• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA to License Kepler and Future GPU IP to 3rd Parties

For example, Intel could license a Kepler design to incorporate as their IGP. (Not that I think will happen)

It stands to reason that Nvidia will have to adapt, maybe this is them trying?

I'm literally guessing though.
 
Martini1991 said:
For example, Intel could license a Kepler design to incorporate as their IGP. (Not that I think will happen)

It stands to reason that Nvidia will have to adapt, maybe this is them trying?

I'm literally guessing though.
Click to expand...

Weren't there rumours about an Nvidia/Intel merge? Something about it falling through because Jen Hsun-Huang wanted to be CEO of both? This could somehow be an attempt to hit AMD with an IGP with Intel... or something. Obviously I don't know anything either.
 
humbug said:
Then nothing has changed, Developers are still locked to Nvidia in using CUDA.

Its the bit that I don't understand, allowing anyone to use your GPU technology will mean anyone can compete with you on a level footing.
That would hurt Nvidia more than it would Intel and AMD given that they are nowhere near as dependant on GPU sales as Nvidia are.
Click to expand...

when you licence something to someone, you specifically write the contract so that either they can't compete directly with you, or you charge them so much that you don't care as you are making the same level of profit as if you had made the sales yourself (but with none of the production risks and costs)

kepler cores could be of use to anyone who currently makes a chip that has an integrated GPU of some description who wants it to be bigger and faster without developing a whole new graphics chip themselves from the ground up, you're thinking too literally
 
PorkPieCat said:
Weren't there rumours about an Nvidia/Intel merge? Something about it falling through because Jen Hsun-Huang wanted to be CEO of both? This could somehow be an attempt to hit AMD with an IGP with Intel... or something. Obviously I don't know anything either.
Click to expand...

that was AMD buying Nvidia - which they looked at before buying ATI for a massive amount of money and funnily enough not actually turning a profit since
 
This is also for you, andybird123.

Martini1991 said:
For example, Intel could license a Kepler design to incorporate as their IGP. (Not that I think will happen)

It stands to reason that Nvidia will have to adapt, maybe this is them trying?

I'm literally guessing though.
Click to expand...

Right :) Intel are pretty confident in their own tech.

Nvidia are experiencing a double pronged attack from both Intel and AMD in this space.

Nvidia sell a lot of low and mid level discrete GPU's to OEM's, with AMD and now Intel making Integrated GFX that are just as capable as Nvidia's low to mid level OEM GPU's, those OEM will chose the cheaper more power efficient, more convenient integrated option over Nvidia's OEM GPU's.

Intel nor AMD will take up Nvidia's tech to integrate into their CPU's, whats more anyone wanting to make x86_64 capable CPU + GPU's in that space will also need the blessing of Intel and AMD who own all the needed IP.

That leaves those already licencing ARM CPU's to then also licence Nvidia GPU tech to integrate and make their own 'APU's'

Intel and AMD have very capable solutions in the ARM space with their own iGPU's, and Nvidia already licence ARM and integrate their GPU's to make Tegra. a Samsung licenced ARM chip with an Nvidia iGPU is just another competitor for AMD, Intel and Nvidia.

Perhaps Samsung might want to make their own version of a GTX 680?
Whose that going to hurt? Nvidia more than anyone.

Perhaps Nvidia think licencing their GPU tech along with CUDA will flood the market with CUDA enabled products, there in negating OpenCL's momentum.

Scenario A: Intel and AMD also take up CUDA and simply integrate it into their existing tech.
The GPGPU market is open, more players can enter, we all benefit and none of the big 3 have any advantage over the other.

Scenario B: Nvidia refuse licencing to Intel and AMD, instead licence to others CUDA and GPU tech to flood the market with CUDA only enabled products on a very narrow part of the market where they don't need Intel - AMD's blessing.
This might hurt Intel and AMD in the growing mobile GPGPU space, a tiny bit.
I think it far more likely developers by enlarge will not back it as the other two big players with the majority of combined market share are incompatible.
No developer wants to sell a product that does not work for the majority of its user base.
 
Last edited:
I might be wrong but for me this sounds a little desperate from nvidia or you could say forced but most likely the right move for them. At the minute Nvidia are doing pretty well but i get the feeling things like hsa could really spoil there day in the pc market and going forward the professional market where they own last time i checked around 80% share. Apu's are only going to get better as we can see from how powerful the ps4 chipset is. I don't know to much about about the mobile market but from a few threads around here it does not sound like tegra is the leading chip there. Nvidia lacks in not having a x86 license which now seems to be there biggest weekness and in the pc market i think it could force them out at some point.

