Nvidia deside to abadon opensource, I deside to abadon Nvidia.

[FunkyRider]

Here's two more: Radeon X300SE, Radeon X600 128MB, also fsck the one minute change limit

[FunkyRider]

Was really tempted to buy a Radeon 5870, but the mediocre Fglrx and 2 years to wait OSS really can't convince me to spend $500 on a card I can't utilize.

[pingufunkybeat]

The things I said are my personal experience over several generations of ATI cards on Linux, I believe instead of blindly loving ATI OSS drivers like you, I have a more neutral stand point to point out the good and bad of ATI's Linux situation.

I believe that you have a good understanding of ATi's OSS support 2 years ago.

Like people say, things have changed a lot recently. In exactly one year, the drivers went from basic 2D and nothing else on recent hardware, to dynamic power management, OpenGL 2.1 acceleration, kernel modesetting and DRI2. This is a huge development, and if you're clinging on to your experience from several generations ago, you're missing out.

At the end of the day, ATi has provided me with a great desktop experience using a modern fanless card and a fully open source stack. Not a 1337 gaming experience, but it's great for everything I do, including the games I play from time to time.

[FunkyRider]

LOL YES! do it!.. please ! LOL do it! if you buy the card send this informaton to bridgman ppplleeaaaaasseee.. LOL!

thats so funny the people (yes me to) buy expensive hardware(openGL4) only to run the opensource driver (openGL2)

If my hardware can help OSS driver development I would certainly be happy to help. Is there anything I can provide to d.bridgman? i.e. dumps, certain command output and so on?

Yeah, I spend heaps on hardware, I run those gears not because I need them, but because ... you know what I mean for a hardware enthusiast. I only bought that $2500 card after I made sure that OSS driver can handle 3D acceleration on RV770 GPU. However, enabling KMS will resulting the DVI port to lose signal, I have to use Display Port to DVI adapter to get a signal in KMS mode. UMS works for all ports. This is a driver bug IMO

[pingufunkybeat]

Yeah, I spend heaps on hardware, I run those gears not because I need them, but because ... you know what I mean for a hardware enthusiast. I only bought that $2500 card after I made sure that OSS driver can handle 3D acceleration on RV770 GPU. However, enabling KMS will resulting the DVI port to lose signal, I have to use Display Port to DVI adapter to get a signal in KMS mode. UMS works for all ports. This is a driver bug IMO

That certainly is a bug and it should be reported.

[agd5f]

It won't be a 2 year wait for evergreen 3D. Likely you'll see support later in the summer. Richard is already working on it and we have documentation in the pipeline as well. Most of the previous delay was due to upgrading the Linux graphics stack in general and catching up on previous generations. The rest of the evergreen stuff is already on par with older generations (modesetting, interrupts, power management, etc.).

[crazycheese]

Nice step you did
I have one question though, I am running similar hardware and am keen to know what the bleeding edge software supports right now!
You are running the newest kernel, 3ddriver and mesa 7.9 devel?
How many frames do you get in Openarena at, say 1024x800 ?
Does 7.9 support Opengl 2.1 and GLSL already?
And have you ever tried opencl?
THXalot for desired info

Ok, here you go!
For ubuntu, I think you need xorg-edgers ppa and be sure to read their forums to set up. However its easier than on gentoo(but gentoo gives you the choice).
For gentoo you need following(~ means testing keyword, either perpackage or global):
~xorg-server,
~xorg-drivers, with radeon driver enabled via make.conf->VIDEO_CARDS="radeon vesa v4l2 v4l" (vesa is good for fallback and radeon is required, rest unimportant.)
~libdrm
~mesa
~radeon-ucode (only for R600-800), those cards will not start without it.

Your kernel should be compiled with fbcon(frame buffer on console, allows tty1-6 switching and drawing) and radeon support. Also say Y to firmware blobs in kernel(ucode).
It is recommended to use initrd, but I managed to run without it.

Ush, thats it, no xorg.conf, starts up correctly.
My versions(current, Im not using GIT):
media-libs/mesa ~7.8.1
x11-base/xorg-server ~1.8.0
x11-libs/libdrm ~2.4.20
sys-kernel/gentoo-sources ~2.6.33-r1

However you can switch to 9999(ie git) packages, as descr. in this thread: https://forums.gentoo.org/viewtopic-...deonucode.html

I decided not to do this, because it puts curtain "filter" in the face of guys that update ebuild on gentoo; I hardly imagine someone would pull next version if its buggy. And if mesa-dev(or anything -dev/9999) breaks for 15 minutes and you upgrade world withing this time frame, you will have one system down(at least on xorg). I think mesa 7.8.1 is pretty usable for this price.

According to dmesg, Im running KMS out of the box.

On Athlon II X4 630(ondemand governor,stock), 2x2GB DDR3-1600, HD4770(stock), 8 bots, I get 40-60 fps on 1920x1080, 32bit, maxed textures, maxed geometry, bloom, anisotrophy 8x. I dont know how much is ignored, but I tried to switch to lowest quality at 1920 resolution and framerate didnot improve.

This means hardware is not running fully utilized, but there is some bottleneck, that keeps everything at this level.

