• 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.
  • Furukawa Speaks! We discuss the announcement of the Nintendo Switch Successor and our June Direct Predictions on the new episode of the Famiboards Discussion Club! Check it out here!

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

Reader mode
Nvidia may already be shipping Drake to non-Nintendo customers.

Nividia released Drive OS 6.0 on November 30th. This is the commercial L4T version that contains closed source software and is designed only to run on their Drive platforms (Orin and Xavier). Drive contains a (closed source?) Vulkan SC implementation - Vulkan SC is a version of Vulkan that is stripped down to make it easier to audit for Safety Critical applications, and each binary must be linked against the Vulkan SC library specific to your Drive GPU.

The documentation includes references to GA10F and the SDK (which is not public and I do not have access to) contains precompiled GA10F binaries for all the Vulcan SC sample code

It's somewhat strange that Nvidia would ship an A78C based chip into the Drive market, especially when paired with Vulcan SC. But it seems possible that Drake is either the entry level ADAS chip (instead of Orin Nano), or is being offered as an in-cab entertainment chip. While overpowered, it certainly would do the job, and would be the one place in the Drive stack that all of it's compromises and advantages over Orin - fewer TOPS, higher perf, lower safety CPU - would make sense.

The September upstreaming of the CPU driver tracks with Drive OS 6's release, as well as the fact that it's not been further pursued (as Drive OS is only available to paying customers, and has closed source components).

If this is true is also means that Nintendo isn't bound to the Drake production schedule.
 
wolftendo said:
I cracked the code:
  1. Early announcement within the next month-ish, Full reveal + trailers at E3, Holiday Season launch.
  2. Early announcement at E3, Full reveal + trailers during holiday season, Launch H1 2024.
Please do not read too much into this
Click to expand...
Remix of the mix 😂😂
 
0
oldpuck said:
Nvidia may already be shipping Drake to non-Nintendo customers.

Nividia released Drive OS 6.0 on November 30th. This is the commercial L4T version that contains closed source software and is designed only to run on their Drive platforms (Orin and Xavier). Drive contains a (closed source?) Vulkan SC implementation - Vulkan SC is a version of Vulkan that is stripped down to make it easier to audit for Safety Critical applications, and each binary must be linked against the Vulkan SC library specific to your Drive GPU.

The documentation includes references to GA10F and the SDK (which is not public and I do not have access to) contains precompiled GA10F binaries for all the Vulcan SC sample code

It's somewhat strange that Nvidia would ship an A78C based chip into the Drive market, especially when paired with Vulcan SC. But it seems possible that Drake is either the entry level ADAS chip (instead of Orin Nano), or is being offered as an in-cab entertainment chip. While overpowered, it certainly would do the job, and would be the one place in the Drive stack that all of it's compromises and advantages over Orin - fewer TOPS, higher perf, lower safety CPU - would make sense.

The September upstreaming of the CPU driver tracks with Drive OS 6's release, as well as the fact that it's not been further pursued (as Drive OS is only available to paying customers, and has closed source components).

If this is true is also means that Nintendo isn't bound to the Drake production schedule.
Click to expand...
It's ogre bros. We got too cocky
 
oldpuck said:
Nvidia may already be shipping Drake to non-Nintendo customers.

Nividia released Drive OS 6.0 on November 30th. This is the commercial L4T version that contains closed source software and is designed only to run on their Drive platforms (Orin and Xavier). Drive contains a (closed source?) Vulkan SC implementation - Vulkan SC is a version of Vulkan that is stripped down to make it easier to audit for Safety Critical applications, and each binary must be linked against the Vulkan SC library specific to your Drive GPU.

The documentation includes references to GA10F and the SDK (which is not public and I do not have access to) contains precompiled GA10F binaries for all the Vulcan SC sample code

It's somewhat strange that Nvidia would ship an A78C based chip into the Drive market, especially when paired with Vulcan SC. But it seems possible that Drake is either the entry level ADAS chip (instead of Orin Nano), or is being offered as an in-cab entertainment chip. While overpowered, it certainly would do the job, and would be the one place in the Drive stack that all of it's compromises and advantages over Orin - fewer TOPS, higher perf, lower safety CPU - would make sense.

