Dear NVIDIA this is directed to You

[COCAYven]

Do You think NVIDIA is holding Us back?

Let Your money and opinion count for something.
Maybe the more of You will post a thread the more finally We will get proper gpu technolgy and hardware for what we pay and especially for the year we are in.

I am a CG artist even if i used to build Chips for space programmes, I am fed Up with NVIDIA releasing something better just every 2 years.

The best Gaming GPU is worth almost a month salary, it sucks a billion watts and is not enough to run 3 30inch displays all set to ultra.(thats if You are a gamer)
Companies like adobe or autodesk and many others are held back into further developing their software comforming totally to CUDA simply because really powerful cuda GPU are affordable to few and anyhow the high end stuff is basically still super slow and poor in computing performance).

It will not render on cudas all the time(eg autodesk maya) and even if faster than CPU still it will neither raytrace perfectly nor dramatically decrease render times in order to have a complex scene ready in few minutes (That meaning the time to smoke a cigarette).

Same is for quadro cards which of which I am a regular customer.

We dont want SLI we dont want to scale.

We want a single GPU to do what nothing else can do DRAMATICALLY for what It costs.

We want the smallest GPU to be able to run in ultra settings for gamers on a small cheap computer, and the biggest GPU to compute and perform at least on a 6 screen config with flawless framerate and not using over 150 watts.

We want the most expensive quadro cards to be released at least every 6 months, and to perform at least 50% better than any GEFORCE, im programs made by companies that are delighted to develop over cuda tech rather than finding direct x more appealing.


Yes no one can compete with NVIDIA however this is not about being the only to dominate the market , this is about You being able but still holding back our lives, our work and our leisure time.

Truth is however that both NVIDIA and AMD top GPU lines suck, they will look amazing simply because they run spartan games, and yes even the most demanding games to day are spartan and way far from realistic from a graphics and physics point of view.(this doesnt mean that a pokemon game or arma 3 look ugly however they are still spartan build compared to what is really available out there in the world in HIGH tech)

It is absurd that we are a handful years from having processor capable to compute faster than a human brain while we still got best games available today way far from realistic graphics.
You at NVIDIA are holding back hundreds of thousands of professionals that just want to get work done, in this moment of history where nanotechnologies are proliferating and biotechnologies are going to offer 200 years life span to human being within the next decade.


You Dear NVIDIA are a wonderful company, but now it is about time You get out the good stuff,and more frequently instead of finding marketing strategies to make people buy more of the same thing all the time or forcing them to scale in whatever they wish to do........ no wonder nobody wants to invest too much in servers......

What is the problem, release soemthing that a client can know that in 5 years time he is still offering a beast and has to have a hard time to decide if reinvest its amortisation savings or if to keep what he got a few years more.....that is what in finance is called added value.

Truth is dear fellows that a top end GPU if it cannot run this simple low ploy games at over 150 FPS in at least a 3 30inch config then it means that the best GPU is a piece of junk for the price asked and Maybe NVIDIA should better think into revisiting their company manpower costs.

Selling top end GPU is about delighting the client AND THE PROFESSIONAL, not about cheating peole into buying something because You are the only one making it and still not being able to offer a GPU that can satisfy the need for the heaviest configurations.

You really wanna sell GPU computing, make GPU that really process the needs of millions of people connected to millions of servers.

It is apalling to have to state as a professional that in year 2014 we still have to conform to offer low poly and spend time optimizing geometries, worse to say wait so much time for new releases in professional graphics.

Release decent stuff once and for all, and You will give the opportunity to every professional to finally release true virtual reality, not this low poly trash we all fed up with.

So Anyone here in the forums that agrees that the smallest gpu should be able to run any game in ultra settings in a small computer with a small screen and the biggest GPU should be able to handle the biggest screen configuration without the need to invest into power supplies speak Your mind.

Please refrain from saying that there are no FABS available or ready to make better processors or that the cost is prohibitive...

The true cost of a system on CHIP on a wafer of 200k units is approx 200k dollars in full production, so basically this means that the cost of teh processor in any GPU is of 1 dollar .

The cost to manufacture in a FAB the prototype is 400 thousand dollars, after debug You will spend another 400k for the final chip.
On top of this you will have to add all the costs of the cars employees drive, light, salaries, investments in assets, dividends, taxation.

To make an example the true cost to assemble an iPhone without the RD before it is 8 dollars.

