Slashdot Mirror

I was going to mention physics processing units, but I really don't think they will take off. Not in the way graphics cards have. Physics just isn't as sexy.

That said, I think that some sort of physical calculation support will end up in the CPU. Given that the GPU is basically a massively parallel computer, and that people have already done physics simulations on GPUs, I predict that the GPU on CPU technology will be used for physics more than any external physics card. It's not quite the same as folding the technology from an external device into the CPU, it's a case of reusing existing CPU grunt for a new purpose.

Use the "Legacy GPU" drivers. GeForce 256 is still supported.
http://www.nvidia.com/object/unix.html

Silicon Graphics 'graphics' engineers are now nVidia.

Commodity PC hardware ain't gonna cut it.

http://www.s3graphics.com/en/index.jsp
http://www.matrox.com/
http://www.tridentmicro.com/

have died at the hands of
http://www.leadtek.com/ (foxconn)
http://www.nvidia.com/
http://www.ati.com/

SGI's fu is weak besides..

Also, in Slackware there is no official Nvidia package. It's still trivial though. Just fire up the Links browser, point it to http://www.nvidia.com/drivers and download the latest Nvidia linux driver. Su to root, run the driver package, let it build the kernel addon for you, and let it modify your xorg.conf. If you've already toyed with your xorg.conf it might not modify the file for you; just manually change "nv" to "nvidia". Start up X and you should see the Nvidia logo. If so, you're all good. This has worked for me for every version of Slackware from 9.0 up.

Again, nVidia the way to go, as they support both FreeBSD and Solaris x86 with excellent and easy-to-install drivers. I play Doom 3 on FreeBSD without as much as a glitch...

There were no GPUs in the mid 90s, the Voodoo/Voodoo 2 didn't have GPUs, they were mere rasterizers. The first GPU was nvidia with the GeForce 256, Aug 31 1999. It took a long while for it to really catch on, since it's just as easy to do the 3d to 2d conversion in memory (CPU).

The 3D engine used by Quake and Quake 2 was pure software, the CPU did all the heavy lifting geometry wise (and still does, for the most part). AFAIK, the 3d geometry part of it is still mainly CPU based, you can't just send every polygon in the world up to the GPU and expect it to sort the shit out in a timely fashion. BSP trees and face culling and all kindsa nifty hacks abound for such things.

We had no fancy hardware T&L business or programmable pixel shaders, and that's how we liked it.

I remember walking uphill 40 miles in the snow just to frag newbies with my nailgun.

While MS may have stopped supporting win98 in terms of patches etc, the industry stopped supporting it a long time ago. MS also stopped supporting it for much of their office products quite a while back. Even hardware such as printers have been not supporting the old OS in the last few years.

    Referring to the use as a gaming OS:

http://www.nvidia.com/object/win9x_81.98.html

    Granted, 2k/XP has a newer driver, but this one is still more recent than several years..

http://www.microsoft.com/Windows/DirectX/productin fo/sysreqs/default.mspx
http://www.nvidia.com/object/win9x_81.98.html

Apparently.

Not to mention that most scientific applications run mostly under *nix like Linux or BSD, for which GFX driver support isn't always incredible, specially for recent models,

This is nothing but FUD, nVidia has always had great support for *nix OS's http://www.nvidia.com/object/unix.html even ATI has some support for *nix OS's https://support.ati.com/ics/support/default.asp?de ptID=894&task=knowledge&folderID=27 try using them sometime rather than listening to your tech gods at the local geek squad....

On the other hand, good luck finding the touch pad driver, random wireless card driver, or any of the other numerous literally non-standardized widgets on a keyboard. The only things that are probably truly standard on a laptop are the keyboard and screen (because screens need to be standardized, luckily).

Not to mention, that every person here saying "run it on a desktop" means a business-class desktop and not some gaming rig that has more tweaks than there are bugs on the planet. Everyone that really knows how to install an OS would still probably tell you that it's much easier to get these gaming rigs (with a few exceptions) running than the laptops for most beta, or weird distributions out there.