It is great result for opensource solution(if you skip the tiny firmware blobs to boot the chip), but average/mediocre in terms of a driver. With 9800gt I was able to get near 120 fps almost constantly. But 2D, even in composite, is FAAST.

Now on AlienArena, where 9800gt got 90fps I get between 10 and 15 with minor artifacts(mostly textures). Unplayable because of spee, but shows very good progress.

I stay with AMD, this is great performance for drivers in testing phase! Keep it on AMD!!

[crazycheese]

Well I voted for yes, however the problem is AMD's implementation won't be ready for another 2-3 years. As you know PC hardware fades like crazy, a brand spanking flagship product will be a mediocre performer within a year. With Nvidia you get the benefit of performance and functionality instantly, while with AMD you have to wait really long, long to the point when the hardware is outdated to useless piece of junk Linux will start to pick function and speed up by then (r300g anyone? who uses an X800 now?). This makes their hardware virtually useless unless you do dual boot to Windows to play games.

I still hold R100, aka Radeon 64DDR, aka Radeon 1, in my hands
I noted the opensource driver progress went all up for AMD recently but keeping constantly, where Nvidia suprisingly stopped any opensource effort, hence I switched.

In terms of perfomance, it is partially true of what you say about flagmanship. Nvidia and AMD have cards that spread both in generations AND classes. And to my calculations enthusiast class keeps perfomance compared to budget for at least 3 generations. How much times got 8800 rebranded on nvidia and still holds the perfomance. 5 years are a big frame. Besides Gallium3D, which AMD and Intel develop, is supposed to make developing drivers much faster and more consistent.

Nvidia is a joke and corner situation. It is windows driver, no matter what they say about "crossplatfoms". It is a windows driver with interfaces remapped to posix. It always(take any test) lags behind windows driver, take any version and compare. The only times where linux outperforms windows is thanks to more efficient kernel.

Supporting nvidia is supporting a company that completely ignores any improvements in linux, ignores opensource and just ports windows stuff to linux. This way you will always need windows box for games, always.

Linux needs games like AlienArena, UrbanTerror, Nexuiz, TORCS, Glest that crunch 3D performance and request for more, request for drivers to be improved. There is supertux and opentyrian, alongside with games from Kenta Cho, if you have low performance card. They are fun to play too.

The thing is, AMD gives the chance to linux openGL to outperform windows in bare GFX hardware as well as supporting opensource operating system. In my opinion the operating system is bare minimum that SHOULD be opensource thing. Userspace applications may go this or that way, depending on build model. Although I prefer opensource. What does nvidia do?

Accelerating horse is much better than a pony, bound to "windows" horse, always lagging behind, always destined 2nd.

[crazycheese]

The problem is, comparing fglrx to nvidia binary blob is not apple to apple, but more like comparing a rotten stinky apple corpse to a fresh juicy apple. :P Please don't use fglrx as the response to 'benefit of performance and functionality' because it got none of them.

Nvidia is not fresh juicy apple.

Opensource AMD is fresh juicy apple.

Nvidia blob is second hand banana, that is colored green and reshaped to look like an apple.

[crazycheese]

NVIDIA shares driver code between the Windows and Linux drivers, but that doesn't make it "use the Windows driver on Linux" or "remap" anything, it just means that common parts of the code are shared. Why would they spend all the time and effort to make two completely different drivers when they can use one everywhere?

Nvidia is windows driver, with IO interfaces made remappable.
Let me show you!
Open this linux driver in one tab and this windows driver in another tab.

Lets compare!
Release date:
2010.04.28 - 2010.04.13 *Linux crew had need some time to patch it to linux.

Weight:
40Mb - 110Mb * Is DirectX roughly size of OpenGL? Windows gets both, linux gets OpenGL only, so you can notice where 60 megs go?

New Supported chipsets(release highlights):
480, 470 - 480, 470

Supported Products:
newest...6xxx - newest...6xxx

Now read the EULA and especially the special Linux exception.
This will tell you, that:
-) Linux drivers lag behind windows drivers(windows ones are used as base)
-) Linux drivers share same code with windows minus directX related stuff.

Lets go further!

http://www.phoronix.com/scan.php?page=article&item=nvidia_qa_linux&num=2

"The NVIDIA kernel module is nearly all cross-platform code; the majority of the same code for this component runs on Windows XP, Windows Vista, Windows Win7, MacOSX, Solaris, FreeBSD, and Linux. In fact, the exact same nv-kernel.o binary file can be used on all three of Solaris, FreeBSD, and Linux. The only bits that are Linux-specific are provided as source code in the kernel interface layer of the NVIDIA kernel module."

"The VDPAU driver leverages a lot of the low-level video decoding algorithms from the Windows video driver."

So now tell me, for truth hidden behind NV doors, which is more appiable:
a) Nvidia develops crossplatform code and then ports it to windows and other UNICes(linux,macos,bsd etc)

b) Nvidia develops windows driver with IO shifted to one point. Then it releases the windows driver for windows users and linux crew. Linux crew remaps IO zone to linux and fixes some troubles/workarounds.

You have 80% windows blob running in your linux box, great way to support opensource system, if you ask me.