Slashdot Mirror

Kernel Driver
GL Files

They compile and work fine on my CVS build of the 2.4.1-XFS tree.

Er, what would it help with?

Does anyone know of any sites dedicated to using 3D gaming engines for non-gaming purposes other than the article we saw a while back about a company using Q2 engine for an architectural walk-through? You know, maybe with resources that help take Autocad files and convert them to Quake maps?

There are plenty of people who use the Unreal engine for these sort of things, like the before mentioned Unrealty and the Notre Dame virtual tour. The Unreal engine is so easy to extend and modify because of its modular and object oriented design. The engine actually recently won a price for being so powerful and making development easier. Here is a list of recent improvements to the engine and the editor. Looks really sweet! DirectX 8 + GeForce + UnrealEngine looks like a good recipe for great upcoming games.

************************************************ ** *

Perhaps before buying a power laptop like that you should've waited for the GeForce 2 Go from nVidia. It features all the full GPU support of the GeForce 2 MX (it's less powerful than the GTS, but come on, it's a laptop), features full 3D acceleration, and when you plug it into any monitor, you've got the ideal LAN party computer. They are due out imminently from Toshiba and other manufacturers as well. Go nVidia!

http://www.nvidia.com/Products/Drivers.nsf/Linux.h tml

The "motion blur" in these scenes is not actually motion blur... it's the same thing as pixel memory (remember those old phosphorous screens and my favorite xscreensaver). It buffers the previous frame then additively disolves it overtop of the new frame. (You can't even call it bluring.) It's poor and if you'll notice, is it really worth losing 40-50 frames/sec?

It's technology hype. Sure, it's kind of neato, but it's used where it doesn't apply (notice the walls get blurred? wtf?) and it only serves to muddy up the images. You wouldn't even *need* a Voodoo chipset to do this if you were willing to sacrifice enough video memory. Think about it. Gimp does this same effect real time too. :-)

NVIDIA's approach, is however, a bit more in the right direction. It uses a real blur, and maintains frame rate. I highly suggest that you check out http://www.nvidia.com/Marketing/Developer/DevRel.n sf/pages/64A26BA3A82A992188256993007AC623 if you want the real deal.

I just recently built a dual CPU linux box, and went with a GeForce 256 based card. The card works wonderfully with nvidia supplied binaries _except_ that there are issues with OpenGL (GLX) when running SMP.

Excerpt from nvidia FAQ concerning linux drivers for TNT/TNT2/GeForce/Quadro Chips:

6.5.7 OpenGL + SMP

Some of our internal testing has revealed random lockups on SMP systems when running OpenGL. We are unsure of the exact cause of this, but we will find it and fix it for our next driver release.

I've personally experienced lockups when trying out OpenGL screensavers, and I look forward to updated binaries from nvidia that will fix the problem.

I think I read on the nVidia FAQ page (which now appears to be broken, so I can't check or quote it) that the 3Dfx shareholders will have to approve the transaction before it can go through.

Of course, if 3Dfx feel that they're in such bad shape that they have to do this, then you may be worse off without the deal.

I doubt it. NVIDIA's Q&A says, "The structure of this deal allows NVIDIA to purchase certain assets that are consistent with our business model without acquiring 3dfx liabilities." One of these liabilities, apparently, is the current Voodoo line: "The remaining 3dfx entity is responsible for their current product lines and retail channel. ... The 3dfx product in the channel and installed base and customer support remain the responsibility of 3dfx. It is best to check those details with 3dfx management."

After this deal, there isn't going to be a "remaining 3dfx entity." According to the 3dfx press release, the "board of directors will recommend to its shareholders that they ... approve a plan to dissolve the company following completion of the asset sale." The creditors will lick the plate clean, any remaining employees will be fired, and Nasdaq will remove TDFX from the ticker.

i'm also very interested to see what happens with the 3dfx brand name. could nvidia keep marketing separate products under the 3dfx brand name? if so, what would be the differences between the products sold as 3dfx and those sold as nvidia. interesting indeed.

The 3Dfx name still provokes images of fast, quality video card for quake (or other 3D games), even though their latest cards may have sucked in comparison to NVidia's or ATI's.