In conjunction with this, nVidia are also today releasing their new Series 90 of drivers for Windows, the biggest visible change is a new configuration panel interface.

Have any of you guys heard of MXM? http://www.nvidia.com/page/mxm.html

While I agree for most people installing these drivers is easier under Windows, that is not because the install procedure is easier or faster, but simply because they are accostumed to doing things this way.

No, the install procedure on Windows really is easier and faster.
- go to ati.com or nvidia.com or whatever
- find and download the driver--usually an .exe
- double-click the .exe
- click 'next' a few times
- reboot
- pretty much always works

Compare that to pages like this and this.

And while it's true that very often we need to compile and/or load some modules in most linux distros for these to work, at least they will work.

Really?

this is my understanding; correct me where I'm wrong as I have not dealt with any of this software personally, this just comes from what I've picked up by reading about it.

The drivers you get from Nvidia/ATI (for simplicity I'll just consider the Nvidia case and assume the ATI case is legally identical) consist of (1) the actual binary driver and (2) the source code to a kernel module that you need to build/link with the kernel, BOTH distributed under the same non-GPL-compatible license seen here http://www.nvidia.com/object/nv_swlicense.html

Kororaa (henceforth "the project") is distributing (1) the same binary driver and (3) a COMPILED binary of the kernel module

Nvidia is the sole copyright holder for (1). The project needs Nvidia's permission to distribute it, which is granted in section 2.1.2 of the previously linked license.

(3) is a derivative work of both the Linux kernel (because of included headers) and (2). The project needs permission from Nvidia and the copyright holders of the Linux kernel of them in order to distribute it.

If (3) is distributed under a GPL-compatible license, the copyright holders of the Linux kernel grant their permission via the GPL.

In order for (3) to be distributed under a GPL-compatible license, Nvidia's license for (2) needs to be GPL compatible for the project to avoid infringing Nvidia's copyright for (2).

Nvidia's license for (2) is NOT GPL-compatible because it does not allow for the preparation of derivative works among other reasons.

Therefore (3) may not be distributed under a GPL-compatible license.

Therefore permission is not granted by the copyright holders of the Linux kernel to prepare a derivative work and the project's distribution of such a work (3) is copyright infringement.

Nvidia's distribution of ONLY (1) and (2) is not copyright infringement because they are the sole copyright holders of (1) and (2).

IMO, the true blame for this rests with ATI and NVIDIA for not releasing the drivers open source in the first place.

Then Nvidia, Mandriva, SUSE, and others, are in the wrong.

The issue appears to be the distribution of the Linux Kernel and nVidia's/ATI binary drivers together. When the binary drivers are built the GPL'ed code is included in the binary result, which is a violation.

From the nVidia Software License:

2.1.2 Linux/FreeBSD Exception. Notwithstanding the foregoing terms of Section 2.1.1, SOFTWARE designed exclusively for use on the Linux or FreeBSD operating systems, or other operating systems derived from the source code to these operating systems, may be copied and redistributed, provided that the binary files thereof are not modified in any way (except for unzipping of compressed files).

Maybe Koraraa could modify their Live CD such that it only includes the source code for the nVidia drivers, this would certainly be legal under the terms of nVidia's software license. Putting the source code for the kernel module on the same CD as other GPL'ed software is certainly no violation of the GPL or nVidia license. Upon boot you could just automatically compile the kernel module from the source and insert the module. The binary kernel module is not being distributed, therefore there is no violation of the GPL.

Just for clarification (I had to look this up myself), the nVidia Linux drivers have two parts: a closed source, binary only, platform independent component; and a kernel module distruted as (non open) source code which acts as an interface layer to the linux kernel. The binary only portion contains no GPL code and can certainly be placed on the same CD as GPL code. The kernel module which would contain the product of compiling GPL and non GPL code would not be distributed on the CD, it would only be compiled on the user's computer upon install.

Would this solution work?

Perhaps the first point of call is that the nVidia license permits the re-distribution of their driver in Linux distributions. Surely if it was a GPL violation they are not allowed to do this, or perhaps it's just down to the distro maintainer to see if the GPL overrides other package licenses.