Now il leave it up to You to fantasize on how much is the true cost of assembling a GTX titan or a quadro K6000 after RD in mass production......

Trust me..... a pack of cigarettes might just cost more....

So OK we want all, NVIDIA to be not profitable but super profitable and grow a lot and give more and more jobs and make all investors happy....but if You agree with me and think that they should start offering really good stuff and stop holding back the world post it here.


ask for added value.

 

[Rationale]

Learn to use grammar and I might take you more seriously. Until then, you can take your random speculation about humans living 200 years, and your entitlement complex (the fact we have to optimize is appalling, really?) and go. There's too much naivety and half-formed pseudo-science in this to even be worth addressing.

Are you aware that humans used to live in shacks and survive on a diet of grain and goat cheese? You should be happy to even have life. You should be absolutely ecstatic to have a PC in a world where so many people can barely survive. The fact that you're basically demanding 50% faster, cheaper PC components every 6 months or it's "holding your life back an unacceptable amount" is entirely disgusting, and a horrible example of consumerism and entitlement gone awry.

 

[Someone Somewhere]

Have you thought that perhaps the MULTI-BILLION DOLLAR FAB FACILITIES cost anything?

R & D ain't free. You still need to cover the cost of those from your sales, not just the cost per unit.

The actual cost of components in an iPhone, IIRC, was somewhere around $200. You may be thinking of the assembly.

 

[wdmfiber]

Dude, Nvidia is a small company, a middle man(by Wall Street standards).
TSCM are the guys in charge, they make the chips for Nvidia, AMD(Radeon), Apple ect...
It's does suck were still at 28nm(late 2011 technology), but the high-end "Maxwell" GPU's are getting closer.

AMD is broke and Nvidia hardy has any money. Microsoft and many others easily make several billion a quarter(every 3 months). But it takes Nvidia a year to make a few hundred million. It's not enough. They have to outsource!

 

[COCAYven]

Have you thought that perhaps the MULTI-BILLION DOLLAR FAB FACILITIES cost anything?

R & D ain't free. You still need to cover the cost of those from your sales, not just the cost per unit.

The actual cost of components in an iPhone, IIRC, was somewhere around $200. You may be thinking of the assembly.

Yes average cost fo a FAB is 5 billion average cost for a SOC 800k , I phone still costs 8 bucks, annual cost for RD 3.5 million. I can send You all spread sheets if You want.

 

[COCAYven]

Learn to use grammar and I might take you more seriously. Until then, you can take your random speculation about humans living 200 years, and your entitlement complex (the fact we have to optimize is appalling, really?) and go. There's too much naivety and half-formed pseudo-science in this to even be worth addressing.

Are you aware that humans used to live in shacks and survive on a diet of grain and goat cheese? You should be happy to even have life. You should be absolutely ecstatic to have a PC in a world where so many people can barely survive. The fact that you're basically demanding 50% faster, cheaper PC components every 6 months or it's "holding your life back an unacceptable amount" is entirely disgusting, and a horrible example of consumerism and entitlement gone awry.

As a matter of fact ......A buon intenditore poche parole.

 

[Rationale]

Learn to use grammar and I might take you more seriously. Until then, you can take your random speculation about humans living 200 years, and your entitlement complex (the fact we have to optimize is appalling, really?) and go. There's too much naivety and half-formed pseudo-science in this to even be worth addressing.

Are you aware that humans used to live in shacks and survive on a diet of grain and goat cheese? You should be happy to even have life. You should be absolutely ecstatic to have a PC in a world where so many people can barely survive. The fact that you're basically demanding 50% faster, cheaper PC components every 6 months or it's "holding your life back an unacceptable amount" is entirely disgusting, and a horrible example of consumerism and entitlement gone awry.

As a matter of fact ......A buon intenditore poche parole.

That quote isn't relevant or insightful.

 

[JackNaylorPE]

You didn't finish .....

Where's the part about sitting on my duff at home and having a big check arrive once a month to support me in the style I'd like to be accustomed to ? Pricing is set by the market ... what people are willing to pay and what the competition is able to put on the market.

nVidia is doing what the public asks them to do. Companies don't stay in business by building what people don't want. They have 20 outta the top 30 GFX cards by market penetration as measured by peeps hitting steam servers (gaming community). That's 4 times as many as Intel and 4 times as many as AMD. They have 7 outta the top 10 (intel has other 3) and 13 outta the top 20 (Intel has 5, AMD has 2).

