Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Nvidia announces Tegra K1, due in the first half of this year
Which comes next in this sequence:
Tegra
Tegra 2
Tegra 3
Tegra 4
If you said "Tegra 5", then you didn't read the title. Nvidia marketing got bored and decided to name it Tegra K1 instead.
Regardless, it is what we expected: 4 ARM Cortex A15 cores together with a single Kepler SMX. As the thread title says, it's due in the first half of this year, though that might mean "Nvidia will start selling it to tablet vendors" and not "devices will be available at retail".
Nvidia also said that in the second half of this year, there will be an updated version with two "Denver" cores, which is Nvidia's custom version of 64-bit ARM v8 cores.
All of this will still be on 28 nm, and the CPU isn't any better than available in Tegra 4. But on the GPU side, the architecture is moving from 2005 state of the art to 2012 state of the art, which is quite a jump. If you want an Android gaming tablet, this is the Tegra chip you've been waiting for. Android devices that can do modern 3D graphics rather than being limited to a really gimpy OpenGL ES API will be a welcome development.
-----
In other rumors, Nvidia is rumored to launch their first next generation Maxwell video cards relatively early this year. They're still on 28 nm, so it's not at all clear whether this will be a meaningful advance over Kepler. But AMD is stuck on 28 nm, too, so Kepler is still competitive with AMD's latest and greatest.
Comments
I wonder if this means that there will be games that are only capable of running on Nvidia powered tablets by the end of 2014, or if developers will still develop for the most common install base.
I can not remember winning or losing a single debate on the internet.
While I don't expect many games to target Tegra K1 tablets only, if you've made a game for desktops that uses OpenGL and a programming language that also compiles for Android, then porting it may well be just some extra debugging and not require restructuring anything in your code or completely forking your code base. Mac OS X and (desktop) Linux don't have enough of a user base to push OpenGL adoption, but Android might.
There's also the issue of future compatibility. If a year from now, most Android tablets don't support OpenGL, but four years from now, basically all of them do, then many games launching four years from now might well use OpenGL. (Remember that for Android and iOS, DirectX is not an option.) A Tegra K1-based tablet would still be able to run those games then, while alternative GPU chips in Android tablets that you could get now or soon won't.
There's also the issue that I think is important and doesn't get enough publicity: more versatile GPU chips let you offload more functionality from the CPU to the GPU. If some computations are GPU-friendly, doing them on the GPU is vastly more energy-efficient than doing them on the CPU, even for an energy-efficient CPU architecture like ARM. Mobile devices with small power envelopes can't exactly afford to waste power by doing GPU-friendly computations on the CPU, which is what you have to do a lot if you don't have a modern GPU chip.
While the Tegra K1 is a stupid name, it at least is not a stupidly long name. Though with two variants coming, I really hope that Nvidia doesn't call them both Tegra K1. Calling the Denver version Tegra K2 would make sense. Which is why Nvidia won't do that.
You forgot to link this
https://www.youtube.com/watch?v=GkJ5PupShfk
Framerate is bad but it is pulling off some great special effects.
Buying an OpenGL-compatible tablet now just so it can use OpenGL software in 2-4 years is a bad idea... a tablet's lifespan is measured in months, not years, before the technology makes them effectively obsolete. They are running on a 9-14 month release cycle right now. That will start to slow down eventually, but probably not within that 2-4 year time frame.
Even if we don't see much advancement on the CPU/GPU side in 2-4 years, there's a lot more to a tablet that can make it obsolete -- connectivity, input, display, battery life, etc.
That being said, I expect we see a lot more of this aimed at the Auto/Embedded industry than we do in tablets.
This is exciting news. The mobile gaming market could potentially push developers away from DirectX, resulting in more games being compatible with non-Windows operating systems. In other words, SteamOS could benefit from all of this.
Uh ....what?
First, you think that improved graphics from phones and tablets will push developers away from DirectX and towards Open GL.
Which, in turn, will improve support for SteamOS?
SteamOS is Linux based and Open GL runs horribly on Linux.
You do know that Android is based on Linux, right?
What makes OpenGL run "horribly" on Linux? Is it the OS, or the lack of support for the engine on the OS? Games made for Linux can't run on DirectX, so what is the next best thing?