I send emails to both ATI and nVidia querying this license issue. ATI didn't even acknowledge my email however nVidia (to their credit) did reply (thank you!). Here is the nVidia license, as quoted in an email from nVidia:

        The NVIDIA Software License:

        http://www.nvidia.com/object/nv_swlicense.html

        permits redistribution of the NVIDIA Linux Graphics driver like you are doing:

        "2.1.2 Linux/FreeBSD Exception. Notwithstanding the foregoing terms
        of Section 2.1.1, SOFTWARE designed exclusively for use on the Linux
        or FreeBSD operating systems, or other operating systems derived
        from the source code to these operating systems, may be copied
        and redistributed, provided that the binary files thereof are not
        modified in any way (except for unzipping of compressed files)."

How is the hardware support?

Why don't you check for yourself?

NVidia releases native FreeBSD x86 drivers (no amd64 unfortunately), ATI support blows (what else is new :), 'though a freebsd port of the ATI linux driver is being written.

Had you continued to read the article you would have read that the chipsets were STILL SUPPORTED, just not under the unified driver.

The Riva TNT is what, 7 years old?

I would hope that code would have worked its way out of the unified driver.

Had you continued to read the article you would have read that the chipsets were STILL SUPPORTED, just not under the unified driver.

The Riva TNT is what, 7 years old?

I would hope that code would have worked its way out of the unified driver.

Had you continued to read the article you would have read that the chipsets were STILL SUPPORTED, just not under the unified driver.

The Riva TNT is what, 7 years old?

I would hope that code would have worked its way out of the unified driver.

What is the output of your Xorg.0.log?

For those users, something along the lines of

(EE) Time to buy new hardware whether you wanted to or not!

Apple finally broke the chicken and egg problem with their 30" Apple Cinema display. They built dual link into their entire product line in preparation for it's launch.

Jobs also introduced a new 30-inch display. The $3,299 display sports 2560 x 1600 pixel resolution, and works only in the Power Mac G5. It requires a new Nvidia GeForce graphics card in order to work, a $599 card that features dual-link DVI interfaces.

...and finally the latest generation of ultra-high end video cards now mostly support dual link.

From what I understand this new standard will be incapable of driving monitors at resolutions above what these 30" displays can do now. That's nice but DVI is there and prepared to surpass that. Why create a new standard that limits display size to a resolution that was reached a year before the standard is even released, especially when dual link support is finally taking hold and the original limitations of DVI are starting to melt away.

DisplayPort, LVDS, DVI, and HDMI comparison

yes, and with newer GPUs, the program size has increased dramatically, which makes them much more versatile. 3 years ago I was cramming a vertex program into 256 lines - now I've got 65535. Fragment programs increased similarly (though I just finally got a card that supports them in the last 3 months to play Oblivion, so I'm still learning the ropes).

Take a look at the GPU based samples (unfortunately, most require Windows) - many are incorporating physics (water, cloth, etc). Another good source is http://www.gpgpu.org/

Unfortunately, I don't know of any open source GPU based physics engines, which sucks, and IIRC, a bunch of patents have been filed on some of the software based solutions.

http://www.ati.com/products/RadeonX1900/specs.html

"Full speed 128-bit floating point processing for all shader operations"
"64-bit floating point HDR rendering supported throughout the pipeline"

http://www.nvidia.com/object/7_series_techspecs.ht ml

"Full 128-bit studio-quality floating point precision through the entire rendering pipeline"

Nvidia drivers windows only??? I suppose it was the magical elves that live in the walls of nvidia's offices that wrote these.

Does NVidia have drivers that work on Sparc Linux? Solaris x86? PPC Linux? The answer is no, and I never expect them to - which IS the problem.

Incorrect about Solaris x86 - I have used those drivers and they work very well. Check it out here. However, the point you are making is valid; if NVidia does not support a platform that you use, you are SOL.