The one thing Nvidia are not is a stupid company though and i bet they got some good plans to stay in the game. I just think this is one step that they did not want to take but have had to for some of the reasons above.
 
My instant and only reaction is, they've basically lost all direct customers for their SOC line up and that they aren't even a tiny bit confident about Denver. We're talking about a company that has sunk billions into developing their Tegra line, where the company has talked about NOTHING but their mobile line growth, how great they are and how the future is mobile and how Tegra will kick ass and be worth a HUGE portion of their yearly revenue in the future.

AMD sell 10-20mil cpu's a year or so, Arm licence IP used in 100's of millions a year, billions maybe, its nuts.... AMD has a revenue of 5-6billion, ARM have a revenue of 500million or so. IP revenue for SOC's is TINY per chip, absolutely stonkingly tiny. The profit from selling actual chips is a magnitude higher than IP, and the devices themselves are a magnitude higher again. Essentially no one volentarily moves from a "higher" level of profit per device to the lower level..... unless pretty much forced to.

IF Nvidia had the best GPU's its a reason for people to buy their entire exclusive SOC with a pretty decent profit per SOC to get that gpu performance, when you sell the IP and allow potentially the best GPU performance to be used in any SOC, leaving no one a reason to buy your SOC's.... you are shooting yourself in the foot unless you have just about zero faith in your SOC's selling in the first place.

Tegra 3 had one major customer, mostly due to having their quad core 40nm chip out early rather than going 28nm like everyone else, they promptly lost this despite the massive sales...... many rumours why, again that people end up angry at Nvidia's promises for power targets, how they act as a partner, delays for their 28nm parts. Either way their easily highest volume Tegra 3 client moved on and dropped Tegra 4, it got delayed, it supposedly missed power targets again(like EVERY Tegra). They promised a lot with Tegra 1, delivered almost nothing, angered people, but paid/convinced people it was a one off. Got more design wins for Tegra 2, again missed targets, again angered customers. Tegra 3 got one major design win due mostly to very short sighted thinking, and then got dropped by that same client for the replacement device(due to the short sighted thinking) and now Tegra 4 is late, over power, late again with no seeming advantage either. People stopped buying the "its not our fault, the next one will be great" excuse.

Nvidia bet their entire future almost on the growth of Tegra sales, and today they announced a reason for no one to ever need to buy a Tegra again... but that they MIGHT be able to sell the gpu IP for a fraction of the profit per device.

It really doesn't come across as good news, it comes across as giving up the profitable high volume SOC sales market for the very low profit, IP market, which hands all the profit over to those making the chips.

The only way this could work out would be some insane deal with Intel, Intel use the IP for IGP's, and as part of the deal they let Nvidia have some foundry capacity for their gpu's..... but I seriously doubt they would announce that as a IP licence rather than a direct partnership with Intel, making me think the chances of them working with Intel are non existant.
 
drunkenmaster said:
My instant and only reaction is, they've basically lost all direct customers for their SOC line up and that they aren't even a tiny bit confident about Denver. We're talking about a company that has sunk billions into developing their Tegra line, where the company has talked about NOTHING but their mobile line growth, how great they are and how the future is mobile and how Tegra will kick ass and be worth a HUGE portion of their yearly revenue in the future.

AMD sell 10-20mil cpu's a year or so, Arm licence IP used in 100's of millions a year, billions maybe, its nuts.... AMD has a revenue of 5-6billion, ARM have a revenue of 500million or so. IP revenue for SOC's is TINY per chip, absolutely stonkingly tiny. The profit from selling actual chips is a magnitude higher than IP, and the devices themselves are a magnitude higher again. Essentially no one volentarily moves from a "higher" level of profit per device to the lower level..... unless pretty much forced to.

IF Nvidia had the best GPU's its a reason for people to buy their entire exclusive SOC with a pretty decent profit per SOC to get that gpu performance, when you sell the IP and allow potentially the best GPU performance to be used in any SOC, leaving no one a reason to buy your SOC's.... you are shooting yourself in the foot unless you have just about zero faith in your SOC's selling in the first place.

