NVIDIA Tegra Architecture

As I said above, I disagree. Some gamepads are really compact, functional and even act as powerbanks for your smartphone. Battery life will only be a problem for people who are on the go without a power plug for the whole day, people who sit at desks like students and most IT/bureaucracy/law/etc. related jobs can probably get easy access to a charger and nowadays many phones will charge really fast up to ~80%.

Regardless, we'll never know because those games are locked for good to an irrelevant amount of people.

It seems that most AAA companies do not agree with you though. Anyone can do games or apps for Android (I do), there is nothing preventing them from that. The hardware and software is available. If there was a valid use case and money to be done, they would. I really do not get your point. How is it that nVIDIA prevents anyone from doing what they wish on such an open platform as Android? It is not like nVIDIA has a gun pointed to their head o_O If they have ports to their hardware is because they are PAYING for them. Without such subsidies companies will not do AAA games for Android, leading me back to the initial point: Android is not ideal for gaming.

EDIT - That gamepad is compact, but is it functional, really? Do you think that that size will fit everyone? It looks even more cramped with worse ergonomics than the PS VITA.
 
Last edited:
The ports of Half Life 2 and Doom3 were surely funded by NV to help sell the Shield Tablet. I think of these games as NV's "exclusives". Just like what MS and Sony do on a bigger scale. They really need touchscreen support though.

I did buy a Shield gamepad to experiment with with my Shield K1 tablet, and it's great, but most of the time I would prefer to use touch controls. Android game support for it is very weak though and the gamepad mapper, while capable, usually adds like 30 minutes setup time because there's always trial and error.
 
Last edited:
It usually adds like 30 minutes setup time though because there's always trial and error.

And that is exactly why Consoles are more popular among consumers than PCs. Put the disc in, let it install and you have a standard controller, no fuss, no definitions to care about. Similar experience to casual gaming on Android (your Temple Runs, Candy Crush, Card Games), hence why its the only genre that is successful. Android mobile phones come in a huge diversity and fragmentation both in hardware and software. Its a nightmare to code for, which coupled with low expectation on return makes it a no go for AAA title companies.
 
And that is exactly why Consoles are more popular among consumers than PCs. Put the disc in, let it install and you have a standard controller, no fuss, no definitions to care about. Similar experience to casual gaming on Android (your Temple Runs, Candy Crush, Card Games), hence why its the only genre that is successful. Android mobile phones come in a huge diversity and fragmentation both in hardware and software. Its a nightmare to code for, which coupled with low expectation on return makes it a no go for AAA title companies.
Yeah I agree completely. I don't really consider a gamepad interesting for a tablet or phone. I just wanted to try the Shield gamepad out with my Shield K1. I connected the tablet to my TV and played a few games that way. It's a strange, very quirky, very limited game console that way.

I hear you on the difficulties developers have. The platform is a mess and the games usually sell terribly. The KOTOR port for Android was interesting. I saw the Aspyr port developers ask the community to help them identify devices with hardware that could possibly run the game, because the Android Play store database isn't detailed enough or some such thing. And then these companies get their game working and the game sells 2,000 copes at $6. Yay. lol. KOTOR appears to have sold somewhere above 100,000 copies for < $10. Oh boy.
 
Last edited:
I hear you on the difficulties developers have. The KOTOR port for Android was interesting. I saw the Aspyr port developers ask the community to help them identify devices with hardware that could possibly run the game, because the Android Play store database isn't detailed enough or some such thing. And then these companies get their game working and the game sells 2,000 copes at $6. Yay. lol. KOTOR appears to have sold somewhere above 100,000 copies for < $10. Oh boy.

I am but a tiny small developer on my free time and its annoying as hell to have, for example, custom implementations on libraries by OEMs like Samsung, Sony, etc. on their baked ROMs sometimes breaking functionality with each Android version. Its not like Windows where everything is standard and "expected". The amount of maintenance that an Android app/game requires is very big, especially large ones. Why do you think you see so many complaints in the Play Store about "update broke it, does not work anymore?" But noooooooooooo, nVIDIA is the big bad wolf who does not let Android gaming grow :rolleyes:
 
I am but a tiny small developer on my free time and its annoying as hell to have, for example, custom implementations on libraries by OEMs like Samsung, Sony, etc. on their baked ROMs sometimes breaking functionality with each Android version. Its not like Windows where everything is standard and "expected". The amount of maintenance that an Android app/game requires is very big, especially large ones. Why do you think you see so many complaints in the Play Store about "update broke it, does not work anymore?" But noooooooooooo, nVIDIA is the big bad wolf who does not let Android gaming grow :rolleyes:
Yeah I've seen lots of examples of broken games on new OS releases. On various devices, including my Shield Tablet and Tegra Note 7. And yeah there are loads of abandoned games on the Play Store. Good thing there's that refund option.
 
And that is exactly why Consoles are more popular among consumers than PCs. Put the disc in, let it install and you have a standard controller, no fuss, no definitions to care about. Similar experience to casual gaming on Android (your Temple Runs, Candy Crush, Card Games), hence why its the only genre that is successful. Android mobile phones come in a huge diversity and fragmentation both in hardware and software. Its a nightmare to code for, which coupled with low expectation on return makes it a no go for AAA title companies.
That might have been true, but many if not most new console games need those day-0 patches and all
 
That might have been true, but many if not most new console games need those day-0 patches and all
That's different. That's some kind of recurring rush-to-release-before-it's-ready business decision thing. There's no reason a PS4 or Xbox game can't be perfect for everyone on release. The platforms are known quantities.
 
Last edited:
So we may not get a dedicated NVIDIA keynote at CES 2017, and as such no new news on future Tegra products.