My knowledge of Linux gaming is very limited, as you might have guessed. I am not aware of any engines designed to run well on Linux, but SteamOS can only succeed if their games run well. The result of having better OpenGL support on mobile Android (Linux) platforms is likely to be enhanced support for OpenGL on other platforms, including Linux and Windows.
OpenGL runs horribly on Linux? As compared to what? DirectX, which flatly does not run at all on Linux? OpenGL runs better on Linux than any other 3D API does.
The standard API for Android (and, incidentally, iOS) right now is OpenGL ES, which is a rather gimpy subset of OpenGL, roughly on par to DirectX 9.0c. Moving to the full OpenGL is a very big deal.
So, getting full Open GL support is not a big deal because the best Linux Open GL drivers is Mesa which still has compatibility and stability issues. Which is why Open GL does not run well on Linux.
Open GL will not get better until the 3rd party drivers that make Open GL compatible with Linux gets better.
What does this ultimately mean?
For your phone to run Open GL instead of Open GL ES it would have to run Open GL and the drivers. And good luck there.
AMD and Nvidia already write OpenGL drivers for Linux, just like they do for Windows. There is some dispute as to whether AMD's Linux drivers are any good, but if Linux had big market share, AMD would make it a priority and the drivers would be good, just as with Windows.
Uh ....what? First, you think that improved graphics from phones and tablets will push developers away from DirectX and towards Open GL. Which, in turn, will improve support for SteamOS? SteamOS is Linux based and Open GL runs horribly on Linux.
OpenGL runs horribly on Linux? As compared to what? DirectX, which flatly does not run at all on Linux? OpenGL runs better on Linux than any other 3D API does. The standard API for Android (and, incidentally, iOS) right now is OpenGL ES, which is a rather gimpy subset of OpenGL, roughly on par to DirectX 9.0c. Moving to the full OpenGL is a very big deal.
I think you're forgetting that for Open GL to run on Linux it needs a 3rd party driver to be compatible the X Windows System. So, getting full Open GL support is not a big deal because the best Linux Open GL drivers is Mesa which still has compatibility and stability issues. Which is why Open GL does not run well on Linux. Open GL will not get better until the 3rd party drivers that make Open GL compatible with Linux gets better. What does this ultimately mean? For your phone to run Open GL instead of Open GL ES it would have to run Open GL and the drivers. And good luck there.
AMD and Nvidia already write OpenGL drivers for Linux, just like they do for Windows. There is some dispute as to whether AMD's Linux drivers are any good, but if Linux had big market share, AMD would make it a priority and the drivers would be good, just as with Windows.
AMD drivers are not great and they are now heavily pushing their own API called Mantle. So, forget about Open GL support from AMD.
nVidia has good drivers ...when using nVidia hardware. So, if you do have a Tegra K1 device then GREAT, if not then SOL.
Uh ....what? First, you think that improved graphics from phones and tablets will push developers away from DirectX and towards Open GL. Which, in turn, will improve support for SteamOS? SteamOS is Linux based and Open GL runs horribly on Linux.
OpenGL runs horribly on Linux? As compared to what? DirectX, which flatly does not run at all on Linux? OpenGL runs better on Linux than any other 3D API does.
The standard API for Android (and, incidentally, iOS) right now is OpenGL ES, which is a rather gimpy subset of OpenGL, roughly on par to DirectX 9.0c. Moving to the full OpenGL is a very big deal.
I think you're forgetting that for Open GL to run on Linux it needs a 3rd party driver to be compatible the X Windows System.
So, getting full Open GL support is not a big deal because the best Linux Open GL drivers is Mesa which still has compatibility and stability issues. Which is why Open GL does not run well on Linux.
Open GL will not get better until the 3rd party drivers that make Open GL compatible with Linux gets better.
What does this ultimately mean?
For your phone to run Open GL instead of Open GL ES it would have to run Open GL and the drivers. And good luck there.
That's where Valve comes in. They are turning benchmarks on Linux that are better than the same games running under Windows. Someone on these forums has had the same experience running their Steam games (the ones that work) under Linux. Right now it's only Nvidia's cards that are supported, but Valve has enough customers and can push the market enough that Linux as a good gaming platform is not out of the question.
