but other components
Then that wouldn't be a silent upgrade like the 2019 one was. I'd be shocked if we randomly start getting higher clocked units in February.
Absolutely. And Nvidia has capacity at both manufacturers for 8, 7 (at Samsung too), (6) and 5 nm. Latest (and plausible) rumors are:
- Hooper will be monolithic (on 7/6 or 5 nm).
- GA103 and the rest of the Ampere series will stay on 8N.
It will depend of the existence of a 2021 lite model. Mariko battery life on the 2019 hybrid model is tied to lite battery capacity. It would be really easier to release a lite model with a 7 nm SOC instead of a 8 nm one.
I would not be shocked if games started coming out that ran better on Mariko devices. All they need to do is unlock faster profiles in firmware.
Yeah fair point.
I doubt they put these out in Feb. Think they do what they can with the stock till the end of the FY.
This means the Lite is getting revised too, I presume?
Considering how Switch has been sold out for the majority of the year in their two primary markets I doubt they really have much stock built up. I'd guess enough to ship for January but not all of February, assuming production of 12nm does end at the start of January.
This stuff about the entire line going to a new node is giving me visions of 2019
Well get a new new Switch at 8nm but at the same clocks. In 2019 the die shrunk gave us the Lite this time it'll give us the Pro at higher clocks with some bells and whistles but it looks like both the base, lite and pro models will use the same chip eventually
Well get a new new Switch at 8nm but at the same clocks. In 2019 the die shrunk gave us the Lite this time it'll give us the Pro at higher clocks with some bells and whistles but it looks like both the base, lite and pro models will use the same chip eventually
we don't know where their bottleneck is. from the NPD thread, October's surge ate into their november allocation which is partially why November was down. so they're allocating stock monthly. they could have plenty of chips but they're not implemented into systems yet for all we know.
Hopper is allegedly the chiplet design for non-gaming stuff while Lovelace (if real) will be the GeForce arch
The Mariko revision solved one of the initial problems of the Switch which was battery life. I don't know the exact efficiency gain but going to 8nm without touching the clocks would probably push the battery life at an unprecedented high level.
Yeah I suppose that's true. Either way this is a very interesting development if true. I don't mean to be disrespectful or anything but it's hard at the moment from our perspective to know whether this info Thugstas has shared is accurate.
Kopite's latest leaks say that the MCM chiplets design has been ditched for Hooper (gaming? compute?) uArch.
Mariko efficiency gains was necessary for the Switch Lite. The upgraded OG model was just an added bonus.
Actually, kopite7kimi said that Hopper is still a MCM GPU, but it has been delayed.
kopite7kimi also said that Nvidia's next-gen gaming GPU is a monolithic GPU, not a MCM GPU, but it's not Hopper.
Interestingly enough, RedGamingTech says he talked to kopite7kimi and said what kopite7kimi said about Nvidia's next-gen gaming GPU lines up with what he has heard about Nvidia's next-gen gaming GPU.
- fabricated at TSMC's 5 nm nodes
- planned for release in 2022
- codenamed "Lovelace" (after Ada Lovelace, a British mathematician who was one of the first computer programmers)
- is a monolithic GPU rather than a MCM GPU
- more of a evolution of Ampere than a brand new GPU architecture
You are saying the X1 product line is closed. So in other words, you are excluding an 8nm TX1 as a possibility? So new chip then?
Mariko also fixed the major security exploit that was in the TX1, which may have been the primary goal of Nintendo in wanting the new chip to revise the whole Switch line-up.
The "pro" replaces OG, and a new "pro lite" replaces the lite. Basically what happened with new 3ds/ 2ds.
Edit: But the options are tx1++ or new chip assuming this new information checks out. The only thing that's excluded here is plain Mariko. Which is great news.
I really doubt there are two separate chips.
Last edited:
Would that happen right away though? I'd be a bit surprised if it does, which is why this whole thing is kinda weird. For my scenarios I'm thinking about what they will be selling throughout the whole year, not just eventually down the line.
Logistically it makes the most sense for all of their models to share the same SoC, so with this new information I'd be willing to bet on scenario #1, where they're all just TX1s. The pro can have more RAM, storage, higher clocks, more premium casing, etc. It would be disappointing but it seems fairly likely if they are indeed shutting down 12nm production within a month.
No problem. So the GPU design planned for 2022 (for EUV nodes) will be monolithic and a sort of tweaked Ampere uArch.
I have the feeling that we will not see a die shrunk TX1 on 8 nm. That said, I'm not saying that the chip we will get on 8 nm will absolutely be a powerful one (a full gen upgrade). However 7 nm is still a possibility and if they decide to go on this node, there is close to no chance to only have a 'pro' upgrade.
Probably, but wouldn't have to be simultaneous. They can get away with making a lot less of those
One (large) piece of silicon, as opposed to being built from several chiplets which is a future trend.
a single chip as we've been having. MCM is more like how Ryzen CPUs work: multiple chips that speak to each other with an IO die
I really think this is leading up to a "New Nintendo Switch" next year
Let's it is a TX1 8mm for all models next year:
How performance gain do we have on the Lite/Standard Switch assuming battery life stay the same ?
How big the gap could be between the updated Switch and the Pro ?
About 2x across the board, with 34 gb/s bandwidth. A pretty nice upgrade, really.
But then there would be no Pro right ? The updated Switch would serve as a "Pro" ?
The problem there is that they would in essence be discontinuing the original Switch. Probably pretty early in the year. That would be pretty strange IMO.
Yep it doesn't line up in timing I agree. Plus, I am pretty sure the strategy is to have another member in the Switch family.
nah, there's no point in making two variations of the same chip. that just means that they'll have stock sitting, waiting to be sold if one variation doesn't sell well
yes, because it'd mean more chips per wafer
- Status