Tegra revenue down 48% Y/Y

[Homeles]

LOL, NVIDIA's "bread and butter" is most certainly not low end GPU's!

You're serious, aren't you?

 

[sontin]

They clearly have their set of APIs that they want developers to stick to- it makes sense, as they want to prevent fragmentation. And besides, what developer would write CUDA code for Android when only a tiny fraction of consumer devices will ever even support CUDA?

Cuda is not bound to Android. If you write one application with Cuda and OpenGL you can port it to Windows, Android and Linux systems.

And even if full OpenGL were available, would the next Angry Birds use it if only a fraction of top end devices were capable of supporting it? OpenGL ES is the guaranteed lowest common denominator, and that's not going to change. Just look at how long we were stuck with DirectX 9 games on PC.

With Tegra K1 there is the first kind of ARM SoC which allow developers to sell their games on more than 1-3 plattforms.
Here is a video showing someone playing TF2 on a Tegra K1 dev plattform with Ubuntu:
https://www.youtube.com/watch?v=hRBPeNzE558

 

[NTMBK]

NVIDIA has never been a big player in the "mainstream" laptop market. They can and will add value to premium laptops and gaming laptops.

When I bought my Core 2 Duo laptop back in the day, there was no way in hell I would have settled for Intel GMA graphics; I went with ATi, and my flatmate went with NVidia. These weren't high end laptops, they were pretty cheap bog-standard laptops- we were just students, we weren't shopping Alienware! But that kind of market is dead now. Intel integrated is simply the best solution for standard laptops, in terms of better battery life, smaller form factor (from integrated cooling), and lower costs.

It is almost inevitable that Windows Phone and Windows RT will merge in the somewhat near future to create one Windows on ARM consumer OS. Almost every single major SoC hardware vendor other than Intel is moving towards using ARM CPU's in the consumer space (including AMD).

AMD have not announced any plans for consumer ARM chips, only server parts- and are pushing ahead with the x86 Mullins APU aimed at consumer tablets. As for the rest of them, of course they're using ARM- it's not like Intel are going to give them an x86 license, is it?

 

[ams23]

Of course the Phi uses a different architecture, but in terms of potential customers you better believe that they're going after the same people.

And so what? Both NVIDIA and Intel can add value to HPC and Exascale computing, and there are plenty of customers for both of them.

NVidia's profits come from selling GPUs, and the bulk of the volume of GPUs are sold in the low end.

The lowest margin and least profitable GPU's for NVIDIA are low end discrete GPU's, period.

No, it's not inevitable. OpenCL has already been blocked deliberately by Google.

If Android is to ever become a serious platform for gaming, it will absolutely need to support full OpenGL.

 

[NTMBK]

Cuda is not bound to Android. If you write one application with Cuda and OpenGL you can port it to Windows, Android and Linux systems.

Or you could write it with just OpenGL and port it to the other 99% of ARM consumer SoCs? Or with OpenCL- except Google have blocked use of OpenCL on Android, just like they probably will with CUDA.

With Tegra K1 there is the first kind of ARM SoC which allow developers to sell their games on more than 1-3 plattforms.
Here is a video showing someone playing TF2 on a Tegra K1 dev plattform with Ubuntu:
https://www.youtube.com/watch?v=hRBPeNzE558

OpenGL ES is now a strict subset of OpenGL- if you target OpenGL ES, then the same code will also run on fully fledged OpenGL platforms. If anybody actually cared about portability (and supporting 99% of Android devices out there), they would use OpenGL ES.

 

[NTMBK]

And so what? Both NVIDIA and Intel can add value to HPC and Exascale computing, and there are plenty of customers for both of them.

How many supercomputers do you really think get built every year? Sure, the HPC market shifts thousands and thousands of accelerators, but they are split between a very small number of individual sites. A single big customer going Intel instead of NVidia means tens of thousands of GPUs that NVidia won't sell. They're in direct competition, I'm not really sure how you can possibly argue about that. (Of course NVidia is clearly winning the competition at this point, but that's a different discussion.)

The lowest margin and least profitable GPU's for NVIDIA are low end discrete GPU's, period.

Lowest profit per GPU, yes, but they sell a MUCH higher number of them.

If Android is to ever become a serious platform for gaming, it will absolutely need to support full OpenGL.

What makes you think Google care about that? The only company currently trying to push Android as a "serious gaming" platform is NVidia, for obvious reasons. Google want to focus on making an excellent tablet and phone platform, not a mediocre games console platform.

 

[ams23]

Or you could write it with just OpenGL

OpenGL is a graphics API that is used for a wide variety of gaming applications. CUDA is a parallel programming model that is used for very specialized compute applications.

 

[sontin]

Or you could write it with just OpenGL and port it to the other 99% of ARM consumer SoCs? Or with OpenCL- except Google have blocked use of OpenCL on Android, just like they probably will with CUDA.

