Qualcomm Adreno 530 matches Tegra X1?!?

[BoFox]

Wow, the competition remains fierce!!  Just when Nvidia thought that they were so far ahead with their Tegra X1, Qualcomm manages a comeback. 

http://wccftech.com/snapdragon-820-benchmarks/

Too bad Nvidia couldn't have had their Tegra X1 in more tablets for several months since it debuted on their SHIELD Android TV, with the recently released Google Pixel C the only tablet on the market with such chip.  Could it have been production issues? 

Nvidia needs to try to go big once again like they did with Tegra 2 and to a smaller extent, Tegra 3.  They might have the tech, but delivering it on more than just one device these days seem to be a problem (Tegra 4 being on the first SHIELD months before being available for EVGA Tegra Note 7, then Tegra K1 only being available on the Shield Tablet for many months until custom Denver Tegra K1 were featured in Nexus 9.  What gives?  Did most companies sign a hidden contract with Qualcomm to stay away from Nvidia, or are these tablet companies avoiding Nvidia for another reason (being burned by delays that broke the deadlines one too many times)?

This isn't the Nvidia that I know.  In the PC industry, Nvidia rose as the world's best GPU company due to its market dominance that required massive product availability.  Nvidia didn't go the 3dfx route and ditch all other AiB's, choosing to market only its own brand alone (or have just one AiB like EVGA make cards along with Nvidia).  What gives?

[SickBeast]

The new node is going to allow them to make some nice gains. This year might be a pretty good year to pick up a new phone actually. After that I expect things to stagnate and God knows how long it will be until they can move to an even smaller node.

[BenSkywalker]

Just when Nvidia thought that they were so far ahead with their Tegra X1, Qualcomm manages a comeback.

Couple quick points.

First- The X1 was a 1H '15 20nm part, the 820 is 1H '16 14nm part.

Second - The X1 is still faster(not much, but it *is* faster).

So, with an additional year and a half node shrink Qualcomm was able to produce a part that didn't fail as badly- I'm not feeling that as being quite a 'win' from a performance perspective.

Too bad Nvidia couldn't have had their Tegra X1 in more tablets for several months since it debuted on their SHIELD Android TV, with the recently released Google Pixel C the only tablet on the market with such chip.  Could it have been production issues?

It was a cost issue. nVidia hasn't hidden the fact that they aren't willing to get into a low margin commodity market. They get their margins, or they go find someone else to buy them. Like the automotive industry.

During the second quarter, NVIDIA generated $71 million of revenue from its automotive business, a 76% year-over-year increase. NVIDIA's Tegra chips are mainly used today to power in-car displays, and according to the company's investor day presentation, 8 million cars currently on the road contain NVIDIA chips. Some of those cars contain more than one Tegra chip, with the Audi Q7 sporting as many as four in order to power all of its displays.

Twenty-five million additional cars containing NVIDIA Tegra chips are in the pipeline, but because the time-to-market for cars is typically measured in years, NVIDIA's automotive design wins take quite a bit of time to actually generate revenue. The silver lining is that it makes NVIDIA's automotive revenue more predictable, and it's safe to say that strong automotive growth is likely to continue.

http://www.fool.com/investing/general/20...table.aspx

I still see them pushing the Android front but mainly as a chicken and the egg answer- which they are starting to see some fruit from. If Android becomes a major 'serious' gaming platform, nVidia is pretty much lapping all the other players in terms of hardware/support/development tools etc. By remaining the 'halo' part in this segment, market shifts would allow them to move quite quickly. Meanwhile, the auto industry can give them the huge margins they want to see. With how small the chips are, $25 per is considered absurdly expensive in the consumer electronics space(I would *GLADLY* pay more then that in addition to the base cost for a premium Parker based Tegra with OLED tablet but I digress), anyone who has ever had to replace an ECU knows that the auto industry considers that peanuts(and due to nVidia using the same base part for lots of different cars, their R&D overhead is by comparison tiny).