It has been announced today that Jen-Hsun is giving the CES opening keynote. This doesn't strictly clash with NVIDIA's events (which are usually a day or two earlier), but it's easy to imagine that Jen-Hsun won't want to give two keynotes about NVIDIA in half of a week. And since the opening keynote tends to be more topic/trend focused, I could easily NVIDIA using this as an opportunity to break away from their yearly product cadence there.
 
Last edited:
Back at the CES in January we saw Nvidia launch its first 'in-car supercomputer', the DRIVE PX2. Partners such as Audi, BMW, Ford, Mercedes-Benz and Volvo lined up were already lined up to use the roadworthy AI system. At the same time we learnt about the supporting Nvidia DriveWorks software and DRIVENET deep neural net technologies.

Now Nvidia has unveiled a new DRIVE PX 2 for auto makers, pictured above. The new in-car computer is palm-sized and energy efficient but offers enough processing power for automated and autonomous vehicles to use for driving and mapping. Consuming just 10 watts, the new single processor DRIVE PX 2 is targeted at providing AutoCruise functions, "which include highway automated driving and HD mapping".


5d9969e9-942f-4a93-8a1a-fc1baf7e28c0.jpg

http://hexus.net/ce/news/automotive/96835-nvidia-unveils-palm-sized-single-soc-version-drive-px2/
 
I find it interesting that Nvidia are now inferring you need 2 x fully configured Drive PX2 for a fully autonomous solution, now would that also mean 1 GPU per PX2 or 1GPU per Parker for 4 total; I assume they mean 2 GPU per PX2 for 4 total for their context of fully configured.
DRIVE PX 2 FOR FULLY AUTONOMOUS DRIVING
Multiple fully-configured DRIVE PX 2 systems can be integrated in a single vehicle to enable autonomous driving.

But that info is in their blog so maybe will change *shrug*.
Cheers
 
How bad is it that I thought that was a joke at first?

If it was a brainfart, then we all have those more than we all would want to. Apart from that I don't remember where but NV had stated that PX2 is scalable from 1 to 4 chips.
 
If it was a brainfart, then we all have those more than we all would want to. Apart from that I don't remember where but NV had stated that PX2 is scalable from 1 to 4 chips.
The architecture allows one discrete GPU per Parker and a board can be one or two Parker processors, so as the above picture shows you would use two PX2 full boards to achieve 4 Parker processors.
Cheers
 
The architecture allows one discrete GPU per Parker and a board can be one or two Parker processors, so as the above picture shows you would use two PX2 full boards to achieve 4 Parker processors.
Cheers

Meaning that the Autocruise PX2 also comes with dedicated Pascal GPU on board besides Parker? Have a look at the text for each module/config:

DRIVE PX 2 FOR AUTOCRUISE
Small form factor, DRIVE PX 2 for AutoCruise, is designed to handle functions including highway automated driving, as well as HD mapping. This platform will be available in Q4 2016.

DRIVE PX 2 FOR AUTOCHAUFFER
DRIVE PX 2 configuration with two SoCs and two discrete GPUs is available today for point to point travel.

DRIVE PX 2 FOR FULLY AUTONOMOUS DRIVING
Multiple fully-configured DRIVE PX 2 systems can be integrated in a single vehicle to enable autonomous driving.

The 3rd above is just double the 2nd module as it is. With scaling from 1 to 4 chips I meant from 1 SoC per module only (autocruise) to 2 SoCs + 2 dGPUs (autochauffer). Nr3./Autonomous driving would be then rather 4+4 than 2+2 or autochauffer@SLi if you prefer :p
 
No idea how many GPUs are needed for each option and probably depends upon tasks-commands, it will be interesting to see if the Autocruise PX2 can work without one discrete GPU (I guess it depends how advanced they want to make it).
Maybe we are at cross purposes.
The architecture historically shows you can have a max of 2 Parker processors and 2 discrete GPUs for a single PX2 solution (full configuration), but it does seem that one full configured PX2 is not enough for complete autonomous driving - maybe this is where we are thinking the same but cross wires.

Unlike before it does seem this time though they infer one needs multiple full configured PX2 for autonomous driving, this seems a newish direction, probably because they are finally getting somewhere with their driving solutions and the experience gained.
So I think we agree just crossed wires.
Cheers
 
It was an honest question about the autocruise module, so if anyone knows a bit more about it any added information is welcome.

As for fully autonomous driving I can understand why it needs such obscene heights of processing power. If it'll ever come to full autonomy for driving in real time is a whole chapter of its own and quite OT for this thread. Long story short IMHO: yes to it for the foreseeable future but NOT for all conditions. I wouldn't want to sit in a airplane either which flies under all conditions with the autopilot turned on.
 
It was an honest question about the autocruise module, so if anyone knows a bit more about it any added information is welcome.

As for fully autonomous driving I can understand why it needs such obscene heights of processing power. If it'll ever come to full autonomy for driving in real time is a whole chapter of its own and quite OT for this thread. Long story short IMHO: yes to it for the foreseeable future but NOT for all conditions. I wouldn't want to sit in a airplane either which flies under all conditions with the autopilot turned on.
Ah OK.
Well at most it can support is one discrete GPU as the architecture is 1 Tegra Parker to 1 discrete GPU.
Looking into it some more at Nvidia, seems the autocruise is without disctrete GPU as it consumes 10W.
The new single-processor configuration of DRIVE PX 2 for autocruise functions -- which include highway automated driving and HD mapping -- consumes just 10 watts of power and enables vehicles to use deep neural networks to process data from multiple cameras and sensors.
So yeah looks like just a more advanced-safety cruise control but it is going to have its limits, still better than existing solutions and something that could also be considered for other transport infrastructure/vehicles.
And yeah totally with you on those considerations, especially cars that have so many environmental and random variables.

Cheers
 
Back
Top