The September upstreaming of the CPU driver tracks with Drive OS 6's release, as well as the fact that it's not been further pursued (as Drive OS is only available to paying customers, and has closed source components).

If this is true is also means that Nintendo isn't bound to the Drake production schedule.
Click to expand...
Interesting find.

Let's crowdfund a car so we can crack open the entertainment system! Il chip in 100$.
 
Last edited:
oldpuck said:
Nvidia may already be shipping Drake to non-Nintendo customers.

Nividia released Drive OS 6.0 on November 30th. This is the commercial L4T version that contains closed source software and is designed only to run on their Drive platforms (Orin and Xavier). Drive contains a (closed source?) Vulkan SC implementation - Vulkan SC is a version of Vulkan that is stripped down to make it easier to audit for Safety Critical applications, and each binary must be linked against the Vulkan SC library specific to your Drive GPU.

The documentation includes references to GA10F and the SDK (which is not public and I do not have access to) contains precompiled GA10F binaries for all the Vulcan SC sample code

It's somewhat strange that Nvidia would ship an A78C based chip into the Drive market, especially when paired with Vulcan SC. But it seems possible that Drake is either the entry level ADAS chip (instead of Orin Nano), or is being offered as an in-cab entertainment chip. While overpowered, it certainly would do the job, and would be the one place in the Drive stack that all of it's compromises and advantages over Orin - fewer TOPS, higher perf, lower safety CPU - would make sense.

The September upstreaming of the CPU driver tracks with Drive OS 6's release, as well as the fact that it's not been further pursued (as Drive OS is only available to paying customers, and has closed source components).

If this is true is also means that Nintendo isn't bound to the Drake production schedule.
Click to expand...
Use for ADAS is out of the question. All the machine learning and vision hardware of Orin was removed, the GPU was reconfigured back to the design of desktop Ampere, new gaming silicon like the FDE was added, the specifically automotive-focused A78AE were replaced with A78C, and the list goes on.

As for auto entertainment, I don't think a lower safety CPU would be permissible there either, since it would still have to be the SoC in the main console for the car, and it's not like A78Cs would be needed for performance. And either way it would just be absurd overkill for that too.

Can you think of an automotive scenario where Nvidia would want to offer this, or a customer would want to buy it, over the existing Orin options? I sure can't. But, let's say it was going to any customers besides Nvidia and Nintendo. Why would the chip even be confidential then?
 
Zemya said:
I hope TotK isn't just "more Botw" considering how long it took.
Click to expand...

I'm definitely with you on that. Covid or not, they started with a solid technical base, and they had 6 years while losing at most 1 year due to covid. Anything less than groundbreaking would be a disappointment in my book.
 
I would wait until it releases to deem it a dissapointment or underwhelming, not 4 months before

ShaunSwitch said:
I knew it, the switch 2.0's new gimmick is that it's a car.
Click to expand...
As a fellow train enthusiast, I'd rather have my switch metro system. Nintendo ads, a dock in every seat, and Waluigi announcing every stop.
 
I'm starting to doubt a fall release. Simply because if it came out 6 months after Zelda I feel Nintendo would just delay Zelda to have it aligned with Switch 2. If Switch 2 isn't getting announced in January/February and not releasing in May with Zelda then I think it's likely coming out March 2024 the earliest.
 
so maybe the die shrink theory makes the most sense, right? nintendo wasn't happy with the power consumption, heat, or both, so they decided to push back their plans until a die shrink that remedies those problems is more viable

if they really can at this point get die shrunk t239s into systems with a relatively short turnaround they could be in a great position, no? sales finally start to weaken, they flip a switch and have a new console ready to go in a year, or maybe even shorter
 
oldpuck said:
Nvidia may already be shipping Drake to non-Nintendo customers.

Nividia released Drive OS 6.0 on November 30th. This is the commercial L4T version that contains closed source software and is designed only to run on their Drive platforms (Orin and Xavier). Drive contains a (closed source?) Vulkan SC implementation - Vulkan SC is a version of Vulkan that is stripped down to make it easier to audit for Safety Critical applications, and each binary must be linked against the Vulkan SC library specific to your Drive GPU.