I can not remember winning or losing a single debate on the internet.
Well of course Nvidia only writes drivers for Nvidia hardware. Just like AMD writes drivers for AMD hardware, Intel writes drivers for Intel hardware, and so forth.
AMD pushing Mantle has about as much to do with AMD's OpenGL support as Nvidia pushing PhysX had to do with Nvidia's OpenGL support: nothing.
AMD pushing Mantle has about as much to do with AMD's OpenGL support as Nvidia pushing PhysX had to do with Nvidia's OpenGL support: nothing.
Its like Im talking my cat here.
AMDs drivers for Open GL work fine nVidia hardware and nVidia's drivers do work AMD hardware just not as well as nVidia because of how they coded it. Why are you even bring hardware in to the discussion about an open source graphics API. It makes no sense. Open GL works with all hardware no matter who made the drivers for it.
Also, why would AMD take the time and money to support someone elses graphics API when they want people using their graphics API. And why are you comparing it to PhysX? Mantle isnt some kind of middleware, its a graphics API. It does the same thing as Open GL.
Seriously, do you even know what Mantle is?
When you start making some sense then talk to me again.
Mantle isn't a 3D API.
Mantle is an API for using GCN hardware. GCN hardware works very well for 3D Graphics, but it can do a lot of other things as well.
PhysX isn't the best analogy for Mantle - GLIDE would probably be the best thing to compare it to. GLIDE was proprietary to 3dfx, and just acted as an API for developers to be able to easily access the hardware functions built into the Voodoo line of graphics chips.
Now, Voodoo was particular to 3D graphics, and had very little other functionality, whereas GCN does have the capability to do a good bit more than just 3D graphics (OpenCL, for instance) and Mantle just allows generic access into the GCN hardware, and while it's written with gaming in mind, it's not exclusive to 3D graphics.
As far as AMD drivers for OpenGL working with nVidia hardware (and vice versa).... I don't know what planet your from, but that's 100% incorrect. OpenGL is an API, it just defines a function name, particular inputs, and what the output should be. It's up to the vendor to decide what to put in the middle of all of that - be it specific hardware calls, some mixture of hardware + generic CPU calls, or just software emulate the entire thing because you can't do it in hardware at all.... AMD and nVidia will use the same function calls, and should produce the same outputs, but that's because they are both implementing the OpenGL standard... how they chose to implement that is very proprietary, and while there do exist software drivers that are hardware agnostic (by virtue of doing everything via software emulation and therefore fairly slowly), hardware vendor drivers are absolutely not cross-compatible.
Maybe you should talk to your cat some more, it obviously has a better understanding.
No, AMD's video drivers don't work with Nvidia hardware, nor vice versa. OpenGL can't do anything without video drivers. OpenGL gives developers a way to say, I want this behavior. It is hardware-agnostic, however, so it doesn't specify which transistors are supposed to do what. For that matter, you can write a fully OpenGL-compliant software renderer that doesn't even touch a GPU but does all computations on the CPU. (This is slow, but you can do it.) You need drivers to translate OpenGL commands into GPU behavior, and those drivers are written by the GPU vendor.
The reason AMD writes video drivers to support OpenGL is the same reason why they write drivers to support DirectX and other APIs. Without drivers, a video card is useless except perhaps as a doorstop. AMD is in the business of selling hardware, and no one will buy AMD video cards if they can't run anything. So AMD writes drivers to make sure that their cards can run lots of programs (both games and otherwise), so that people looking to play games will have reasons to buy AMD hardware. But AMD won't write drivers for Nvidia hardware, as if no one bought Nvidia hardware, that wouldn't trouble AMD.
As for exactly what Mantle is, I'd contend that we don't entirely know yet, since it hasn't been publicly released. If you want to know what OpenGL is in all its gory detail, you can read the full specification of exactly what it does, published by the Khronos group and probably mirrored elsewhere in a number of places. For Mantle, you can't do that, so for now, we only have descriptions put out by AMD's marketing department.
"Mantle is AMD's 3D graphics low-level API created to replace Direct3D and OpenGL"
/facepalm
There Is Always Hope!