Nvidia results

[mrmt]

Just a bit off topic for the forum but I gave a quick look at Nvidia results last quarter. They were quite good, and it seems that Nvidia strategy of small die chips is definitely paying off, as they are able to retain margins when the market is bad and are better positioned when the GPU market warms up.

The consumer GPU markets had a nice quarter growing 15% YoY, reversing the downward trend present last year, and guess what, Nvidia was ready to grab the opportunity and posted very, very good results even if we exclude their Tegra sales, which were reasonable enough.

Tegra now amounts for more than 30% of Nvidia sales, which means that the business should get critical mass to stop bleeding AMD money soon. Their very conservative forecasts for Q4 means that they are finally feeling Krait pressure and that the PC slowdown should impact them soon.

Unless the PC market plunges in Q4 they will fare well until Haswell, which means we'll have a pretty good idea of Nvidia 2013 in Q2 only.

[Idontcare]

Originally Posted by mrmt

Tegra now amounts for more than 30% of Nvidia sales, which means that the business should get critical mass to stop bleeding AMD money soon. Their very conservative forecasts for Q4 means that they are finally feeling Krait pressure and that the PC slowdown should impact them soon.

Unless the PC market plunges in Q4 they will fare well until Haswell, which means we'll have a pretty good idea of Nvidia 2013 in Q2 only.

I had no idea that Nvidia had so successfully diversified themselves. 30% from Tegra, that's crazy. Pretty soon Nvidia will be known as the MPU company and not as the GPU company.

[mrmt]

Originally Posted by Idontcare

I had no idea that Nvidia had so successfully diversified themselves. 30% from Tegra, that's crazy. Pretty soon Nvidia will be known as the MPU company and not as the GPU company.

Once you put it perspective, Nvidia is far smaller than Qualcomm, Samsung or even TI, so whatever gains they are making now must not be seen as definitive, but so far the company is managing the changing environment far better than others like TI or the MIPS ecosystem.

They pushed quad cores ARMv9, they are developing a custom ARM core, they pulled out of the GPGPU chips for consumers on the right time, and they fielded the smallest and most efficient high end GPU chip on the market today.

It is an impressive track record for a company that four years ago was taking losses and fielding dogs like Fermi.

[Idontcare]

xbitlabs Nvidia: Denver Is Custom ARMv8 64-Bit Processor with Secret Sauce.

“Project Denver is a 64-bit processor, it is a full-custom ARMv8 64-bit processor with some exciting secret sauce and we have not revealed much about it. There are no 64 bits processors shipping today, as you know. ARMv8 [cores] was just announced. […] I guess, the only thing I could tell you is Denver is going great,” said Jen-Hsun Huang, chief executive officer of Nvidia, during quarterly conference call with financial analysts.

Any ideas on when Denver is supposed to see the light of day?

[boxleitnerb]

Probably 2014. Could only find a German article at heise.de, sorry:
http://www.heise.de/newsticker/meldu...M-1426462.html

[Sable]

I wouldn't say it's that surprising. Tegra 3 has been a huge success for them. The Nexus 7 win alone must have been HUGE for them.

edit:

Shows a decent market awareness from them though. AMD at the moment have nothing in this space and their main source (CPU) they're getting whipped in.

[tviceman]

Originally Posted by Idontcare

xbitlabs Nvidia: Denver Is Custom ARMv8 64-Bit Processor with Secret Sauce.
Any ideas on when Denver is supposed to see the light of day?

It's being designed for 20nm at TSMC, although there have been reports that they have taped out tegra designs with Samsung's fabs too.

[Sable]

So we're gonna have AMD and Nvidia competing in the ARM 64 space. Is it interesting that ARM 64 had an announcement with AMD's ARM 64? Although did I read correctly that AMD are using a standard ARM design? Are they producing their own custom graphics like NV or is it pure ARM with sever aspirations and a completely different market to the mobile concern that NV is concentrating on?

[Intel17]

Love how the stock sold off after great results + huge cash balance + dividend.

Added more today.

[boxleitnerb]

It's ironic that AMD follows Nvidias path, but always later. HPC and GPGPU, then ARM 64bit. To jump into a market that is so competitive and then to do it quite late was a dumb move in my opinion.

[lambchops511]

One of the difference between NVDA and AMD is layoffs. NVDA never really laid off any engineers vs. AMD laid off tons of engineers during the recession. Technology is a cut throat business, R&D or die.

