NVIDIA Tegra Architecture

I think Nvidia Shield has fan ... helps to manage the heat dissipation and score higher than other T4 Tablet ..

21613-116.jpg
 
I suppose the Toshiba implementation of Tegra 4 with its heat problems is Toshiba pushing the chip too far in the fanless form factor. Notebook vendors are doing the same thing and relying on throttling. They win benchmarks and can make a svelte package but the user experience is compromised in some scenarios.
 
I suppose the Toshiba implementation of Tegra 4 with its heat problems is Toshiba pushing the chip too far in the fanless form factor. Notebook vendors are doing the same thing and relying on throttling. They win benchmarks and can make a svelte package but the user experience is compromised in some scenarios.

Things like that are generally caught and called out at sites like Anandtech, for example. But we generally don't see performance dropping by up to 50% on other ARM based SOCs over an extended and repeated run. If and when they do throttle, it's usually from start to finish as they throttle to maintain their advertised power consumption which in turn maintains their advertised TDP.

Tegra 4 basically needs either much lower clocks to operate in a fanless tablet, or if it wants the higher clocks and thus performance it requires active cooling. Either that or Nvidia just gave Toshiba all their "bad" chips. :p

Regards,
SB
 
This Toshiba tablet is definitely a poor design. But I think it's premature to say it's T4's fault. The biggest power consumer in this tablet is not the SoC, but the display, and the way they've integrated the display and SoC seems terrible. Let's see if someone else does a better job designing the system before judging the SoC.
 
But if the GPU under heavy load while the CPU is under light load throttles then drops to 50% performance after being used for a while then what's the point of even having this level of GPU capability in this tablet? 10 runs of 3dmark is nothing compared to an extended gaming session. And if games can't and won't even utilize that level of performance to begin with then I still ask, what's the point?
I see what you're getting at, but have you considered the opposite question: why would you not use the thermal headroom available? Especially as mobile GPUs are now used for some tasks that benefit from maximum burst performance.

Being aware that throttling exists, reviewers and benchmark developers need to take the necessary steps to ensure they test what they actually want to test (i.e. if you want to test sustainable games performance, you can't do that with a 2 minute run).
 
What is the expected competition for Grey? I figure Qualcomm's S600, and perhaps a refreshed QC lineup with Krait 400 everywhere, but what about other companies?

Does Mediatek have integrated LTE, for example?

Grey/T4i is aiming for mainstream smartphones.
 
Asus Eee Pad / T4 same frequencies as Shield yet lower performance.

Clearly the T4 variant in a tablet is not using the same GPU max clock operating frequencies as the T4 variant in Shield. And with respect to the CPU, the max clock operating frequencies are likely different when more than one CPU core is active.

Bad news for Grey

NVIDIA stated many months ago that T4i "Grey" will not be available until early 2014, so nothing new there. Carrier certifications take time, there is no good way to shorten that process.
 
Last edited by a moderator:
I see what you're getting at, but have you considered the opposite question: why would you not use the thermal headroom available? Especially as mobile GPUs are now used for some tasks that benefit from maximum burst performance.

Being aware that throttling exists, reviewers and benchmark developers need to take the necessary steps to ensure they test what they actually want to test (i.e. if you want to test sustainable games performance, you can't do that with a 2 minute run).

If there are legitimate uses for hurry up and sleep that utilize on mobile GPU's today then that's worthwhile, although like you say I think that sort of performance testing has to be a whole different category from gaming. I get the impression that there isn't an awful lot of that going on yet, maybe image processing from the camera.

In my mind turbo stuff was always about balancing available resources so you can get as much performance as possible for a given task. So you can clock one CPU core higher if the other ones are asleep or you can clock GPU higher if the CPU load isn't too demanding. The benefit of being able to go past some power consumption limit for a short period of time if you're guaranteed to then go significantly lower right afterwards seems secondary to me, although I know that Intel for instance strongly promotes this.
 
That is the million dollar question :) Who knows what will be available in mainstream phones by early 2014, but T4i should be competitive at it's price point in that timeframe (assuming that it is priced close to $200-$300 unsubsidized). It is my understanding that, pricing wise, it will be closer to the S400 SoC.
 
Assuming Qualcomm is the marketshare leader, Nvidia may have to make price concessions to draw customers to use their SOC.
 
On an aside, Tegra automotive is making some headway. Tegra automotive revenue is said to be $100 million for Fiscal Year 2014 (ie. this fiscal year), $200 million for Fiscal Year 2015, and $400+ million for Fiscal Year 2016. Note that these are long-term contracts that are already in place. If NVIDIA hits their stride with Logan (and all signs point that way), Tegra will be a really good business for them.
 
On an aside, Tegra automotive is making some headway. Tegra automotive revenue is said to be $100 million for Fiscal Year 2014 (ie. this fiscal year), $200 million for Fiscal Year 2015, and $400+ million for Fiscal Year 2016. Note that these are long-term contracts that are already in place. If NVIDIA hits their stride with Logan (and all signs point that way), Tegra will be a really good business for them.

Marketing wash to keep stock holders and investors in place. At least NVIDIA sets examples in cases like that, unlike IMG for instance where with the idiotically phrased market warnings their stock values have dropped below half their value and are still struggling to recover.
 
At CES this year, didn't Audi partner with nVidia to integrate Tegra into the A3? Doesn't this

http://www.wired.com/autopia/2013/08/audi-ekurzinfo/

kind of defeat the purpose of having Tegra in the car? From the article, it seems like Tegra is non-existant (in the car); shouldn't you be able to access the same or similar "stuff" from the "Tegra screen/dash"?

Also, http://gigaom.com/2013/08/15/the-fi...-debuts-in-europe-u-s-must-wait-until-spring/

So what's the point of Tegra in cars and why didn't they fully utilize a Tegra3 (since it should already have been in the car and have LTE...right?)? Purely infotainment and virtual dashboard?

And switching to tablets, going by what Taner Ozcelik (nVidia’s general manager, automotive) said about the modularity of Tegra http://www.digitaltrends.com/ces/audi-nvidia-tegra-powered-infotainment-system/ he makes it sound like it's a drop-in change/"upgrade" (a la CPU). I'm specifically refering to this quote

“Nvidia’s modular VCM approach lets companies like Audi quickly move from a Tegra 2 processor, to a Tegra 3 and beyond,”

Why aren't there more Tegra4 tablets if it supposedly allows companies to quickly move to a new Tegra processor? So far, there's only the Toshiba, the HP SlateBookX2 (which is double the thickness of the Toshiba-0.81" depth vs 0.4" for the Toshiba) and an upcoming Asus (infinity?).
 
Just what I always wanted: an AR app to scan over the dashboard with my phone like an idiot. ;)

Irrespective of whether or not Tegra is in a cars, this AR app is not going to be the reason for it not to be.

What I do see, is that many families are moving to a one tablet per family member situation (or past it). Give it another 5 years and nobody will care about those fixed screens against the front seat headrest.

But the dashboard has still ways to go...
 
As for switchable modules: he never said that the dies are pin compatible. A switchable module can work when form factor isn't an issue. That's never going to happen for a tablet...
 
Apple has an initiative with car makers to integrate with their devices, to mirror the display on the devices to the dash screen.

But car makers also sell those expensive tech packages with Nav, Bluetooth, HID, etc. They may need to put smarts in those packages to justify the prices they charge but any mobile hardware they put in will quickly become outdated.
 
Back
Top