Tegra 3 had one major customer, mostly due to having their quad core 40nm chip out early rather than going 28nm like everyone else, they promptly lost this despite the massive sales...... many rumours why, again that people end up angry at Nvidia's promises for power targets, how they act as a partner, delays for their 28nm parts. Either way their easily highest volume Tegra 3 client moved on and dropped Tegra 4, it got delayed, it supposedly missed power targets again(like EVERY Tegra). They promised a lot with Tegra 1, delivered almost nothing, angered people, but paid/convinced people it was a one off. Got more design wins for Tegra 2, again missed targets, again angered customers. Tegra 3 got one major design win due mostly to very short sighted thinking, and then got dropped by that same client for the replacement device(due to the short sighted thinking) and now Tegra 4 is late, over power, late again with no seeming advantage either. People stopped buying the "its not our fault, the next one will be great" excuse.

Nvidia bet their entire future almost on the growth of Tegra sales, and today they announced a reason for no one to ever need to buy a Tegra again... but that they MIGHT be able to sell the gpu IP for a fraction of the profit per device.

It really doesn't come across as good news, it comes across as giving up the profitable high volume SOC sales market for the very low profit, IP market, which hands all the profit over to those making the chips.

The only way this could work out would be some insane deal with Intel, Intel use the IP for IGP's, and as part of the deal they let Nvidia have some foundry capacity for their gpu's..... but I seriously doubt they would announce that as a IP licence rather than a direct partnership with Intel, making me think the chances of them working with Intel are non existant.
Click to expand...

This ^^^^ it doesn't make any sense.
Nvidia are about to loose $100m a year revenue from PS3 IP sales, that's going off 2012 figures of $115m.

Its a lot of money but only a shadow of what their revenues are from Tegra, the PS3 is just IP at $10 a pop, while Tegra is a SoC chip bringing in $30 to $60 a pop..... oh wait a minute?!?!?!?! not anymore....
 
Think of it this way, Nvidia supposedly was aiming for around $1billion in revenue over 10 years for the Ps3 deal, AMD is banking on pretty much 1 billion a year for their console deals(its unclear if thats X1 and PS4, or just PS4, revenue is predicted to be $80mil in Q3 and 250mil in Q4 from the console deals, but these deals are weird, IBM made the first run of chips for the PS3, then GloFo took over. It might be that IBM were making the first batches of wafers but production will switch to glofo through AMD for the X1... who knows).

1 billion a year, minimum pretty much for AMD, 1billion over the entire life of a console for Nvidia. No one, I mean NO ONE gives up the former for the latter... unless they think the latter is screwed royally. Its massively massively more profitable to make the SOC yourself and sell it than sell the IP, always has been, always will be. Everyone in the industry wants to move the other way, IP guys would kill for the chance to become manufacturers and up revenue/income, guys making chips are killing for the chance to become device makers because that again massively increases the income.

Think of the numbers along the lines of ARM get say 10p per chip made at Samsung for Apple chips made, Samsung get $10 profit per chip they sell to Apple(and its just manufacturing and Apple's design, if Apple were buying samsungs design it would increase the profit significantly) and Apple sell a phone with $150 tech in it for $400. Arm have a revenue of 500mil, Samsung have a revenue of say $5billion(from Apple alone) and Apple have a revenue in the $100billion +.

You move UP the chain when things are going well, you move down the chain when things are going badly... no one wants to move from manufacturing to IP sales and you don't start selling a "market leading design" that if it was so good, would make people buy your entire SOC to get it.
 
drunkenmaster said:
Think of it this way, Nvidia supposedly was aiming for around $1billion in revenue over 10 years for the Ps3 deal, AMD is banking on pretty much 1 billion a year for their console deals(its unclear if thats X1 and PS4, or just PS4, revenue is predicted to be $80mil in Q3 and 250mil in Q4 from the console deals, but these deals are weird, IBM made the first run of chips for the PS3, then GloFo took over. It might be that IBM were making the first batches of wafers but production will switch to glofo through AMD for the X1... who knows).

1 billion a year, minimum pretty much for AMD, 1billion over the entire life of a console for Nvidia. No one, I mean NO ONE gives up the former for the latter... unless they think the latter is screwed royally. Its massively massively more profitable to make the SOC yourself and sell it than sell the IP, always has been, always will be. Everyone in the industry wants to move the other way, IP guys would kill for the chance to become manufacturers and up revenue/income, guys making chips are killing for the chance to become device makers because that again massively increases the income.