From the NVidia Q&A:

9. What is NVIDIA's intention for the 3dfx and Voodoo brands?
We believe that the 3dfx and Voodoo brands are well known and respected throughout the industry. We have not finalized our plans for these brands at this time.

6. Are you also acquiring the board-related business?
No, the board business remains with 3dfx.

So what chips are they going to put on these boards, NVidia??? Now that would be ironic...

• Nvidia's 3dfx Q&A
• 3dfx's third quarter earnings webcast
• A letter from 3dfx founder Scott Sellers about the sale

Ahem, NVidia Linux support? Try here.

They even have BeOS support finally, here.

-Julius X

I'm just curious why Voodoo5 5500 went on the list. If we're going for true-geek, which means the fastest and best available, then why not go for the NVIDIA GeForce2 Ultra? It's much faster than the 5500 and has onboard Transform & Lighting abilities.

Not only that, but you'd be secure that you're buying from a company that will definitely support its product....I'm not sure how long 3dfx is gonna be in the biz anymore....

-Julius X

Yeah, that does suck... NVidia wanted to release specs (they even made a good start), but NDA's prevented them from doing so.

------

Even for the enormous Linux kernel module that's required to use their drivers? Really?

Yes. The only part that is Linux-dependent is the abstraction layer, for which the source code is provided. The same kernel module with a different abstraction layer is used on Windows. (If you don't believe me, head on over to that Linux dev page at nvidia -- the one with the register-level specs and such. Too bad the specs are incomplete due to NDA's...)

Does their Windows driver, after less than a week of use, bloat to consume over 200MB of virtual memory?

No one knows -- Windows itself bloats faster. :) OK, that memory leak is obviously something they are working on. It is beta software. Does it hurt so much to restart X once every few days?

they don't seem to care about keeping up with development kernels

Do you honestly expect them to?

I haven't exactly put much work into fixing the problem (but how can I, when I can't even recompile with debugging symbols?)

You have the source code for the abstraction layer in the NVidia kernel module. Any changes necessary can be made there.

Voodoo2 glide; the frame rate may not have been as fast, but the rate of driver improvment certainly was faster.

That's because NVidia's Linux driver does not require much in the way of improvements. It is pretty much complete, except for some minor bug fixes. Compare this to the Voodoo 5 driver, which was supposed to be ready a month after the release of the hardware. It is still in very poor shape (only supports one processor, no FSAA) despite having open source code.

------

1) I just downloaded the Linux version. The installer is pretty slick and I got high transfer rates, but unfortunately I had to do it 6 times before it would actually complete the install without hanging.

2) Upon installing it insists that you register for their Netcenter website and the fonts where it asks you if you'd like spam along with your registration are so small that they're barely legible. Suggestion: Just click the link says "I'm under 13 years old" to get around the mandatory registration.

3) So far this version shows no improvement over the nightly snapshots I've been downloading from Mozilla's site. Suggestion: Download a recent nightly build from mozilla.org instead if you really insist on upgrading.

4) It won't render NVidia's Linux Drivers page. Lame...

numb

Since tom'shardware is slashdotted, you might choose to while away the hours gazing earnestly at Nvidia's press release. A good sample quote:

"GeForce2 Go allows business users, artists, and gaming enthusiasts to create, present and entertain anywhere, anytime."

Remind you of a certain software giant's claim of "anytime, anywhere, and on any device"?

Did you check to see if all the previous mesa garbage is gone? Read the FAQ at Nvidia's site

Don't forget to get the XF ree 86 4.0 drivers from NVIDIA, even if you're not using 3d; they're faster for 2d stuff than the open-source drivers included with XFree86.

Don't forget to get the XF ree 86 4.0 drivers from NVIDIA, even if you're not using 3d; they're faster for 2d stuff than the open-source drivers included with XFree86.

All CSSMP does is say that there is now a mobil-phone media type, and that these rules are used for it. In a way, it's a completely different spec, but you'd need it to be separate. Lumping voice browsers, TV browsers, and console-browsers into one standard would get... messy. CSSMP just gives another set of rules for rendering HTML content.