The documentation includes references to GA10F and the SDK (which is not public and I do not have access to) contains precompiled GA10F binaries for all the Vulcan SC sample code

It's somewhat strange that Nvidia would ship an A78C based chip into the Drive market, especially when paired with Vulcan SC. But it seems possible that Drake is either the entry level ADAS chip (instead of Orin Nano), or is being offered as an in-cab entertainment chip. While overpowered, it certainly would do the job, and would be the one place in the Drive stack that all of it's compromises and advantages over Orin - fewer TOPS, higher perf, lower safety CPU - would make sense.

The September upstreaming of the CPU driver tracks with Drive OS 6's release, as well as the fact that it's not been further pursued (as Drive OS is only available to paying customers, and has closed source components).

If this is true is also means that Nintendo isn't bound to the Drake production schedule.
Click to expand...
Maybe this is over my head but I'm having trouble understanding why the documentation talking about GA10F suggests this is Drake or uses A78C cores. GA10F is one of the code names for Orin, right?
 
Skittzo said:
Maybe this is over my head but I'm having trouble understanding why the documentation talking about GA10F suggests this is Drake or uses A78C cores. GA10F is one of the code names for Orin, right?
Click to expand...
GA10F is the name of Drake's GPU.
 
oldpuck said:
Nvidia may already be shipping Drake to non-Nintendo customers.

Nividia released Drive OS 6.0 on November 30th. This is the commercial L4T version that contains closed source software and is designed only to run on their Drive platforms (Orin and Xavier). Drive contains a (closed source?) Vulkan SC implementation - Vulkan SC is a version of Vulkan that is stripped down to make it easier to audit for Safety Critical applications, and each binary must be linked against the Vulkan SC library specific to your Drive GPU.

The documentation includes references to GA10F and the SDK (which is not public and I do not have access to) contains precompiled GA10F binaries for all the Vulcan SC sample code

It's somewhat strange that Nvidia would ship an A78C based chip into the Drive market, especially when paired with Vulcan SC. But it seems possible that Drake is either the entry level ADAS chip (instead of Orin Nano), or is being offered as an in-cab entertainment chip. While overpowered, it certainly would do the job, and would be the one place in the Drive stack that all of it's compromises and advantages over Orin - fewer TOPS, higher perf, lower safety CPU - would make sense.

The September upstreaming of the CPU driver tracks with Drive OS 6's release, as well as the fact that it's not been further pursued (as Drive OS is only available to paying customers, and has closed source components).

If this is true is also means that Nintendo isn't bound to the Drake production schedule.
Click to expand...
Just fell to my knees inside a Wal-Mart
 
The only thing 100% certain about Nintendo is that they will always come up with things totally out of left field.
 
LiC said:
Use for ADAS is out of the question. All the machine learning and vision hardware of Orin was removed, the GPU was reconfigured back to the design of desktop Ampere, new gaming silicon like the FDE was added, the specifically automotive-focused A78AE were replaced with A78C, and the list goes on.
Click to expand...
While Drake is overkill, I agree, is it inappropriate for ADAS? You don't need ASIL-D and double speed tensors core to put two red lines on a backup camera. Jensen was explicit about wanting to offer a chip which would replace that usecase for manufacturers who wanted to use a single software platform across the fleet.
LiC said:
As for auto entertainment, I don't think a lower safety CPU would be permissible there either, since it would still have to be the SoC in the main console for the car, and it's not like A78Cs would be needed for performance. And either way it would just be absurd overkill for that too.
Click to expand...
Does it? The incredibly short datasheet for for Drive IX, the in-cabin part of Drive, has the entire in cabin experience running on a different chip than any external sensing or driving control.

LiC said:
Can you think of an automotive scenario where Nvidia would want to offer this, or a customer would want to buy it, over the existing Orin options? I sure can't. But, let's say it was going to any customers besides Nvidia and Nintendo. Why would the chip even be confidential then?
Click to expand...
Fucked if I know man. I live in a city. I've never driven a car in my life ;) So it's not exactly my area of expertise.

But two things. One, I think it's arguable that the chip is confidential. It's listed in public docs, and while the Jetson devkits are publicly documented, the internals of the Drive platform are hidden behind actually having a contract with Nvidia. I can't tell you with confidence that Drake isn't all over Drive.