Think of the numbers along the lines of ARM get say 10p per chip made at Samsung for Apple chips made, Samsung get $10 profit per chip they sell to Apple(and its just manufacturing and Apple's design, if Apple were buying samsungs design it would increase the profit significantly) and Apple sell a phone with $150 tech in it for $400. Arm have a revenue of 500mil, Samsung have a revenue of say $5billion(from Apple alone) and Apple have a revenue in the $100billion +.

You move UP the chain when things are going well, you move down the chain when things are going badly... no one wants to move from manufacturing to IP sales and you don't start selling a "market leading design" that if it was so good, would make people buy your entire SOC to get it.
Click to expand...

Perhaps with Intel and AMD now making decent SoC mobile chips, cite Silvermont and Temash, and 'if' under cutting Nvidia to get a foot-hold in that space, Nvidia are no longer able to sell Tegra at a rate that brings in meaningful profit, so they are moving to IP.

Nvidia have been able to sell Tegra at a much higher price than other ARM derived chips on the promise that its much better (Samsung ARM + Power VR) with AMD and especially Intel pushing ahead of ARM in performance and potentially what your getting for your watt, Nvidia are finding themselves on the loosing side and high expense.

If that's whats going on you have to feel for Nvidia being squeezed out like that, doesn't seem fair given that they inevitably will get squeezed from the other side with very up-to-date and powerful new Game consoles.
Some people think AMD have problems, some even feel sorry for them, I think they will be just fine, they have a lot of pies they can stick their fingers into, Nvidia are much more constrained in that regard.
 
Last edited:
Surely it is more profitable to sell product A and license the IP rather than just sell product A, just because they are licensing the IP it doesn't mean they will no longer make the product themselves.

Companies who want the IP will license it an companies who want the product will still want the product, yes of course there will be cross overs, some companies who might have originally wanted the product will now only want the IP, but overall I'm sure Nvidia knows what it is doing.
 
bru said:
Surely it is more profitable to sell product A and license the IP rather than just sell product A, just because they are licensing the IP it doesn't mean they will no longer make the product themselves.

Companies who want the IP will license it an companies who want the product will still want the product, yes of course there will be cross overs, some companies who might have originally wanted the product will now only want the IP, but overall I'm sure Nvidia knows what it is doing.
Click to expand...

This is clearly why as such, no one does this. If the GPU was SO killer that people would even put up with Nvidia's SOC problems to get it... why let other people make a better version and screw yourself out of money. It basically doesn't happen and makes little to no sense.

In most cases you also have the situation that if your gpu is so mindblowing that everyone needed it... people would buy your chip anyway, and very few companies have bothered yet. If they refuse to buy the chips because you're always late, over power and not as good as you say, then, well this gives you one option of licencing the IP and praying people put your gpu in their soc to get at least some money.

If there was more total money to be made doing so... Apple, AMD, Intel, Samsung, Qualcomm would all do so, not a one of them does.

This also ignores that Nvidia's mobile gpu for this gen(28nm) is looking like its going to be 50% slower than the fastest GPU this gen for mobile ARM parts. Its also likely cheaper, supports openGL 5.0(Nvidia only just getting on to 4.3 I think it said). Its no where near the market leader in mobile, no where near and despite being a gpu manufacturer gone ARM soc maker, they've failed to dominate a single generation of other products with their GPU's, and they are putting Kepler, with I assume DX11 compatibility, into products where that die space/transistors/power simply hurt them rather than help, leaving them miles behind the competition.

You would ONLY licence this IP if you assume the soc won't sell, and their IP won't sell because its no where near the best mobile gpu IP. We're talking about expecting say Samsung, Intel(are they still using third party GPU in Atom?) or any of the other guys to tear up existing contracts with BETTER mobile GPU makers... to help out Nvidia.... who have screwed them over on device after device since Tegra started.... its a complete non starter. Maybe a gen, or two, of paid for chip designs and device wins which once again never materialise.
 
bru said:
Surely it is more profitable to sell product A and license the IP rather than just sell product A, just because they are licensing the IP it doesn't mean they will no longer make the product themselves.

Companies who want the IP will license it an companies who want the product will still want the product, yes of course there will be cross overs, some companies who might have originally wanted the product will now only want the IP, but overall I'm sure Nvidia knows what it is doing.
Click to expand...

The problem arises when a company who licensed your IP builds a competing product. Each sale made by the competing product is a lost sale for you, even though you make some money from the IP license.

As mentioned previously it is far more profitable to sell the entire product rather than the IP.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom