NVIDIA Open-Sources Tegra K1 Graphics Support 66
An anonymous reader writes "NVIDIA's next-generation Tegra K1 ARM processor now has open-source support for its Kepler-based graphics. NVIDIA decided to submit a large queue of patches to the open-source, reverse-engineered Nouveau project for supporting their ARM Kepler graphics with the open-source driver. The patches are still experimental but this is the first time NVIDIA has contributed open-source code to Nouveau."
Re: (Score:2)
And if we imagine there's some unlikey rationale we can use to weave your incredibly less than statistically significant chance of weaving corn into the already convoluted story line, your comment is only slightly less rational.
Re: (Score:2)
The word rational appears twice in your post, but would have been good if it wasn't preceded by
the word Christmas. As such, you can't be taken serious and any use of the former word has
to be seen as clownesque -- which is good, since clowns can hardly stand in the way of a free
internet.
Clickity-click (Score:4, Funny)
NSA cookies
Are you saying I could harness the resources of the NSA to make cookies [wikia.com]? Is it more powerful than an antimatter condenser [wikia.com]?
Nvidia has NOTHING to lose at this stage (Score:5, Interesting)
Tegra has been a horrid disappointment for Nvidia till now, and the competition in the ultra-mobile SoC market is ramping up at a terrifying rate.
-Tegra 1. The equivalent of Microsoft's Windows 1,2. If it ever existed, no-one noticed.
-Tegra 2. Horribly late, missing NEON, and missing hardware acceleration for H264 video decode. Used in devices only because Nvidia was forced to give it away.
-Tegra 3. First ARM SoC part from Nvidia worth using. Late, but good enough to get still get some major contracts as a highish end part.
-Tegra 4. Pretty much an unmitigated disaster. Late and expensive enough to lose the small progress Tegra 3 had made. Wrongly specced, so Nvidia had to announce the 4i.
-Tegra 5, renamed the K1. Built on the wrong process (not really Nvidia's fault- TSMC and others have failed to make the shrink progress expected years ago when this part was first planned). Using the wrong ARM core (A15), so Nvidia had to announce a later version of the K1 that will come with Nvidia's own 64-bit ARM core. Of course, this means the first K1 is already obsolete, long before it is on sale. First Tegra with PC class GPU cores, but not the NEW Maxwell GPU architecture Nvidia launches on the desktop in a few weeks time (750TI). So, the GPU is also out of date before the K1 goes on sale.
The Tegra 5/K1 has a lot of graphic clout for an ARM SoC, BUT cannot use that power in a phone/normal tablet form factor. Therefore, Qualcomm and Apple will best the K1 in performance per Watt, once again.
So, Nvidia has zero (ZERO!!!!!) to lose by throwing out all the tech details of the K1 into the public arena. Intel pulls the same stunt with its laughably poor integrated GPUs on its current CPU chips. If you can't compete, make your documentation open-source in the hope this will boot-strap some extra business.
Re:Nvidia has NOTHING to lose at this stage (Score:5, Insightful)
If you can't compete, make your documentation open-source in the hope this will boot-strap some extra business.
Too little too late. For YEARS we have been screaming for nvidia drivers that aren't buggy, closed and unstable, to the point of writing Nuveau, an open source hack (remarkably good but still crippled). Rot in hell, NVIDIA - I have wasted enough money on your hardware.
Re: (Score:1)
NVIDIA Linux drivers are closed source and complex, but you must admit they have good quality.
They used to be far better than ATI/AMD proprietary stuff.
Only Intel write good opensourced graphic drivers, but their hardware was so limited that their
drivers didn't need the complexity of serious OpenGL drivers.
Re:Nvidia has NOTHING to lose at this stage (Score:4, Interesting)
Stop spouting this rubbish. Intel's pretty decent performance wise these days and NVIDIA's drivers have always sucked no matter how much better the performance was.
NVIDIA's graphics drivers don't' f'ing work without head pounding. I shouldn't have to f around with the terminal for hours to install a proprietary driver that only half works on a select set of distributions.
The days where Intel's graphics sucked are long over. It's not the 1990s. Intel's graphics are pretty good. Intel's 3rd generation graphics were decent. Almost comparable at the low end with NVIDIA. The newer 4th generation stuff is pretty impressive although unfortunately Iris Pro has been restricted to integrated CPUs and thus no socketed CPUs have it. As a result motherboard manufacturers have chosen to opt out in protest. Nobody ships an Intel Iris Pro mini itx motherboard. In fact there are very few Iris Pro systems. I have one of the very few that exist in fact. It's an ultrabook-like form factor 14” screen.
AMD's drivers still suck and they are still non-free despite the public relations stunt to “open” them.
While I hope this actually helps improve the free drivers for NVIDIA's graphics chips I'm doubtful. I'm not that familiar with these chips although I'm pretty sure they are targeted at and only available in cellular devices and similar. It won't help the desktop users.