Second, I agree, 100%, it's a stretch - just speculation. But I would counter with asking why in the heck Nvidia would go out of the way to add support for t239 to the Drive platform if it wasn't part of the Drive platform? Drive OS only supports Xavier, Orin and now Drake, and seems to include a Vulkan SC driver that is unusable outside of Drive applications.
 
Quoted by: LiC
1
oldpuck said:
haha, let's not go overboard. @LiC is right about Drake being a bad fit, which is why it's just blue moon speculation based on little data.

Also those floors are bad on your knees
Click to expand...
Well it still puts in perspective what we thought were strong hints of Drake being close to production. We have a possible explanation that does not involve Nintendo hardware. Which puts us back to square one in terms of plausible expectations...
 
Pretty much the only reason I could see them using Drake in a car entertainment system, is the absence of better options. The lesser evil.
 
0
I'm guessing it could also just be holdover code or references since Orin and Drake are so closely linked? I dunno.
 
0
oldpuck said:
While Drake is overkill, I agree, is it inappropriate for ADAS? You don't need ASIL-D and double speed tensors core to put two red lines on a backup camera. Jensen was explicit about wanting to offer a chip which would replace that usecase for manufacturers who wanted to use a single software platform across the fleet.
Click to expand...
"Overkill" is kind of burying the lede. It's just fundamentally not the use case of the chip. I'm also very certain T239 doesn't even support cameras, since its list of modules has removed entries -- which Orin has -- for PVA (programmable vision accelerator), VI (video input), ISP (image signal processor), and NVCSI (camera serial interface).

oldpuck said:
Does it? The incredibly short datasheet for for Drive IX, the in-cabin part of Drive, has the entire in cabin experience running on a different chip than any external sensing or driving control.
Click to expand...
A different chip -- as in a separate SoC on the same board -- might handle infotainment separately from an SoC doing the sensors and driving. I don't think that exempts it from safety requirements.

But more than this technical argument which neither of us really know anything about, the important point is more that whatever in-vehicle entertainment would be needed in the Drive AGX Orin platform would be provided by AGX Orin. It already comes in all kinds of shapes and sizes, from more powerful than Drake to less powerful, across any given workload. It's already designed and certified for cars. There's no conceivable reason not to use it.

oldpuck said:
One, I think it's arguable that the chip is confidential. It's listed in public docs, and while the Jetson devkits are publicly documented, the internals of the Drive platform are hidden behind actually having a contract with Nvidia. I can't tell you with confidence that Drake isn't all over Drive.
Click to expand...
I feel like this is trying to have it two ways? It shows up in documentation (in fleeting references), but Nvidia never announced it, never put it on a roadmap, and at this very moment does not list it as a supported product in the SDK for the Drive OS we're discussing. Anyone could order Orin devkits all last year, but you still can't order a T239 devkit. If it's "all over" Drive under contract, but this invisible outside it, even in the public documentation of Drive, that is confidentiality.

And Nvidia announces everything. There's literally one other production chip that Nvidia never publicly mentioned, and that's Tegra X1+, a die shrink of an existing chip used only internally and by Nintendo. We have a full list of leaked Nvidia GPUs and SoCs, so we can say for certain that if this is what was actually going on, it would be a first.

T239 being used by someone other than Nintendo is possible, even expected. It's just 3,000% too much of a stretch to me that being hidden away in Drive as a very bad substitute for Orin is that.
 
CocaFormula said:
Well it still puts in perspective what we thought were strong hints of Drake being close to production. We have a possible explanation that does not involve Nintendo hardware. Which puts us back to square one in terms of plausible expectations...
Click to expand...
yeah, we found the, shall we say, victim to our smoking gun

back to square one indeed
 
0
ShaunSwitch said:
I knew it, the switch 2.0's new gimmick is that it's a car.
Click to expand...
Nintendo copying sony confirmed

Raccoon said:
so maybe the die shrink theory makes the most sense, right? nintendo wasn't happy with the power consumption, heat, or both, so they decided to push back their plans until a die shrink that remedies those problems is more viable

if they really can at this point get die shrunk t239s into systems with a relatively short turnaround they could be in a great position, no? sales finally start to weaken, they flip a switch and have a new console ready to go in a year, or maybe even shorter
Click to expand...
no it doesn't. these are things that they would have known about in their simulations. they were already producing 8nm GPUs when this project started so they know how the power scales. there is no surprise late stage die shrink
 