R&D doesn't necessary mean profits, see NVDA chipset business where they dumped lots of resources and capital and it fails with zero revenue (on Intel platforms). Or see Intel Larabee where they dedicate literally 50% of the company towards and it still hasn't produced a dollar revenue after 5 years. But once in a while, you will have a successful product line (see Tegra / CUDA). The most valuable CEOs are the ones with vision to pick out the winners (see Steve Jobs); obviously luck and timing has a lot to do with it as well (see Steve Jobs before returning to AAPL).

I don't like Jensen, but you have to admit, his ability to guide to the company towards Tegra and CUDA years before anybody even thought to enter them.

[Intel17]

Originally Posted by lambchops511

One of the difference between NVDA and AMD is layoffs. NVDA never really laid off any engineers vs. AMD laid off tons of engineers during the recession. Technology is a cut throat business, R&D or die.

R&D doesn't necessary mean profits, see NVDA chipset business where they dumped lots of resources and capital and it fails with zero revenue (on Intel platforms). Or see Intel Larabee where they dedicate literally 50% of the company towards and it still hasn't produced a dollar revenue after 5 years. But once in a while, you will have a successful product line (see Tegra / CUDA). The most valuable CEOs are the ones with vision to pick out the winners (see Steve Jobs); obviously luck and timing has a lot to do with it as well (see Steve Jobs before returning to AAPL).

I don't like Jensen, but you have to admit, his ability to guide to the company towards Tegra and CUDA years before anybody even thought to enter them.

50% of the company did not work on Larrabee. Not even close.

[Idontcare]

Originally Posted by lambchops511

One of the difference between NVDA and AMD is layoffs. NVDA never really laid off any engineers vs. AMD laid off tons of engineers during the recession. Technology is a cut throat business, R&D or die.

R&D doesn't necessary mean profits, see NVDA chipset business where they dumped lots of resources and capital and it fails with zero revenue (on Intel platforms). Or see Intel Larabee where they dedicate literally 50% of the company towards and it still hasn't produced a dollar revenue after 5 years. But once in a while, you will have a successful product line (see Tegra / CUDA). The most valuable CEOs are the ones with vision to pick out the winners (see Steve Jobs); obviously luck and timing has a lot to do with it as well (see Steve Jobs before returning to AAPL).

I don't like Jensen, but you have to admit, his ability to guide to the company towards Tegra and CUDA years before anybody even thought to enter them.

When it comes to R&D, the adage is: "You don't always get what you pay for, but you never get what you don't pay for."

[mrmt]

Originally Posted by Intel17

50% of the company did not work on Larrabee. Not even close.

That 4 billion/50% of the company statement smells Theo Valich, the clueless plagiarist.

[ShintaiDK]

I think I read somewhere that the entire development cost for Larabee, MIC and the 80 core prototypes was 1 billion$. This also includes technologies that is/can be used in other products. So thats maybe 200mio$ a year on a what, 8billion$+ R&D budget? Far from 50%...

Also that 1 billion$ should pay itself back sooner or later in the HPC segment.

A product like Haswell for example most likely sits on 1½billion$ a year.

[sontin]

Originally Posted by lambchops511

R&D doesn't necessary mean profits, see NVDA chipset business where they dumped lots of resources and capital and it fails with zero revenue (on Intel platforms).

Woot? After they got the Apple deal and put their chipset into the netbook market (ION) it was a $1 Billion per year division of nVidia.

[Cerb]

Originally Posted by mrmt

It is an impressive track record for a company that four years ago was taking losses and fielding dogs like Fermi.

There was one bad Fermi chip, the very first one, GF100. All the rest were great. They're still selling them, as premium parts, even.

Originally Posted by Sable

So we're gonna have AMD and Nvidia competing in the ARM 64 space. Is it interesting that ARM 64 had an announcement with AMD's ARM 64? Although did I read correctly that AMD are using a standard ARM design? Are they producing their own custom graphics like NV or is it pure ARM with sever aspirations and a completely different market to the mobile concern that NV is concentrating on?

I think it's going to be more like NV, Samsung, and Apple. AMD may be there, but I would be quite surprised if they'll be competing much against JHH's company. I think AMD is just reaching for something to do, honestly, having missed so many past opportunities.

Originally Posted by boxleitnerb

It's ironic that AMD follows Nvidias path, but always later. HPC and GPGPU, then ARM 64bit. To jump into a market that is so competitive and then to do it quite late was a dumb move in my opinion.

AMD never invested in it. They acted like they were going to, and then NV actually did.

[Idontcare]

Originally Posted by ShintaiDK

I think I read somewhere that the entire development cost for Larabee, MIC and the 80 core prototypes was 1 billion$. This also includes technologies that is/can be used in other products. So thats maybe 200mio$ a year on a what, 8billion$+ R&D budget? Far from 50%...