Music: Foobar2000 0.8.3 (iTunes and dumbed down fb2k annoy me)

Video: Media Player Classic with ffdshow

If you are using an iMac or MacBook Pro, then you might be interested in the Windows-only software that enables the ATI Radeon 1600's GPU-accelerated H.264 playback and video transcoding. For GPU-accelerated H.264, I think you need to purchase CyberLink's H.264 decoder. ATI's Avivo Video Converter is integrated into the latest Catalyst Control Center, which I'm not sure is included on Apple's Windows driver disc image.

Does anybody know if GPU-accelerated H.264 playback and video transcoding is enabled on OS X yet?

My point was why would Apple throw in a DirectX card into a Mac, when its(DirectX's) sole use will be on Windows.

Even when negotiating with card vendors, wouldn't it be cheaper to get a custom graphics card with all the DirectX circuitry taken out? Why waste transistors on capability you will never use.

This might just provide laptops with enough power to run Aero Glass.

Aero Glass requires a DirectX 9 class GPU that supports Windows Display Driver Model (WDDM). Low-end mobile GPUs that meet this requirement include GeForceFX Go5100, Mobility Radeon 9500/X300, and Intel GMA 950. Even GMA 900 (which a lot of current Centrino users have) should work if they write WDDM drivers for it, but I doubt they will.

Here's some links for those who want to see the Aero Glass mobile GPU requirements:

• NVIDIA GPUs "Windows Vista Ready"
• ATI Vista Ready GPUs
• Current GPU Guidelines for Windows Vista Capable PCs

BTW, the "Vista Basic user experience" (formerly known as "Aero Basic") does not look like Windows XP (the GPU requirements will be similar to XP). In fact, I think many users will prefer this interface to Aero Glass. Here's some screenshots:

Windows Vista February CTP Screenshot Gallery 8: Windows Vista Home Basic

Start with Fedora Core 4.
Browse to: http://nvidia.com/drivers
Select linux IA32/AMD64 as your platform.

Download, make executable, run. Should automagically set up everything for you.

It worked fine for me...

It's difficult to read through the marketing words on the tech sheet, but they claim 128-bit FP precision through the entire pipeline. That figure, however, is from using 32-bit floating point per color channel (RGB and alpha) to represent colors during the rendering process. I can't seem to find any details on how a dedicated graphics pipeline is being adapted to service physics calculations. I'm going to go out on a limb and say it's all happening on the vector units in the T&L engine and not involving the rasterization part very much at all.

FYI, SLI in the nVidia context is 'scalable link interface'. 3dfx did the same thing (Well, similar. Really more like ATI's Crossfire) and used the same acronym, but expanded it to 'scan line interlace'.

i'll be dead before using a gamer chipset for serious usage

what you don't realise is that it's not your regular NForce4 gamer's chipset. nVidia has a separate professional line, see here to which this one (nForce pro 2200) belongs.

*Rolls eyes*

Your graphics driver doesn't yet support the GLX_EXT_texture_from_pixmap extension so every time that call is made (near enough any time the window changes) it is going to be emulated in software. As you do that opaque resize just think how many times it might be being called... Don't worry though - your 3D chipset vendor says they will implement it in their next driver release.

However at the end of the day what is acceptable to me could well not be acceptable to you - just two months ago everyone was complaining window dragging was too slow whereas I thought it was acceptable. On Xgl it seems quite fast to me but now we're on to the next of the never ending issues. I guess since you want something finished yesterday you had better start saving for / using that Mac...

just provide the drivers... the community will deal with the rest...

Dell doesn't make the hardware, they integrate it and support Windows users that don't want to deal with the "community." The hardware they use (Intel chipsets/wireless, ATI/NVIDIA graphics, etc.) have Linux drivers available from the hardware manufacturers. The rest are already provided with Linux.

The "community" complains too much.

In this paper, we make the case for using the existing XFree86/X.Org DDX loadable driver framework to achieve a production-quality composited X desktop, as opposed to the X-on-OpenGL model. While the X-on-OpenGL model demonstrates what the graphics hardware is capable of, everything that the X-on-OpenGL model can achieve is equally possible with the current framework. Furthermore, the current framework offers flexibility to driver developers to expose vendor-specific features that may not be possible through the X-on-OpenGL model.