ILikeFeet said:
no it doesn't. these are things that they would have known about in their simulations. they were already producing 8nm GPUs when this project started so they know how the power scales. there is no surprise late stage die shrink
Click to expand...
damn

I wonder what changed, if anything
 
Skittzo said:
But for real what if the Lapsus hack was a very very very elaborate fake? Some people would probably go to jail to troll Nintendo fans
Click to expand...
Then I would slightly tip my hat* to Nintendo and carry on with my day

*Editor's note: I am not wearing a hat
 
0
Question...... what the hell happened in the last 2 weeks I just had the time to check the thread and I see a post about a switch train are we going wacko mode?!?!?!?
 
0
Serif said:
If T239 will be used in vehicles this year, wouldn't there have been some hint at CES earlier?
Click to expand...
I agree with that, I would had expected some form of hint or announcement during CES that the T239 drake would be used in upcoming vehicles...Just would be odd that a totally custom chip would be used in a car. Wouldn't that be an expensive development for an auto company to create their own SoC through nvidia? it would had been more cost effective to use Orin AGX or nano if it was an auto manufacturer.
 
0
yohoopzcrazy said:
I was a Grinch Leak believer... those scars still haven't fully healed...
Click to expand...
I mean just to clarify this is just an errant thought, there's really very little chance of that somehow being fake. It would just help explain a bunch of stuff.

But I don't believe it can be fake.
 
0
@oldpuck If I understand your post clearly, "GA10F" GPU is referenced in the Drive OS 6 documentation and also in the Vulcan SC sample codes. Sorry at the moment I don't have time to check this myself, but I assume that you also searched for "T239" or "Drake" but found no reference of either, correct? If that's case, wouldn't this be an indicator that Nvidia is/was planning another variant of Orin with the GA10F GPU (instead of the GA10B on the current Orin variants)? It doesn't seem a probable evidence of the T239/Drake being part of the Drive platform.
 
Without even mentioning the storage capacity that the future hardware will have, do you think we can reasonably expect more frequent use of larger capacity game cartridges? I'm thinking not only of third-party games, but also of ambitious Nintendo games like Prime 4, future Zeldas, or even Xenoblades. If The Witcher 3 with all its DLC got a physical release, I hope Nintendo will be able to make that kind of effort. But I admit I don't know how difficult it will be financially to increase the capacity of the cartridges or even Drake's internal storage.
 
boodream said:
Without even mentioning the storage capacity that the future hardware will have, do you think we can reasonably expect more frequent use of larger capacity game cartridges? I'm thinking not only of third-party games, but also of ambitious Nintendo games like Prime 4, future Zeldas, or even Xenoblades. If The Witcher 3 with all its DLC got a physical release, I hope Nintendo will be able to make that kind of effort. But I admit I don't know how difficult it will be financially to increase the capacity of the cartridges or even Drake's internal storage.
Click to expand...
Nintendo can easily eat the costs of higher capacity cards, third parties can't. So there wouldn't be that much more usage
 
fwd-bwd said:
@oldpuck If I understand your post clearly, "GA10F" GPU is referenced in the Drive OS 6 documentation and also in the Vulcan SC sample codes. Sorry at the moment I don't have time to check this myself, but I assume that you also searched for "T239" or "Drake" but found no reference of either, correct? If that's case, wouldn't this be an indicator that Nvidia is/was planning another variant of Orin with the GA10F GPU (instead of the GA10B on the current Orin variants)? It doesn't seem a probable evidence of the T239/Drake being part of the Drive platform.
Click to expand...
I don't think GA10F exists outside T239. And there's no other T23x SoC out there to be the "Orin but with GA10F."

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

I don't know why GA10F is in that documentation, but I would believe almost any random explanation more than I believe that GA10F/T239 actually exists in Drive products but Nvidia is keeping it secret for some reason. If it was being used in cars, even despite its unsuitability and the existence of Orin (as in, the "Drive AGX Orin" platform we're talking about here) to already fulfill its role, then Nvidia would have announced it like they do every other product.
 
Reader mode
You must log in or register to reply here.
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