They also have the most desired workstation cards ..... people spend money on those cards because they are able to make money themselves. No business invests in technology that doesn't have a return on investment. Your argument that some mysterious cheap technology exists out there which nVidia is holding back is specious because if it was so easy, a competitor would have pursued it and released it.

Now do I think that nVidia ever holds back their "ace in the whole" once the competition reveals their hand ? Of course they do. It's long been assumed that the 7xx line was "renumbered" and the top card shelved after AMD revealed their lineup. But why should they release the "ace in the hole" when the hand they are holding already beats the competition .... why not save it till the competition comes up with something ? The more complex the card, the lower the yield. So if their no. 2 card can top the competition's No. 1 card, why would they release the top card ? If AMD sets the bar for their top of the line product at say $600 ..... then why would they release their top card which costs $575 to make to compete with it when their No. 2 card costs $475 and can compete or even top it ? Why not leave it on the shelf, ready to go until the competition releases a new lineup then take out the ace in the hole and restructure the pricing of the rest to compete ?

Do I wish a pair of 780 Tis was $200 ? Of course I do ..... but if I had the proverbial 3 wishes.... that wouldn't be anywhere near the top of my list.

 

[COCAYven]

Dude, Nvidia is a small company, a middle man(by Wall Street standards).
TSCM are the guys in charge, they make the chips for Nvidia, AMD(Radeon), Apple ect...
It's does suck were still at 28nm(late 2011 technology), but the high-end "Maxwell" GPU's are getting closer.

AMD is broke and Nvidia hardy has any money. Microsoft and many others easily make several billion a quarter(every 3 months). But it takes Nvidia a year to make a few hundred million. It's not enough. They have to outsource!

You didn't finish .....

Where's the part about sitting on my duff at home and having a big check arrive once a month to support me in the style I'd like to be accustomed to ? Pricing is set by the market ... what people are willing to pay and what the competition is able to put on the market.

nVidia is doing what the public asks them to do. Companies don't stay in business by building what people don't want. They have 20 outta the top 30 GFX cards by market penetration as measured by peeps hitting steam servers (gaming community). That's 4 times as many as Intel and 4 times as many as AMD. They have 7 outta the top 10 (intel has other 3) and 13 outta the top 20 (Intel has 5, AMD has 2).

They also have the most desired workstation cards ..... people spend money on those cards because they are able to make money themselves. No business invests in technology that doesn't have a return on investment. Your argument that some mysterious cheap technology exists out there which nVidia is holding back is specious because if it was so easy, a competitor would have pursued it and released it.

Now do I think that nVidia ever holds back their "ace in the whole" once the competition reveals their hand ? Of course they do. It's long been assumed that the 7xx line was "renumbered" and the top card shelved after AMD revealed their lineup. But why should they release the "ace in the hole" when the hand they are holding already beats the competition .... why not save it till the competition comes up with something ? The more complex the card, the lower the yield. So if their no. 2 card can top the competition's No. 1 card, why would they release the top card ? If AMD sets the bar for their top of the line product at say $600 ..... then why would they release their top card which costs $575 to make to compete with it when their No. 2 card costs $475 and can compete or even top it ? Why not leave it on the shelf, ready to go until the competition releases a new lineup then take out the ace in the hole and restructure the pricing of the rest to compete ?

Do I wish a pair of 780 Tis was $200 ? Of course I do ..... but if I had the proverbial 3 wishes.... that wouldn't be anywhere near the top of my list.

NVIDIA is the market maker, if 780 Tis would cost 200 a pair , would You still buy them if volta was released this june instead of maxwell?..........

Do You really think that Volta is not ready for mass production and fully debugged?

The whole point of this thread is about NVIDIA being late, and feeling comfortable to sell and resell the same old stuff over and over , which to most might seem a way to optimize sales however this causes costs to rise, reduce sales compared to its true potential.

If In june volta was released instead of maxwell, what kind of games or virtual simulations do You think companies would work on? same as they do currently?
If a pair of 780 TIs would cost 200$ do you think TV sales would increase or decrease?