Android is open, the application is talking directly with the driver. They can't remove the ability because for that they need control over the driver.

Google has removed OpenCL from their devices but not from Android.

OpenGL ES is now a strict subset of OpenGL- if you target OpenGL ES, then the same code will also run on fully fledged OpenGL platforms. If anybody actually cared about portability (and supporting 99% of Android devices out there), they would use OpenGL ES.

OpenGL ES lacks to many features and there is no opposite on the Windows front. Using OpenGL ES limit the developers to Android and iOS.

 

[NTMBK]

Android is open, the application is talking directly with the driver. They can't remove the ability because for that they need control over the driver.

Google has removed OpenCL from their devices but not from Android.

Google has been steadily locking down their platform, actually. Ars Technica did a good series of articles on it, I'd highly recommend them- latest is here. The basic gist is that you can go it alone with Google and just use Android Open Source Project, but good luck without up to date versions of core apps like the internet browser or text messaging, and with no access to the Google Play Store.

OpenGL ES lacks to many features and there is no opposite on the Windows front. Using OpenGL ES limit the developers to Android and iOS.

Did you literally not read what I wrote? D: OpenGL ES is now a strict subset of OpenGL. That means that literally any valid OpenGL ES code is also valid OpenGL code. You can write an OpenGL ES app and it will run just fine on OpenGL, but not the other way around.

 

[ams23]

They're in direct competition

And again, so what? They are competing against each other using completely different architectures, and have different strengths and weaknesses in this area.

Lowest profit per GPU, yes, but they sell a MUCH higher number of them.

LOL, so you have statistics on NVIDIA GPU's sold by model number in this day and age? If NVIDIA's low end GPU business is hurting them as much as you claim, then why are FY 2014 GPU revenues 7% higher than FY 2013 GPU revenues (from the link in your very first post)?!?

What makes you think Google care about that?

Do you think Google is stupid? Of course they (and others) would want to promote Android as a viable gaming platform for the future.

 

[jpiniero]

then why are FY 2014 GPU revenues 7% higher than FY 2013 GPU revenues (from the link in your very first post)?!?

The Crypto Coin Craze, of course.

 

[xpea]

Choice is great, but when the choice is between a company that consistently delivers subpar products, and a couple of companies that don't, it's fairly obvious that there's no need for the subpar company.

good, so if I follow you, we don't need intel and his subpar GPUs, we don't need mediatek and their overall sh*t, we don't need Qalcomm and their broken paper feature set (OGL3.0 on 320), and so on... :hmm:

 

[xpea]

Did you literally not read what I wrote? D: OpenGL ES is now a strict subset of OpenGL. That means that literally any valid OpenGL ES code is also valid OpenGL code. You can write an OpenGL ES app and it will run just fine on OpenGL, but not the other way around.

you still don't understand. OGL ES is a crippled useless low end API that can't use for some applications, like the new Audi dashboard running on TK1.

 

[NTMBK]

you still don't understand. OGL ES is a crippled useless low end API that can't use for some applications, like the new Audi dashboard running on TK1.

Can't have a dashboard without tessellation!

 

[mrmt]

LOL, NVIDIA's "bread and butter" is most certainly not low end GPU's!

Nvidia business model is built around Quadro making the bulk of the profits and GeForce bearing the brunt of R&D costs by spreading the bill for millions of cheap dGPU chips.

Kill that bottom market and suddenly Nvidia doesn't have as many units to foot the R&D bill, and that means either longer R&D cycles and/or smaller improvements between generations.

 

[ams23]

Nvidia business model is built around Quadro making the bulk of the profits and GeForce bearing the brunt of R&D costs by spreading the bill for millions of cheap dGPU chips.

Kill that bottom market and suddenly Nvidia doesn't have as many units to foot the R&D bill, and that means either longer R&D cycles and/or smaller improvements between generations.

Their business model is not quite that simplistic nowadays. Tesla, Quadro, Grid, and high end Geforce are all highly profitable segments that lift gross margins. And while it is true that lower end discrete revenues are declining, this is offset by higher revenues in other areas. At the end of the day, NVIDIA's GPU revenue actually increased this latest fiscal year vs. last fiscal year, which indicates that their GPU business is sufficiently diversified to withstand dips in some areas.

 

[mrmt]

Their business model is not quite that simplistic nowadays. Tesla, Quadro, Grid, and high end Geforce are all highly profitable segments that lift gross margins. And while it is true that lower end discrete revenues are declining, this is offset by higher revenues in other areas. At the end of the day, NVIDIA's GPU revenue actually increased this latest fiscal year vs. last fiscal year, which indicates that their GPU business is sufficiently diversified to withstand dips in some areas.

Maybe you should check their annual reports. Tesla isn't profitable, Grid is non-existant and Geforce just moved to blue after Kepler, but they are not that big. The bulk of the profits is Quadro.