 Nvidia didn't go the 3dfx route and ditch all other AiB's, choosing to market only its own brand alone (or have just one AiB like EVGA make cards along with Nvidia).  What gives?

Companies don't want to pay. I was seriously upset when I saw that Google was going with shitty ass Wal-Mart/Apple level displays for the Pixel C. If it had a reasonable display in it, I would have one right now. That is, to me, the big problem. Nobody is making a truly high end Android tablet or phone that uses Tegra SoCs(well, the Pixel C outside of its piss poor display would be). But who is asking for them? I am. I have a Shield. I have a Shield TV. nVidia makes a Parker/Tegra shield with an OLED screen I'm all over it. Outside of me, I don't know anyone else who is really asking for it.

In the old early days of PC gaming, *EVERYONE* was asking for it- now, not so much.

After that I expect things to stagnate and God knows how long it will be until they can move to an even smaller node.

Just as a heads up, while GPUs have been stuck at 28nm for a long time we have already seen 20nm, 16nm and 14nm SoCs- the demands for fabbing SoCs is much, much lower then for monolithic GPUs so smaller nodes are going to continue to happen quicker in that segment(simply a function of a *MUCH* smaller die- 18 defects can destroy an entire batch of desktop GPUs, or 5% of a batch of SoCs).

[SickBeast]

Ben the nVidia Tegra chips don't have integrated modems and their power consumption has been high. As a result no one is using them in phones. I have only ever seen them in tablets. This severely limits the market for nVidia. I really don't know why they are doing this. Their mobile APUs are top notch. All they need is the modem. If they dial back the power consumption a bit they would probably have their processors in a whole bunch of phones and they would be making billions of dollars.

Usually I find that nVidia is very well run but in this particular area I really think they could be doing better. I find it hard to believe that they can engineer such complex GPUs but then they are unable to integrate something as simple as a modem.

[BenSkywalker]

Ben the nVidia Tegra chips don't have integrated modems and their power consumption has been high.

Margins aren't in the league of being where nVidia wants them in the cell phone market, Qualcomm was under 26% the latest numbers I found in a quick search- nVidia's are at 56%. The modem is a non issue because they don't want to play in that segment. As far as power consumption- you have any links with identical workloads losing to anyone on power consumption?

I really don't know why they are doing this.

Margins. No reason to go after the cell phone market with margins that low. They don't want to be a low price commodity company. 

Their mobile APUs are top notch.

SoC, there is actually a pretty huge technical gap between the two

If they dial back the power consumption a bit they would probably have their processors in a whole bunch of phones and they would be making billions of dollars.

Neither Apple nor Samsung have an integrated modem that can handle all of the various basebands, they are using external chips from Qualcomm. So we have Qualcomm owning the lions share of the SoC market, and also selling modems to everybody and they have annual revenue of around $25B which gives them about $6B in margin. Not a bad number by any stretch of the imagination. Then look at nVidia who is in the $5B in sales range and the ~$2.2B in margin range. So nVidia could use your approach, try to increase Tegra volume by a couple orders of magnitude and reach comparable margins, or take the approach they currently have and shoot for more modest volume gains while simultaneously becoming *THE* dominant player in truly smart cars while remaining the halo part in the mobile sector.

I'm not saying one is the right way to do it, just pointing out that both paths have potentially lucrative outcomes, just one is more in line with nV's long standing goals while the other is quite a different direction.

I find it hard to believe that they can engineer such complex GPUs but then they are unable to integrate something as simple as a modem.

They already had it done, and shut the whole thing down because the margins weren't there. This isn't me speculating, Icera is quite public information.

[BoFox]

Just when Nvidia thought that they were so far ahead with their Tegra X1, Qualcomm manages a comeback.

Couple quick points.

First- The X1 was a 1H '15 20nm part, the 820 is 1H '16 14nm part.

Second - The X1 is still faster(not much, but it *is* faster).

So, with an additional year and a half node shrink Qualcomm was able to produce a part that didn't fail as badly- I'm not feeling that as being quite a 'win' from a performance perspective.