Yes by delaying release You defend the pricing however while You may sell some units at a high end price to few, the industry which is where the big money is not offered an up to date product to compete with alternative solutions to invest in nor financially neither in human resources. This is the main reason for which cuda is NOT the standard and big tech giants as medium ones who can afford it will have their processors build outsourced for their need with their own software.
Cuda should have been way ahead of android which was nice and ready in 2006. So why is tegra so important now? Seems to me NVIDIA is kinda 7 years late on schedule.
What was a 21 year old company like NVIDIA doing in 2006? fighting on how much to sell cuda to Google?

No problem You got a workstation You better stick to NVIDIA, but hell yeah waiting for quadro cards 2 years that worked 30% on average faster than fermi is kinda really poor performance for 9 thousand employees.

And yes pricing is indeed set by the market this is why NVIDA sells so little.
If there was someone else competing really they wouldnt sell at all....so the founder may go back to work at AMD where he started.
NVIDIA

Revenue $4.2 billion (2013)
Operating income $648 million (2012)
Net income $581 million (2012)

Total assets

US$ 6.412245 billion (2013) [1]
US$ 5.552928 billion (2012) [2]

Total equity $4.14 billion (2012)
Employees 8,800 (2014)

 

[JackNaylorPE]

Sells so little ? As in 7 outta the top 10 GPUs in the market ? As in 20 of the top 30 ? .... 13 outta the top 20. If that's "little" gotta feel sorry for the company that has none in the top 10 and 2 in the top 20.

If you really think that nVidia could drop the next generation on the market today, don't you think they would do so ? AMD just dropped their new R7 / R9 lines so next round is a ways off.... if nVidia could drop not 1 but 2 generations ahead, don't you think they would do so ? They'd have a 2 generation advantage and if they could drop the cards now at the prices you think they can, AMD would have no answer and become irrelevant .... end result no competition and they could charge what they want. So either nVidia is unique in corporate culture and has a conscience or they simply can't do what you think they can do.

 

[Someone Somewhere]

I'd put AMD at competing quite happily with Nvidia...

If Nvidia sold 780Tis for $100 each, then they'd be selling them at a loss, I expect. They certainly wouldn't get the next generation out the door, plus no-one would upgrade for several years.

CUDA isn't the standard because it's only made by one vendor. As any procurement person will tell you, vendor lock-in is a Bad Thing.

I'm completely lost as to why you're comparing CUDA to Android.

I highly doubt Volta is ready for mass production, considering Maxwell isn't really out yet. If it was ready, why wouldn't they be pushing it out to market ASAP to beat AMD to market? Obvious answer; it's not ready yet.

 

[COCAYven]

Sells so little ? As in 7 outta the top 10 GPUs in the market ? As in 20 of the top 30 ? .... 13 outta the top 20. If that's "little" gotta feel sorry for the company that has none in the top 10 and 2 in the top 20.

If you really think that nVidia could drop the next generation on the market today, don't you think they would do so ? AMD just dropped their new R7 / R9 lines so next round is a ways off.... if nVidia could drop not 1 but 2 generations ahead, don't you think they would do so ? They'd have a 2 generation advantage and if they could drop the cards now at the prices you think they can, AMD would have no answer and become irrelevant .... end result no competition and they could charge what they want. So either nVidia is unique in corporate culture and has a conscience or they simply can't do what you think they can do.

Most laptops dont gear with NVIDIA.
As Most Car displays nor its electronic set ups.
Nor Nautical displays, nor radars nor GPS nor loran.
No airplane display nor nothing computing on an airplane.
No Tv (yes that computes too) is geared with NVIDIA either.
No laundry machine has anything with nvidia written on it.

The short list above and there is much more is a 80 billion market to compete in.....how much was NVIDIAs sales in the last year?.... a mere just above 1 billion......but yes double of what was sold the year before.

As for selling little yes they sell little.....You tend to forget that the bulk of the income made by companies like NVIDIA is made by selling GPUS as processing devices stacked into servers where IBM is as much as a competitor as AMD.

If You want to really make a ranking for the restricted gaming world You should NOT compare NVIDIA with amd but EVGA Sapphire and Asus, then once You did that You can go and see how each of these companies classify compared to AMD and NVIDIA...oh yes dont forget Intel in the soup.

 

[COCAYven]

I'd put AMD at competing quite happily with Nvidia...

If Nvidia sold 780Tis for $100 each, then they'd be selling them at a loss, I expect. They certainly wouldn't get the next generation out the door, plus no-one would upgrade for several years.

CUDA isn't the standard because it's only made by one vendor. As any procurement person will tell you, vendor lock-in is a Bad Thing.