BTW, how do you expect ISP's to rewrite Amazon.com's main page to work on a cellphone? It's kinda graphically intensive... How about Slashdot? For simple pages, it'd be easy, but for complex pages like Slashdot it would be all but impossible. (Especially pages that are designed to look a certain way, which cellphones, and lynx, usually choke on. Try nVidia's webpage under Lynx some time - it's all graphics without ALT tags.)

Actually, as it turns out, CSSMP might be exactly what you want - since a page designed with a CSS2 style sheet can have multiple media types, the cellphone section would describe how to render the page on a cellphone, while the browser sections would identify how to display it under a browser. A properly designed page would work both under a CSS2 compliant browser (there aren't any!) and under a CSS2 compliant cellphone (and... there aren't any of those, either). But it would be the same HTML document, just different styling rules for different ways of displaying the same content.

And while you're giving the W3C credit for being forward looking, realize that there is no (finished, I think Mozilla trys to) browser that currently implements CSS2 - and the W3C is currently working on CSS3 .

Troll eBay looking for either an SGI 1600SW or a Radius Artica ($1500-$1900) - both are the same device: 1600x1024 resolution, wide screen, awesome image quality, uber geek. Only two cards that I know of drive it directly; the now defunct Number Nine Revolution IV or the current 3Dlabs Oxygen VX1-1600. Looks like Xi Graphics have decent X support for both. Otherwise you can get the SGI multi-link adapter ($495) that will take analog DB-15 or DVP/DVI digital inputs and drive it that way. Though, whatever you do, get 100% digital from video card to display. DVI is the current standard in the PeeCee world with support from Matrox, nVidia, ATI on the video card side and more flat panels are coming out that have a DVI-D connector (i.e. Philips 150P) - see Tom's Hardware for a good write up.

No? And it's only 450x450x450? Geez, I'll stick to my GeForce2 Ultra and a cheap-ass ViewSonic CRT.

That made pretty good sense, right up until the word 'Glide'. OpenGL's extension mechanism allows vendors to expose groovy new features without having to wait for Microsoft to rev the API. In fact, nVidia is already doing this.

But Glide? It provides no functionality that isn't already available through DX5-7 and it only runs on Voodoo hardware. And Linux? As a replacement for a gaming console? [makes indignant sputtering noises] I think you should have stopped typing after the word 'OpenGL'.

If the 'The Evil Dwarf from Hell' had actually bothered check out the facts he would have noticed that the Geforce2 MX supports two monitors:

http://www.nvidia.com/P roducts/GeForce2MX.nsf/twinview.html

Now if the linux driver doesn't support dual monitors then that's a whole different matter.

Read about it HERE

Strange... I submitted this to Slashdot yeaterday, and still some other sites came out with it first.

Slashdot seems to be slipping a bit.

Did you install the kernel-headers rpm for 2.2.16? - you'll need it/them - thats why you're getting refrences to 2.5 - same place you got your kernel - I couldn't see any obvious traps otherwise - had a quick look through the FAQ @:
Nvidias page
--

Some people are doing it !! Look at Digital Connections. They have info on Home Theatre PC's - what equipment to use etc. as well as tested home systems.
There is also talk about tweaking the G400 into displaying HDTV signals from DVD's. I have yet to find a card that exploits this, but according to the documentation on the GeForce2, it should be able to display in all HDTV modes (480p to 1080i)as well as record HDTV !!

Contrast that to CivII CTP. This setup was a breeze. Maybe iD should team up with Loki to help them work out the distribution/setup kinks.

The agreement reads, in part:
2.1.2 Limitations.

No Reverse Engineering. Customer may not reverse engineer, decompile, or disassemble the SOFTWARE, nor attempt in any other manner to obtain the source code.

Interesting, eh?

A few weeks before the GeForce2 GTS launch, nVidia was good enough to fly me out to their headquarters in Santa Clara and check out their GeForce2 GTS before anyone else had even seen the card.

The staff at nVidia has bent over backwards to ensure that I have been treated fairly and have ensured that I have had information at my disposal to write informative and accurate reviews of their products.

If I ever wrote something that was in error or made their product look bad you can bet that nVidia would do everything in their power to protect their intellectual property.