 

[ams23]

Maybe you should check their annual reports. Tesla isn't profitable, Grid is non-existant and Geforce just moved to blue after Kepler, but they are not that big. The bulk of the profits is Quadro.

That is incorrect. Tesla, Quadro, Grid, and High End Geforce all lift gross margins, they don't drag gross margins! Maybe when Tesla was a very nascent business it was not profitable, but I doubt that would be the case today.

 

[mrmt]

That is incorrect. Tesla, Quadro, Grid, and High End Geforce all lift gross margins, they don't drag gross margins!

Gross margins don't have a direct relationship with profits.

 

[ams23]

Gross margins doesn't have anything to do with profits.

Gross margins obviously indicate a measure of revenue less cost of goods sold, reflecting a measure of profitability excluding R&D and operating expenses. NVIDIA doesn't actually break down profitability net of expenses by line of business in their GPU segment. The GPU R&D cost is largely shared between multiple lines of business anyway, so it wouldn't be a very fruitful exercise in the first place.

 

[mrmt]

Gross margins obviously indicate a measure of revenue less cost of goods sold, reflecting a measure of profitability excluding R&D and operating expenses. NVIDIA doesn't actually break down "true" profitability by line of business in their GPU segment. The GPU R&D cost is largely shared between multiple lines of business anyway.

I'm sorry, but what do you mean by "obviously indicate"? Gross margin *is* revenue - COGS. A few things more if you want to be more pedantic, but the basic idea is revenue - COGS. And what do you want to say with "excluding operating expenses and R&D? For IC design companies such as Nvidia, R&D *is* operating expense. And OPEX isn't the only thing separating gross profit from net profits, there's at least SG&A and taxes.

And Nvidia doesn't break down the Tesla segment, but you can get the numbers from historical data and market research reports. And Tesla isn't profitable, regardless of what you believe. And Tesla's case, it isn't only a matter of the R&D of the chip, but the heavy software support (CUDA, drivers, engineering teams supporting customers). Btw, Nvidia used to break down profitability by three lines of business, consumer GPU, professional GPU and chipsets. As the chipset business died and they aren't forecasting big growth rates for the entire GPU market, they put everything under the same umbrella.

 

[sontin]

And Nvidia doesn't break down the Tesla segment, but you can get the numbers from historical data and market research reports. And Tesla isn't profitable, regardless of what you believe.

What are you talking? Tesla uses the same GPU like the Geforce, Quadro and in the future the GRID and VCA business. There is only little R&D addition from a hardware standpoint. The rest is software and support. And believe me, this doesnt cost hundreds of million of dollars.

btw, Nvidia used to break down profitability by three lines of business, consumer GPU, professional GPU and chipsets. As the chipset business died and they aren't forecasting big growth rates for the entire GPU market, they put everything under the same umbrella.

nVidia's GPU business is up 14% from last year. It's huge for them. And it will be growing again this year.

 

[krumme]

Nv can really use its brand value on the b2c side. For low to mid range nv equals gfx. That should give superior total profit to the rest of the portfolio. As said by mrmt profit is not margins. Its damn difficult to interprete results as eg r&d can go everywhere to please shareholders.

My take is consumer dgpu low and midrange is the most profitable and in the future jhh is betting its going the arm way. I think thats the right way to go.

And Denver is the most exiting here in years and i am looking forward to it and hopes the best.

 

[Genx87]

And your point is wrong. AMD always had the very bottom of the x86 market to scrap, so regardless of the success or failure of their top of the line products, they could earn some cash in the bottom. Which product do you think that saved AMD's bacon in the last couple of year, derpdozer FX or derpdozer APU and that huge dies of them, or Brazos and Jaguar, small, cheap chips that had a niche all for themselves in the x86 market since Intel wouldn't go after it?

Nvidia doesn't have that luxury in the mobile market. If they don't beat Qualcomm or Samsung to the punch in performance, there's always Mediateks and Rockchips to undercut them product on price so they don't sell. That's the reason on why their Tegra sales crashed YoY. Nvidia faces a much tougher situation, and a far more dangerous one in the ARM market than AMD on the x86 market.

If you look at AMD financials. They rarely ever turned a profit. Ever! They generated enough cash flow to keep the business alive. But the majority of the qtrs they have lost money. Nvidia can keep plowing money into this project as long as they can afford it. Right now they can clearly afford it. And the potential payoff is more than any short term losses.

Microsoft still sells the xBox though through its lifecycle it has lost 6 billion dollars.
Nvidia right now has the luxury of professional markets to keep funding Tegra.

 

[mrmt]

What are you talking? Tesla uses the same GPU like the Geforce, Quadro and in the future the GRID and VCA business. There is only little R&D addition from a hardware standpoint. The rest is software and support. And believe me, this doesnt cost hundreds of million of dollars

Of course it does, because if it didn't you'd see in the old PSB business operating profits close to gross revenue, and we don't see that. Given that COGS is marginal there, there's a sizable support cost.