"XGL is a different X server. This is a more incremental change which is slated to become part of Xorg. We don't believe that replacing the entire X server is the right path, and that improving it incrementally is a better way to modernize it. After talking to people at xdevconf, it felt like much of the upstream Xorg community shares this view. You can search Adam Jackson's notes for "large work for Xgl" to get the blow-by-blow or NVidia's presentation from XDevConf 2006 on using the existing model.

We've been working on the AIGLX code for a some time with the community, which is in direct contrast with the way that XGL was developed. XGL spent the last few months of its development behind closed doors and was dropped on the community as a finished solution. Unfortunately, it wasn't peer reviewed during its development process, and its architecture doesn't sit well with a lot of people.

The other question is Wait, can I use compiz? The answer there is a theoretical yes, although no one has actually gotten it to work. We love compiz and we think it's great stuff and is well polished, but it's often confused with the underlying architecture of XGL. Much like the code that we've added to metacity, compiz is a composite manager. With a bit of work, it should be possible to get compiz working on this X server. There's an excellent post from Soren on the topic of compiz vs. metacity."

Out of curiosity, do you have a geforce2go, or a geforce4 2 go?

The 4 to go has the nfiniteFX II engine which has pixel shaders.

Ref: http://www.nvidia.com/page/geforce4go.html

Like I said before, I disagree in theory. However, in practice you're right. The cards would have to have a PCI express-like card-to-card bus to bypass the AGP return data rate limitation (and being handed off across independent busses), but I believe that existing PCI express cards are using the card-to-card link for almost the same purpose (to keep from passing data back through the PCI express switch). In the case of a PCI express card, it's probably just a matter of dedicating a portion of the lanes supported by the GPU to the link. In the case of an AGP card, it would require a wholly different type of bus tacked on in addition to the AGP circuitry, which would make it cost prohibitive and/or impractical.

BTW: According to an NVIDIA GPGPU presentation the AGP return data path in an AGP 8X bus runs at 1/3rd the data rate of the send data path (page 51 of 52). These same people claim that AGP 8X transfers 2.1GB/sec, which means that the AGP 8X return path is around 700MB/sec, which is slightly slower than the 64 bit 133 MHz PCI-X bus (~1064MB/sec). When the graphics people claim that the bus crawls, just remember that their view of bus speeds is jaded in comparison to most.

http://www.nvidia.com/object/IO_21673.html
COMPUTEX 2005--TAIPEI, TAIWAN--MAY 31, 2005
NVIDIA GeForce Go 6 Series and Mobile PCI Express Module (MXM) brings state-of-the-art NVIDIA GeForce 6 Series performance to notebook platforms with a consistent graphics interface enabling interchangeable graphics cards in notebook PCs.

They have had upgradable video cards for laptops for a while now. All the Dell M170/Insp90000/XPS have them, among others. You just limited yourself by cheaping out on a budget laptop.

You get what you pay for, and you don't get what you don't pay for... in your case: upgradeability.

The latest integrated chipset from Nvidia seem to support Shader Model 3.0, which I believe is a requirement for Aero Glass.

I ordered myself an Asus board recently with GeForce6150 + nForce430 onboard, infact it should be arriving tommorow. Will be fun to finally test some Vista beta's but to be honest I didn't have Aero Glass in mind when I made my purchase, and if it doesn't work then I won't give it a second thought.

Heh, I just noticed my choice seems to top of the line integrated from nVidia. That pleases me somewhat, since i'm not a gamer integrated graphics do me fine.

The latest integrated chipset from Nvidia seem to support Shader Model 3.0, which I believe is a requirement for Aero Glass.

I ordered myself an Asus board recently with GeForce6150 + nForce430 onboard, infact it should be arriving tommorow. Will be fun to finally test some Vista beta's but to be honest I didn't have Aero Glass in mind when I made my purchase, and if it doesn't work then I won't give it a second thought.

Heh, I just noticed my choice seems to top of the line integrated from nVidia. That pleases me somewhat, since i'm not a gamer integrated graphics do me fine.