I really can't fathom why Taco said anything about admitting wrongdoing... this non-apology is yet another reason I won't be purchasing anything from nVidia ever again. I highly encourage anyone who hasn't already to let info@nvidia.com know that you'll be doing the same. Ask them for their side of the story; you'll receive enough bullshit to snow ten strong men.

The correct spellings are as follows:

• NVIDIA
• RIVA
• TNT
• GTS
• GeForce
• Quadro

The reason is that graphics cards are designed to do just that: graphics. According to NVIDIA's developer website (check the FAQ for the GeForce), transformed data cannot be read back from the GeForce system. Even with lower systems, like the Voodoo I, there's no practical way to read data back from the card, even if you did come up with someway to encode your finished data into an image and process it later. Graphics cards are designed to go from main CPU to monitor incredibly fast; they can't go very fast in the other direction, if they can go in the other direction at all. With OpenGL one can use glReadPixels() and glReadBuffer() to read the values from any available buffer (or at least the Z-buffer and color buffer), but that's usually slow because, as I said, graphics cards weren't really designed to go the other way.

I'm sure most of us remember the days of math coprocessors like the 487 (since the 486 SX had no FPU) and even earlier ones like the 8087; it seems the hardware industry is beginning to regress to the math coprocessor days of yore with fancy names like GPU to make itself sound 31337 :) I can't say it's a bad idea; it's a very good idea, but until someone markets a graphics card that can be programmed to perform general-purpose algorithms, we're probably better off with an actual coprocessor or coprocessor board. Though, I can't really say that there are lots of coprocessor boards either; maybe a graphics-card builder can beat the coprocessor builders to the punch?

While it is perhaps not the best journalism practice to accept donations from th manufacturers, it can be the only way to cover expenses and/or the only way to get the card early. So, sites of all sizes do this, from Tom's Hardware to much smaller sites. The result is that nVidia can use these tactics to manipulate sites. It seems to me that the answer is for us, as concerned consumers, to write to nvidia and tell them we don't like their tactics. Even if you think these tactics are inevitable, I don't think anyone would say they are good. So, write to them here to tell them exactly what you think. My letter will go something like: As a technically savvy consumer who reads web sites for reviews of products, I am deeply disappointed with your tactics in dealing with these sites [yadda yadda talk about specifics some]. As a result, my next graphics card purchase will be heavily influenced [yadda yadda]. Not only will my purchase be affected, but, because of the fact that I pay attention to these things and reccomend hardware to friends, theirs will be too. I hope that you decide to change your PR tactics so as to promote good, sound journalism where your cards can shine solely on their merits and not on paid reviews. I know you are capable of producing great cards, lets see your ethics meet the same standards. [signature, etc].

Okay, it's time this NVidia thing was more fully explained. In fact, I daresay the NVidia situation would make an excellent Slashdot feature, assuming we could get anyone at NVidia to go on record. I, sadly, must remain off-record, since we're trying to get this very information out of NVidia so we can support their chips, and I wouldn't want to upset them. However, I also think a more complete picture of the situation -- at least, as I have come to understand it -- may help Slashdotters and Open Source advocates understand the situation more fully and allow youse guys to direct your intellect and zeal in a more constructive manner.

First of all, the Big Book Of NVidia Register Definitions won't tell you a thing. NVidia's design philosophy is very unlike any other graphics chip. Everything, including rendering context maintenance, is dealt with at the hardware level. In reality, a lot of duties get thunked off to software via interrupts, but by taking this approach, NVidia can migrate functionality into and out of the hardware at will, and all client software will continue to work (that's the theory, anyway). Thus, to write an effective driver, you need considerably more information than a register list since, in actual fact, half the time you aren't dealing with real registers.

I have it on reliable authority that this information isn't really written down anywhere; unlike Intel, NVidia does not have a technical documentation division. The design philosophy remains primarily in the engineers' brains. If you want to become a customer of NVidia (that is, a board OEM), they'll give you enough access to their brains so that you can customize your board offering. In terms of NVidia's engineering time, this is not cheap, which is why they charge OEMs handsomely for this privilege.