Too bad Nvidia couldn't have had their Tegra X1 in more tablets for several months since it debuted on their SHIELD Android TV, with the recently released Google Pixel C the only tablet on the market with such chip.  Could it have been production issues?

It was a cost issue. nVidia hasn't hidden the fact that they aren't willing to get into a low margin commodity market. They get their margins, or they go find someone else to buy them. Like the automotive industry.

During the second quarter, NVIDIA generated $71 million of revenue from its automotive business, a 76% year-over-year increase. NVIDIA's Tegra chips are mainly used today to power in-car displays, and according to the company's investor day presentation, 8 million cars currently on the road contain NVIDIA chips. Some of those cars contain more than one Tegra chip, with the Audi Q7 sporting as many as four in order to power all of its displays.

Twenty-five million additional cars containing NVIDIA Tegra chips are in the pipeline, but because the time-to-market for cars is typically measured in years, NVIDIA's automotive design wins take quite a bit of time to actually generate revenue. The silver lining is that it makes NVIDIA's automotive revenue more predictable, and it's safe to say that strong automotive growth is likely to continue.

http://www.fool.com/investing/general/20...table.aspx

I still see them pushing the Android front but mainly as a chicken and the egg answer- which they are starting to see some fruit from. If Android becomes a major 'serious' gaming platform, nVidia is pretty much lapping all the other players in terms of hardware/support/development tools etc. By remaining the 'halo' part in this segment, market shifts would allow them to move quite quickly. Meanwhile, the auto industry can give them the huge margins they want to see. With how small the chips are, $25 per is considered absurdly expensive in the consumer electronics space(I would *GLADLY* pay more then that in addition to the base cost for a premium Parker based Tegra with OLED tablet but I digress), anyone who has ever had to replace an ECU knows that the auto industry considers that peanuts(and due to nVidia using the same base part for lots of different cars, their R&D overhead is by comparison tiny).

 Nvidia didn't go the 3dfx route and ditch all other AiB's, choosing to market only its own brand alone (or have just one AiB like EVGA make cards along with Nvidia).  What gives?

Companies don't want to pay. I was seriously upset when I saw that Google was going with shitty ass Wal-Mart/Apple level displays for the Pixel C. If it had a reasonable display in it, I would have one right now. That is, to me, the big problem. Nobody is making a truly high end Android tablet or phone that uses Tegra SoCs(well, the Pixel C outside of its piss poor display would be). But who is asking for them? I am. I have a Shield. I have a Shield TV. nVidia makes a Parker/Tegra shield with an OLED screen I'm all over it. Outside of me, I don't know anyone else who is really asking for it.

In the old early days of PC gaming, *EVERYONE* was asking for it- now, not so much.

Thank you for this awesome explanation!  This is what I was suspecting, and you pretty much confirmed it all the way for me. 

So, do you have lots of Nvidia stock?  Like tens of thousands invested into NV stock? 

Ya know, perhaps it is possible for NV to get LG to manufacture OLED panels for the next Shield Tablet design?  Outside of you, there are many others who would really like exactly what you would like. 

Yet, the best and biggest thing for Nvidia is to just make a 4K gaming console with upgradeable Pascal modules (along with 16 custom Denver cores). (EDIT - or 8 next-gen Denver cores, plus 8 A72 cores, or even an Intel CPU, if Nvidia can force a deal).

[BenSkywalker]

So, do you have lots of Nvidia stock?  Like tens of thousands invested into NV stock?

Haha, not a single share, well- if any of the funds I have invested in hold nVidia I am not aware of it, but I don't study the particular holdings that much as long as the returns are decent. I was at one point invited to the nVidia focus group and turned that down, my opinion is my own, always. 

Yet, the best and biggest thing for Nvidia is to just make a 4K gaming console with upgradeable Pascal modules

Well, they already have a 4K gaming console(yes, it does actually play some games at 4K too)- just not quite upgrade able Pascal level. Hopefully they can make something like what I am looking for, I have the money- make it and the money is yours nV