Re:Nvidia has NOTHING to lose at this stage (Score:4, Interesting)
The nVidia drivers on FreeBSD are pretty solid, but they got a poor reputation for their open source drivers in the early releases. I was running a room full of Linux boxes about 10 years ago, and they'd all kernel panic about once a day, typically while running nothing more strenuous graphically than the log-in screen, and always with a backtrace in the nVidia drivers. The open source ATi drivers of the same era (R200) were a lot slower, but were very stable.
nVidia also had that embarrassing incident where a crafted image could cause arbitrary code execution in the kernel, which turned out to be exploitable by just putting a picture on a web page, and didn't fix it until about two years after they were first notified of it. For the last 4-5 years, their proprietary drivers have been pretty reasonable though.
Re: (Score:2, Insightful)
The days where Intel's graphics sucked are long over. It's not the 1990s. Intel's graphics are pretty good. Intel's 3rd generation graphics were decent. Almost comparable at the low end with NVIDIA. The newer 4th generation stuff is pretty impressive although unfortunately Iris Pro has been restricted to integrated CPUs and thus no socketed CPUs have it. As a result motherboard manufacturers have chosen to opt out in protest. Nobody ships an Intel Iris Pro mini itx motherboard. In fact there are very few I
Re:Nvidia has NOTHING to lose at this stage (Score:5, Interesting)
it is decisively beaten by pretty much every graphics card over $100. You can't game at 1080p or use MadVR with maximum settings on Iris Pro.
To be competitive on the desktop, Intel needs something about as powerful as a Radeon HD 7850 or GeForce GTX 650 Ti Boost. As of now they aren't even close.
I built my desktop with a 3.4GHz Core i5-3570K Ivy Bridge. Anyone telling you the HD Graphics 4000 is "good enough" for gaming is full of shit. Even my low rez 1200x1080 monitor, most games struggled to get 30 FPS at anything about the lowest detail level. When I got into Dota 2, that was the final straw. I caved in and bought a Radeon HD 7850 for $150. The difference is night and day. Integrated graphics are still garbage, worthless for anything beyond angry birds.
I dual boot to Linux and have a decent steam library. The only thing I'll give Intel, is that they do make decent open source drivers that perform nearly as well in Linux as Windows. The AMD open source drivers are terrible for gaming. They get 30-80% of the proprietary drivers FPS and have major issues with micro stuttering. And yes, I use the dev drivers from the edgy PPA along with all the tweaks like SB backend. They still suck.
Re: (Score:3, Informative)
" but you must admit they have good quality."
When a driver update kills the fan and thus kills the GPU by failing to cool it (320.18) you call that good quality?
Really?
Re: (Score:1)
Re: (Score:2)
"good" evidently does not mean what you think it does.
Nvidia Linux drivers are diabolical, as in "work of the devil" and I am bitterly disappointed. The devil (and presumably Microsoft, if a different entity) may well be pleased.
Re: (Score:2)
Not at all. It's never too late to see the light and do the right thing.
Re: (Score:2, Interesting)
So you are stuck with Intel with their FOSS drivers but terrible GPUs. AMD also was all proprietary and their FOSS drivers are not as feature rich as the closed source. So you are made about them not supporting FOSS in the past... But they are doing it now and you are still mad.... Seems counter productive to keep complaining after a company goes FOSS.
Re:Nvidia has NOTHING to lose at this stage (Score:4, Insightful)
"AMD also was all proprietary and their FOSS drivers are not as feature rich as the closed source."
Their Free drivers may not be as 'feature rich' but they're a heck of a lot more stable and compatible than the blobware.
I'm planning to buy new video hardware about the middle of the year and their chances of getting my money just went from 0 to... well to nonzero at least.
Re:Nvidia has NOTHING to lose at this stage (Score:5, Informative)
Don't get too excited. It's not like nvidia are actually opening up here:
This is only about leveraging the hard work already done by nouveau hackers, in order to bring their embedded SOC product to market more quickly. There was no documentation dropped, and they're specifically refuting the idea desktop linux support.
Re: (Score:2)
Re: (Score:2)
For Linux systems, NVIDIA was the worse graphics card manufacturer. Except, of course, for all their competition.
Now, with nouveau drivers starting to mature, things are getting better. I just install Linux Mint 16, and for the first time in installing linux systems with nvidia graphics drivers (probably in 2004?), I didn't install the nvidia graphics driver and things work at an acceptable speed.
Thanks, nouveau developers, for your excellent work so far. I hope Nvidia gives you more support in the futur
Re: (Score:2)
Re: Nvidia has NOTHING to lose at this stage (Score:1)
Re: (Score:2)
You mean last gen consoles ... current gen consoles are nearly an order of magnitude removed from the K1.
Re: (Score:3)
-Tegra 5, renamed the K1. Built on the wrong process (not really Nvidia's fault- TSMC and others have failed to make the shrink progress expected years ago when this part was first planned). Using the wrong ARM core (A15), so Nvidia had to announce a later version of the K1 that will come with Nvidia's own 64-bit ARM core. Of course, this means the first K1 is already obsolete, long before it is on sale. First Tegra with PC class GPU cores, but not the NEW Maxwell GPU architecture Nvidia launches on the de
Re: (Score:3)
64 bit is better because you use up all your cache memory in half the time. It's better when your cache is full, right?