So, without access to NVidia engineers, the register spec is of little aid to a driver-writing effort. Fuhgeddabouddit. The lowest level they're willing to document at all is the virtualized hardware interface and, as it happens, all that information can be found here.

Another reason NVidia is so protective of their specs is partially due do historical accident. Several of the principals at NVidia used to work at Sun Microsystems, where they developed the GX graphics accelerator. One of the SunOS releases inadvertently contained a single #include file detailing the GX registers. Using that file, a debugger, and a logic analyzer, Weitek managed to develop a clone of this chip in a mere six months. The clone was so good that Sun's own drivers worked on it. Needless to say, this left a bad taste in their mouth.

Astute observers will note that Weitek is now dead dead dead , but NVidia has more recent experiences to keep them gun-shy (Gnu-shy? :-) ). It seems that NVidia is of the opinion that the few Open Source releases they've made have directly resulted in getting hit with patent infringement suits. Apparently their competitors (who evidently have nothing better to do :-) ) download their source release, pick through the code, make some reasonable conclusions about the underlying hardware implementation, and start filing suits. Yes, it's bullshit, but if you're a cost-based accounting kinda guy (and most accountants are these days), NVidia's Open Source releases have been hideously expensive, and haven't resulted in the sales of additional chips.

These two last points (fear of cloning and fear of lawsuits) are the two big sticking points for NVidia at the moment and, sadly, I don't see them changing any time soon. Personally, I think their chip is so monsterously complex to design and fabricate that their fear of cloning is virtually unfounded. The patent thing, however, is another matter. Perhaps they might choose to hook up with Jeff Bezos in his stated intention to pursue patent reform. But that seems unlikely, since it could distract them from the very important job of staying ahead of 3Dfx, S3, ATI, and Matrox.

Keep sending them kind, polite notes of encouragement toward an Open Source release. But be prepared for a long, long wait.

I'm no real expert in these matters, but I would go as far as to say that any company having the skillz to reverse engineer and manufacture a clone board, surely should have the ability to disassemble the driver. Sure, giving them it in source form from the start might lessen the amount of work required for the reverse engineering, but I don't think it prevents it much. I'm thinking a bit about NVIDIA here, since I find their approach interesting: a partially open-sourced kernel driver, which interfaces with a binary-only userspace driver. Maybe that could work for the AnonymousIntern's company's products?

Right her e.

The NVIDIA Linux FAQ has been updated too. In particular, 0.94 Changes

Right her e.

The NVIDIA Linux FAQ has been updated too. In particular, 0.94 Changes

Right her e.

The NVIDIA Linux FAQ has been updated too. In particular, 0.94 Changes

It's secure, but note the realm that it's in - /marketing/oem/. I'd say that this pretty much lends credence to the rumors that Apple has struck a deal with nVidia.

-asy

It is not unusual for companies to deny any type of partnership until both parties are ready for an announcement. I would suspect that nvidia and Apple and are having serious contract negotiations.

Take a look at http://www.nvidia.com/apple. A password protected realm labeled "/marketing/oem"? It sounds like there might be some truth to the rumors after all.

Daniel

---

MS is using COTS hardware to speed R&D time and ultimately, time to market. For a company brand new to the gaming world, I'm worried. They're going to integrate existing hardware and software.

Granted the CPU and memory are COTS (I assume you mean Cheap Off The Shelf) hardware, but why did Microsoft pay nVIDIA $200 million - surely the GPU is a custom design, not just cobbled together?

nVIDIA Developer site

How do you define a "bog-standard consumer PC"? I just checked Gateway's cheap boxes (I've never bought anything from Gateway, it was just the first mainstream consumer box vendor that popped into my head). It turns out that these machines, which retail from $799, include integrated Intel 3D graphics hardware. Without a doubt, that hardware doesn't exactly compete with the big guys' stuff, but it might be able to handle the probably rather modest fillrate requirements of a GUI. Then again, it's a completely different question when we can expect to actually find this kind of hardware in the homes of the consumers... Hm.

I was in a hurry and my machine fell over as I was about to post this interesting link from the nVIDIA site. Lots to learn about therein. Sorry if I posted twice, give me a break.

So, where is the difficulty?