Yes, but that creates other problems:

http://download.nvidia.com/XFree86/Linux-x86/1.0-8 178/README/appendix-s.html

You mean like video drivers (http://www.nvidia.com/object/unix.html) for Linux maybe?

The people you know with 64 chips are idiots then. If you own a copy of Windows XP 64, you should have no problem finding 64 bit drivers for most modern hardware, with the exception of some older HP printers:

http://www.nvidia.com/object/winxp64_81.98.html

http://www.nvidia.com/object/nforce_nf4_winxp64_am d_6.69.html

https://support.ati.com/ics/support/default.asp?de ptID=894&task=knowledge&folderID=367

http://h10010.www1.hp.com/wwpc/pscmisc/vac/us/en/s m/network_software/universalprintdriver_overview.h tml

And for the uneducated and/or ignorant, the 64 bit version of windows STILL RUNS 32 BIT APPS JUST FINE. The only time an app may not function is if it communicates to a hardware device like a printer or CD burner. Which, by the way, CD burning apps like Nero 6 & 7 support just fine. Also, McAfee has a 64 bit verison of their antivirus software, which is included in the SAME installer as the 32 bit version, so no need for 2 seperate installers. I won't even got into detail about all the 64 bit linux distributions which have been out for several years now.

If you have a 64 bit chip and you couldn't find drivers for modern based hardware, you didn't try hard enough, or at all for that matter. Windows Vista will be 64 bit by default, yet include a legacy 32 bit version as well that won't have all the features of the 64 bit version, such as requiring signed drivers. All you have to do is google for the info, and you will find it.

The people you know with 64 chips are idiots then. If you own a copy of Windows XP 64, you should have no problem finding 64 bit drivers for most modern hardware, with the exception of some older HP printers:

http://www.nvidia.com/object/winxp64_81.98.html

http://www.nvidia.com/object/nforce_nf4_winxp64_am d_6.69.html

https://support.ati.com/ics/support/default.asp?de ptID=894&task=knowledge&folderID=367

http://h10010.www1.hp.com/wwpc/pscmisc/vac/us/en/s m/network_software/universalprintdriver_overview.h tml

And for the uneducated and/or ignorant, the 64 bit version of windows STILL RUNS 32 BIT APPS JUST FINE. The only time an app may not function is if it communicates to a hardware device like a printer or CD burner. Which, by the way, CD burning apps like Nero 6 & 7 support just fine. Also, McAfee has a 64 bit verison of their antivirus software, which is included in the SAME installer as the 32 bit version, so no need for 2 seperate installers. I won't even got into detail about all the 64 bit linux distributions which have been out for several years now.

If you have a 64 bit chip and you couldn't find drivers for modern based hardware, you didn't try hard enough, or at all for that matter. Windows Vista will be 64 bit by default, yet include a legacy 32 bit version as well that won't have all the features of the 64 bit version, such as requiring signed drivers. All you have to do is google for the info, and you will find it.

The parent is gravely mistaken. The following NVIDIA notebook GPUs are "Windows Vista Ready:"

GeForce Go 7800 GPUs
GeForce Go 7600 GPUs
GeForce Go 7400 GPUs
GeForce Go 7300 GPUs
GeForce Go 6800 GPUs
GeForce Go 6600 GPUs
GeForce Go 6400 GPUs
GeForce Go 6200 GPUs
GeForce FX Go5700 GPUs
GeForce FX Go5650 GPUs
GeForce FX Go5600 GPUs
GeForce FX Go5200 GPUs
GeForce FX Go5100 GPUs

http://www.nvidia.com/page/technology_vista_home.h tml

If you want to run Glass (the GUI) you need to make sure you have a compatible video card. I have found in Vista that the biggest perf issues stem from low memory or not having a compatible video card. Here is nvidia's list of supported video cards, note that there are no notebook cards on it right now. Here is ATI's list of supported video cards. If you want the slick UI, just make sure you get a laptop that supports LDDM.

They are already out

Nvidia has always been good about getting Linux drivers out for desktop parts.