Re: (Score:2)
ARMv8 supports both AArch32 (32-bit ISA) and AArch64 (64-bit ISA), similar to how AMD (and now Intel) CPUs support both x86 and amd64 ISAs.
Meaning, you can run a 32-bit OS on a 64-bit chip, and get access to all the improvements to the architecture, and it will run like a faster 32-bit chip.
Or, you can run a 64-bit OS on the 64-bit chip, and still run 32-bit apps, and get access to all the improvements to the architecture, and it will run like a 32-bit chip with access to a full 64-bit address space (for th
Re: (Score:1)
You're partially right, Tegra has been pretty poor so far. However it's to early to make a call on Tegra 5 yet. There are have been no independant benchmarks done on it. Also, we do not know which process it uses. It could well be 20nm. 20nm has been in test production since the start of 2013.
Re: (Score:1)
Sorry, forget that post. Indeed, k1 is 28nm, not 20nm.
The light bulb ... (Score:2)
finally pops on.
Finger power (Score:1)
See, if you get management's attention with a few precisely calibrated and executed gestures, stupidity can sometimes end.
Re: (Score:2)
Whoa, what have we got here, Nvidia manager with mod points? Come on, admit it, closing your driver source when you're trying to sell hardware is just plain stupid, no two ways about it.
For all the naysayers out there. (Score:4, Informative)
That jump all over how terrible Tegra SOCs are, the chips still power a crapload of cheap devices all over the planet. What opening up the source on these chips will do is make it easier for smaller companies to create Android and other OS based devices for the expanding cheap device market. What some here refuse to realize is that China is a have and have not market. Those who cannot afford iPads and iPhones will go for the best cheap alternatives and Samsung's products are not significantly cheaper than Apple products. The stuff that they make that is low end could easily be blown out of the water by other companies that clone both the iPad the iPhone and high end Samsung products like the top end Note series and Galaxy 4 phones. You can bet within a very short period of time there will be a flood of cheap knock offs that do everything that these devices do and with just as much grunt but much cheaper.
The high end portable device market that is run by SOCs is undergoing the same thing the pc market went through, aggressive competition and a patent portfolio will not adequately stop the production of knock offs. This is most likely what NVidia is counting on happening, all they care about is selling a gazillion SOCs as fast as they can, just like everybody else that relies on sales of hardware for revenue. NVidia realizes that their SOCs are not going to make it into iPads and Win8 RT is a complete bust so they are instead taking a run at Samsung by opening up their software specs and making cheap but much more powerful versions of Android and even things like Firefox OS and Ubuntu on arm a real possibility. No doubt this will make many more powerful cheap devices possible than what we currently see coming out of the east. This sort of game change is only to be expected, even if many would like to see NVIDIA FOD they are in a position to change the game simply by not playing by the old closed source software design rules that killed many manufactures in the PC market place. My prediction is the next company to bite the dust will be Creative unless someone like NVidia buys them out and teams up with someone like Lenovo to produce killer pro devices and the like as well as consumer do dads. There will be a huge consolidation in the industry and this time Microsoft and their so called "hardware partners" could be left in the dust, perhaps NVidia sees the writing on the wall this time and is breaking free from Redmond's apron strings for a change.
Microsoft's loss (Score:3, Informative)
This resulted in Microsoft effectively having billions of dollars worth of drivers that they didn't even have to pay for; a serious competitive advantage. But as many power users have moved over to Linux for various needs such as servers, rendering, and large scale computing; certain classes of drivers have become valuable for hardware manufacturers to port properly (or assist in the porting).
This won't kill Windows but it is a nice step toward leveling the playing field somewhat.
Re: (Score:3)
Microsoft don't have a library of drivers for ARM, in fact it's the other way round where windows for arm has an extremely limited set of hardware support and linux is far ahead because a lot of the drivers for x86 can be trivially recompiled.
Re: (Score:2)
deceptive title (Score:5, Informative)
The title should be "NVIDIA publishes Nouveau patches to support Tegra K1"
Nothing has been "open sourced" as it was never closed sourced to start with. It's all original code written specifically for Nouveau.
Re: (Score:2)
If you feel like being a pedant, consider that it was closed until released.
Or, do you want to argue about whether software with an open source license, that sits on a disk inaccessible to anyone but the author, is open?
I understood that the code was released with an open license. Finding that it is just patches does not significantly deviate from that meaning, certainly not enough to deserve bold.
Re: deceptive title (Score:1)
Perhaps you should realise that they were trying to distinguish between writing small patches to improve Nouveau with the intention of releasing those patches vs open sourcing existing code from their proprietary codebase.
Re: (Score:2)
this is probably a good way for nvidia to test the waters with regard to oss.
they probably have some brass that see it as:
"paying our engineers to write drivers, to give away for free to our enemies"
hopefully nothing will scare them off. ...ad try not to breath to much!
noone make any sudden moves! whisper!