Also that 1 billion$ should pay itself back sooner or later in the HPC segment.

A product like Haswell for example most likely sits on 1½billion$ a year.

I read $1B as well. Not to trivialize the magnitude of that investment, but it was far less expensive than say spending 5x as much to acquire a discrete graphics company.

[Intel17]

Originally Posted by Idontcare

I read $1B as well. Not to trivialize the magnitude of that investment, but it was far less expensive than say spending 5x as much to acquire a discrete graphics company.

LOL

[mrmt]

Originally Posted by Cerb

There was one bad Fermi chip, the very first one, GF100. All the rest were great. They're still selling them, as premium parts, even

I think it is clear by now NV has a new direction for the company, and that direction is mobile but it wasn't like that before. back in the time, by the time Fermi was being developed,consumer graphics was falling and JHH was pushing GPGPU as the future. Those expectations didn't pan out and today Nvidia PSB is more or less where they are since that time: Quadro is their principal PSB business and Tesla is just a footnote in their SEC fillings.

In order to push GPGPU they had to make GF1x0 which were very big chips, to the point they pushed the entire consumer business into losses in 2010 and reduced profits in 2011 without correspondent increase in Tesla revenues.

I said that Fermi was a dog not because their manufacturing woes, but because the design trade offs Nvidia did to give Fermi a lot of compute power didn't pay off. It was a wrong strategic direction, one that was quickly corrected with Kepler and Tegra.

[blastingcap]

Don't forget cloud computing. NV sees cloud gaming as a legitimate competitor to consoles, and maybe even PC gaming. NV is shut out of next-gen consoles which are reportedly all using AMD APUs, so cloud gaming is of particular importance to them. NV will continue taking their share of PC gaming market as usual.

[Fox5]

Originally Posted by mrmt

I think it is clear by now NV has a new direction for the company, and that direction is mobile but it wasn't like that before. back in the time, by the time Fermi was being developed,consumer graphics was falling and JHH was pushing GPGPU as the future. Those expectations didn't pan out and today Nvidia PSB is more or less where they are since that time: Quadro is their principal PSB business and Tesla is just a footnote in their SEC fillings.

In order to push GPGPU they had to make GF1x0 which were very big chips, to the point they pushed the entire consumer business into losses in 2010 and reduced profits in 2011 without correspondent increase in Tesla revenues.

I said that Fermi was a dog not because their manufacturing woes, but because the design trade offs Nvidia did to give Fermi a lot of compute power didn't pay off. It was a wrong strategic direction, one that was quickly corrected with Kepler and Tegra.

The thing is, AMD provided a ton of compute power in their cards without making as many sacrifices as Fermi. Granted, it's still bloated compared to a gaming focused card like Kepler, but if AMD had a decent GPGPU market, the bloating would be acceptable. (after all, their cpus are bloated as hell too, a move I've never understood when they could just flood the low end market with tiny dual core chips that provided good enough performance)

[mrmt]

Originally Posted by Fox5

The thing is, AMD provided a ton of compute power in their cards without making as many sacrifices as Fermi. Granted, it's still bloated compared to a gaming focused card like Kepler, but if AMD had a decent GPGPU market, the bloating would be acceptable. (after all, their cpus are bloated as hell too, a move I've never understood when they could just flood the low end market with tiny dual core chips that provided good enough performance)

Except for the HPC where GPGPU provides a sizable advantage, the GPGPU advantage when it appears isn't compelling enough to make software companies optimize for it. Worse, in some cases, code that show improvements running on AMD/Nvidia GPUs is simply slower than running conventional code on Intel CPUs. GPGPU didn't take off for the mass market, and that's why Nvidia diverted most of their resources from GPGPU. So, it is fair to say that nobody has a GPGPU market.

GNC takes GPGPU more seriously than Fermi. It is almost a co-processor, one that would fit nicely side by side with a CPU if the logic structure and software support were in place. Problem is that there is no software support for that hybrid approach and AMD is not strong enough to provide it, so in a sense Fermi and GNC are just two of the same bread, chips that ballooned die size trying to chase a market that simply isn't there.

[Cerb]

Originally Posted by mrmt

I said that Fermi was a dog not because their manufacturing woes, but because the design trade offs Nvidia did to give Fermi a lot of compute power didn't pay off. It was a wrong strategic direction, one that was quickly corrected with Kepler and Tegra.

That I can agree with. I took Fermi being a dog to refer its performance, which was actually pretty good.

[pablo87]

Somewhere in those Nvidia results and balance sheet there's a lesson for AMD's Board of Directors and CEO's (past and present)...