I'm completely lost as to why you're comparing CUDA to Android.

I highly doubt Volta is ready for mass production, considering Maxwell isn't really out yet. If it was ready, why wouldn't they be pushing it out to market ASAP to beat AMD to market? Obvious answer; it's not ready yet.

When You switch on Your app on Your smart phone ask Yourself if that app is running on CUDA code or android, then look at all the other apps and ask Yourself teh same question, then once YOu did that go to the APPS store and look for a CUDA coded app and android coded APPS, when you finished if you bothered at all ask Your self if it is necessary to have a tegra SOC in Your device and then You will understand why CUDA didnt compete with android within the WIMAX consortium back in 2006, while Steve Jobs was killing motorola and samsung marketing how to crack Iphone codes on the net for the common user to learn how to make apps.

 

[Someone Somewhere]

CUDA is *not* a competitor to Android. Possibly to the Dalvik VM (which is based on Java, and can thus run on practically anything; there are even x86 phones out there now).

Samsung is the largest manufacturer of Android phones. Why would they put a slower Nvidia chip (that didn't exist yet) in their phones instead of going with their in-house Exynos designs?

Also, remember that CUDA is proprietary. Any company whose software relies on CUDA is going to see reduced sales (because people are going to care more about the hardware and less about the software, especially in the early days of smartphones), and Google would probably never let it into the Play Store/Market.

Also, CUDA tends to be more difficult to develop with, and is thus only used for the most processing power intensive and extremely well threaded tasks. Most of the software is run on the CPU.

 

[Rationale]

I'd put AMD at competing quite happily with Nvidia...

If Nvidia sold 780Tis for $100 each, then they'd be selling them at a loss, I expect. They certainly wouldn't get the next generation out the door, plus no-one would upgrade for several years.

CUDA isn't the standard because it's only made by one vendor. As any procurement person will tell you, vendor lock-in is a Bad Thing.

I'm completely lost as to why you're comparing CUDA to Android.

I highly doubt Volta is ready for mass production, considering Maxwell isn't really out yet. If it was ready, why wouldn't they be pushing it out to market ASAP to beat AMD to market? Obvious answer; it's not ready yet.

When You switch on Your app on Your smart phone ask Yourself if that app is running on CUDA code or android, then look at all the other apps and ask Yourself teh same question, then once YOu did that go to the APPS store and look for a CUDA coded app and android coded APPS, when you finished if you bothered at all ask Your self if it is necessary to have a tegra SOC in Your device and then You will understand why CUDA didnt compete with android within the WIMAX consortium back in 2006, while Steve Jobs was killing motorola and samsung marketing how to crack Iphone codes on the net for the common user to learn how to make apps.

You can't really believe Steve Jobs killed Samsung. Samsung is both larger and growing faster than Apple.
http://static2.businessinsider.com/image/51a6234decad04fa61000005-960/slide-411-1.jpg
http://rack.0.mshcdn.com/media/ZgkyMDEzLzA2LzEyLzgyL1NoYXJlczEybW9uLmJkNDBjLmpwZwpwCXRodW1iCTEyMDB4OTYwMD4/160000f5/560/Shares-12-months-Samsung-vs-Apple-vs-Microsoft1.jpg

 

[ToineF]

We need more competition imo!

 

[ewok93]

Competition will only affect prices, not the actual technology. If we have three big companies, AMD, Nvidia, and one other, then those two will have less money for advancing technology, as competition will reduce costs, but not increase people buying parts. And are you trying to say you wish technology didn't advance as quickly so you could have the best for longer? Or are you trying to say the best were cheaper? Because if it gets any cheaper, those companies will not have the money to produce newer chips, and we'll be stuck.

That would be a bad thing.

 

[ToineF]

Competition will only affect prices, not the actual technology. If we have three big companies, AMD, Nvidia, and one other, then those two will have less money for advancing technology, as competition will reduce costs, but not increase people buying parts. And are you trying to say you wish technology didn't advance as quickly so you could have the best for longer? Or are you trying to say the best were cheaper? Because if it gets any cheaper, those companies will not have the money to produce newer chips, and we'll be stuck.

That would be a bad thing.

imo it goes both ways. Apple pushed microsoft to inovate for years and years. Now both companies steal ideas from eachother. I don't think it's a matter of price in this case.

 

[Someone Somewhere]

No; competition also drives product innovation. In a monopoly there's no reason to spend money on R & D; it all goes straight to the shareholders.

