Nvidia: The Cost of Arrogance

Recently Nvidia was scolded by Linus for withholding source code for their video drivers but the chickens really come home to roost when it hits the bottom line. A large Chinese project went to AMD over the same issue. The Chinese wanted Nvidia’s hardware to work with MIPS CPU…

“One of the leading Chinese companies were looking for GPUs to power their pilot school PC project. These PCs were running Linux on their own MIPS CPU. The company approached Nvidia to supply GeForce/Quardro  GPUs for their project. Unfortunately, the GeForce/Quardro drivers are available only for x86 based architectures. It’s not available for MIPS or even ARM. Nvidia’s officials told Chinese that it would cost them millions of dollars to port the code-base.

Chinese did not want to spend this much money on just porting the code-base. Since Nvidia never releases the source code of their drivers, the Chinese went to AMD which has open source drivers”

Needless to say, if Nvidia had opened their source code, the world would have ported their drivers to ARM and MIPS for no cost to Nvidia… When ARM invades the “PC” space, what will Nvidia do? Cede the field to AMD/ATI?

see Muktware – Nvidia Loses a Large GPU Order Due To Closed Source Drivers.

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
This entry was posted in technology. Bookmark the permalink.

10 Responses to Nvidia: The Cost of Arrogance

  1. oiaohm says:

    Viktor “A stable API/ABI would be something that’s useful for a change.”

    Userspace ABI has always been stable from kernel.

    One was provided from kernel space is was a common binary driver between BSD, Linux and all the other Unix’s. It was built and no one used it.

    Instead companies like nvidia said it was too slow and went after the internal structs of the OS’s. You cannot provide a stable ABI if people will just go past it anyhow and use the internal structs of the OS then attempt to say you cannot change that even that it was marked internal usage only. Mind you nvidia does this under windows.

    Basically unless you can prove that companies like Nvidia will use a common binary driver format there is no point adding it.

    The firmware loading system in Linux is basically identical between BSD and Linux. So yes there is no reason why a new stable binary ABI could not be cooked up almost instantly.

    Embed world use fuse interfaces it implement the binary drivers they need in userspace that has a perfectly stable abi. So one section of the market is perfectly happy with Linux driver abi setup. The other group containing items like video cards and companies like Nvidia do there own wrappers from there o so secret core binary blob they use in everything.

    You can complain about lack of ABI when nvidia and others have put forward a common ABI they agree to use. FOSS did the first attempt and was slapped in the face by closed source driver makers. You cannot expect them to step up again its the turn of the other side now.

  2. Viktor wrote, “Nvidia is a spot of sanity”.

    Requiring people to jump through hoops to install drivers is “sane” when Viktor thinks people have no need to touch source-code in most cases and GNU/Linux makes loading drivers completely transparent to users? Silly.

  3. kozmcrae says:

    Viktor wrote:

    “And Linus should just shut his mouth.”

    You are nothing compared to Linus Torvalds, Viktor. You are a piece of she it compared to him. You have done nothing to make this World a better place. You come here and work to make the World a crappier place to live. Linus Torvalds has done much to help this World and has the cred to prove it. You are dirt.

  4. Viktor says:

    Nvidia is a spot of sanity in the f*cked up world of GNU/Linux. And Linus should just shut his mouth. Why doesn’t he instead beat some sense in Greg Kroah-Hartman? A stable API/ABI would be something that’s useful for a change.

  5. Clarence Moon wrote, of smart thingies, “why are they not showing up with half or more of the web usage?”

    Unique IPs will show a lot more mobile gadgets than page views or some other statistic. That’s because the screens are smaller. Smart phones still outnumber tablets by a wide margin. When tablets are more numerous, share of page views will increase dramatically.

    Still 17% of page views for wikimedia are mobile. It was only 7.5% a year earlier, more than 100% per annum growth. I expect it to slow down a bit because much of the English world is ARMed but then there is Asia where the trend is much stronger.

  6. Clarence Moon says:

    ARM PCs are already common products, counting smart phones as PCs which I do

    Let us know when someone else of any note thinks the same, Mr. Pogson. A PC is not a phone is not a tablet and all other combinations of that.

    There are already, from your own claims, about as many feature phones, smart phones, and various kinds of tablets with internet access as there are PCs in the world. So why are they not showing up with half or more of the web usage?

    Any way, you have several months to go before you will have to rescind your claim that there will be lots of “GNU/Linux” on the store shelves for Christmas this year. Do you still like your chances?

  7. kozmcrae wrote, “If the only computer you own is a mobile phone, then it’s your personal computer.”

    True, but that’s still limiting. My home has more than a dozen personal computers ranging from smart phone to Beast. They are all PCs, and servers and they run like the wind.

  8. kozmcrae says:

    If the only computer you own is a mobile phone, then it’s your personal computer. Microsoft is known for changing the definition of words to suit their aims. It looks like the definition of PC is changing without their help and without their blessing.

  9. Clarence Moon wrote, “Their decision to not support MIPs is simply economic in that they think that there isn’t enough money in such a backwater configuration to make it worth their while to participate.”

    Nope. They cannot see over the horizon to China becoming independent in IT within five years, their national goal. China is a market all by itself worthy of Nvidia or any other global corporation.

    ARM PCs are already common products, counting smart phones as PCs which I do. Everyone I know uses their smart phones to access lots of stuff on the network. My son, for instance, can check a price at a local supplier, in real time, without interrupting his conversation and feed the new data into the conversation at his home, in my home, in the bush etc. It’s his main PC outside of work. At work he brought his own PC, a notebook, but it’s all mobile. This weekend he was target shooting, got a call about a server and talked a guy through a script using his smart phone. Took five minutes and he was back to the real world. A smart phone is a real PC. The only question is how many smart phones the world will have online compared to x86/amd64 PCs. I expect it will be more, far more, because of price, size, mobility and utter usefulness.

  10. Clarence Moon says:

    When ARM invades the “PC” space, what will Nvidia do? Cede the field to AMD/ATI

    IF/When ARM PCs become common products, Nvidia will likely provide products compatible with the hardware and OS environments involved. Their decision to not support MIPs is simply economic in that they think that there isn’t enough money in such a backwater configuration to make it worth their while to participate.

Leave a Reply