It's the 3D stuff. Sure, linux.3dfx.com has a fairly easy step-by-step for the various Voodoo cards, and both nVidi a and the Utah-GLX crew have come a loooong way towards making it easier.

Still, they all practically require recompiles (kernel modules, or in the case of Utah-GLX -- grab the CVS copy and compile the whole thing, then patch your kernel, if you want decent performance). That is a little daunting to Joe average.

Mind you, you only need the investment for Q3, HG2, Heretic2, SoF, and UT (or the older Q1, Q2, or Kingpin), but it's still significantly harder than the equivalent setup under Windows.

Note: I really like Utah-GLX. It makes the G400 in my work box hum and do good things. I also think that the state of 3D hardware under Linux has come a REALLY LONG WAY in the past year. It's where Windows was the first time I tried to get a 3D app running hardware-accelerated (say, 1.5 years ago, or so). We have a ways to go. We have come a really long way. We're getting there.

We will be there for John Q. Average. Just not right this minute.

Also: there have been attempts to bundle known-good versions of Mesa (HG2, Q3A), but that has proben problematic -- after all, just 'cause it works for Q3A doesn't mean it really *is* a good version of Mesa for, say, HG2 which stresses Mesa quite differently.

We've come a long way, baby, but that doesn't mean there isn't room for improvement...

I am getting really sick of everyone attacking nVidia for supporting Linux.

Figure 1: Level of OpenGL support provided by various Windows hardware vendors

3dfx: Works with Quake 3.
ATI: Works with Quake 3.
Matrox: Works with Quake 3.
nVidia: Full OpenGL support on all cards.

As you can see, nVidia has put a LOT more into their drivers than anyone else. I am working on a 3D game engine, and I have ported it to Windows, but it ONLY works on nVidia cards simply becaule no other company supports OpenGL lighting (yes, if it is not done in hardware, you are still supposed to have a software implementation. nVidia has both). If they released that open-source, they would lose an edge over the competition. YES, a significant portion of their drivers could be used in drivers for competing cards.

Figure 2: Register level specs for nVidia RIVA chipsets (128, TNT, TNT2)

http://www.nvidia.com/nv/nvarch.n sf/Home?OpenView.

If you don't believe me, go look. They also have un-obfusicated source code implementations there, and "driver development kits" for various OS's. Unfortunately, because they licensed some technology from other companies, they were not allowed to put up complete specs. The missing peices are rather important, so nVidia decided to write drivers for us so we could use them.

Figure 3: nVidia's options

• Hire PI to make DRI drivers. Cost: high
• Make their own DRI drivers. Cost: high
• Port their *possibly better* Windows infrastructure to Linux. Cost: low
• Release source code and let the community do the work. Cost: $millions in lawsuit

I would like to see the drivers released open source as well, and I am not convinced that it won't happen. However, I talked to one of the engineers, and I can tell you one thing for sure: The zealot attitude of most Slashdotters is only making the situation worse. Anger will get us nowhere.

Now I am going to go install my GeForce 2, which came today. Joy.

------

It's not really hard to do shadow casting in hardware, in fact standard OpenGL can do it, with a bit of creative use. See Nvidia's ShadowMap demo for an example. The source for a lot of the latest effects is Wolfgang Heidrich's thesis. Lots of really cool ideas, needs a reasonable computer graphics background, though.

If you want to get higher precision and speed you'll need some extensions, but not a lot. Depth textures and copy from framebuffer are enough, and have been available on high-end sgis for years, so the design is sort of stable.

Moral: with a good API and some creative use you can get really cool effects. Hardware can make it fast, but we'll have to see if the ATI chip delivers on that part...

I have the installation FAQ in front of me, and it's a long and careful list of things to do. From a quick scan, it looks like people of a nervous disposition should think twice before going down this list - making a quick backup of your current Xfree installation might not be a bad idea, or at least keep the old Xfree86 rpms at hand in case of crisis. Beyond that, it looks like it may conflict a bit with Mesa, so those modules need to be deleted or renamed as well (all in the FAQ).

For a speed comparison under the new drivers, Linux Games has a First Look up which gives me hope that I'll finally see some speed on my TNT2 card!

Cheers,

Toby Haynes