 

[Pinhedd]

ah yes, another one of these "I'm an artists and I don't design chips but if I did..." posts.

You don't design ASICs, so don't tell those who do what they can/can't/should/shouldn't do.

The cost of prototyping a new design is in the order of millions of dollars, and the cost of bringing one to full production on an existing fab is in the tens of millions to hundreds of millions depending on scale.

Production costs for a tiny 20 square mm chip might be around $1 and unsurprisingly that's not much less than what they sell for. Scale that to a 500+ square mm in an environment in which defects render more than half of the products worthless and the price skyrockets.

 

[ewok93]

There IS competition already, more brands will not increase how quickly new components are being created, as they're basically going as fast as current technology allows them to. A lot of the time each brand will just release two equal cards, rather than leapfrogging the other.

 

[COCAYven]

CUDA is *not* a competitor to Android. Possibly to the Dalvik VM (which is based on Java, and can thus run on practically anything; there are even x86 phones out there now).

Samsung is the largest manufacturer of Android phones. Why would they put a slower Nvidia chip (that didn't exist yet) in their phones instead of going with their in-house Exynos designs?

Also, remember that CUDA is proprietary. Any company whose software relies on CUDA is going to see reduced sales (because people are going to care more about the hardware and less about the software, especially in the early days of smartphones), and Google would probably never let it into the Play Store/Market.

Also, CUDA tends to be more difficult to develop with, and is thus only used for the most processing power intensive and extremely well threaded tasks. Most of the software is run on the CPU.

Totally agree....what im saying is that if cuda was ready as android was back in 2006.......cuda would have been the standard and power efficient tegra would have been released 7 years ago. this is just an example of what happened to the mobile market development.... if google wanted to buy cuda today im pretty positive that NVIDIA would be more than glad to sell it and recapitalize, we would easily see NVIDIA shares grow 70 times in 24 months.
p.s.time will tell if people will be glad to code in cuda on maxwell, or if its words to the wind...as for now seeing titans sold at 1k is like seeing Intel selling Phi coprocessors sold less than half price and begging people to code commercial programs for it.

 

[Someone Somewhere]

CUDA and android are semi-standards for completely different things. You might as well say that we should be rendering web pages using UEFI.

Why on earth would Google want to buy CUDA?

 

[Pinhedd]

CUDA is *not* a competitor to Android. Possibly to the Dalvik VM (which is based on Java, and can thus run on practically anything; there are even x86 phones out there now).

Samsung is the largest manufacturer of Android phones. Why would they put a slower Nvidia chip (that didn't exist yet) in their phones instead of going with their in-house Exynos designs?

Also, remember that CUDA is proprietary. Any company whose software relies on CUDA is going to see reduced sales (because people are going to care more about the hardware and less about the software, especially in the early days of smartphones), and Google would probably never let it into the Play Store/Market.

Also, CUDA tends to be more difficult to develop with, and is thus only used for the most processing power intensive and extremely well threaded tasks. Most of the software is run on the CPU.

Totally agree....what im saying is that if cuda was ready as android was back in 2006.......cuda would have been the standard and power efficient tegra would have been released 7 years ago. this is just an example of what happened to the mobile market development.... if google wanted to buy cuda today im pretty positive that NVIDIA would be more than glad to sell it and recapitalize, we would easily see NVIDIA shares grow 70 times in 24 months.
p.s.time will tell if people will be glad to code in cuda on maxwell, or if its words to the wind...as for now seeing titans sold at 1k is like seeing Intel selling Phi coprocessors sold less than half price and begging people to code commercial programs for it.

Android is barely even ready in 2014. The top Android manufacturers can't get the lockscreen to respond in a timely fashion half the time and Chrome can't go an hour without crashing after it runs into a bad script. Android isn't terrible, but it's free and you get what you pay for. Scientific and medical imaging companies aren't in the business of "free", they're in the business of reliable and predictable and they're willing to pay big bucks for it.

CUDA is the key to exploiting the massive vector acceleration exposed by NVidia's hardware in a very vendor specific fashion. OpenCL is more platform portable, but lacks the specific optimizations enabled by CUDA. NVidia relies on CUDA to pitch their hardware as being more than just a render chip. It's inseparable from the hardware, there's no way in hell that they'd consider selling it without selling the core of their business in a takeover. There's also no reason for Google to want it in the first place.