Slashdot Mirror
NVidia + OpenGL + Linux
BJH writes "Saw this on Ars Technica - NVidia have announced their new workstation-class graphics board, and say that it's going to have OpenGL drivers for Linux. Check it out their press release for more information.
" The hardware looks really, really nice too.
I am shopping around for a new system at the moment. You wouldnt believe how hard it is to find a decent card with Windows, Linux and Be support.
Time to buy the techie book and start on some device drivers I suppose.
Well, I can tell you some of the things that I have used or seen high-end graphics used for.
1. Visualization of the 3d structure of molecules. Many chemical reactions require an understanding of the fit between say a molecule and a zeolite in 3d. In order to visualize this in 3d I used an evans-and-sutherland graphics workstation with a mechanical shutter and jittering display image to project a 3d image into the space in front of my eyes. This type of application is big big big in the pharmacueticals industry. SGI has a very strong market share here.
2. Visualization of CFD simulations. Real-world work often requires a multi-dimensional projection of data onto a 2-d surface of large data sets - say flow fields obtained from computational fluid dynamics. Ideally you would like the ability to view the 3d time dependent result and rotate or pan the 3d field in real time. Most of the CFD work I have seen is done on HP or Sun workstations these days. Important in all sorts of places - example - modelling flows in an oil field, or in a tornado.
3. CAD/CAM. Computer aided design on a large scale. My brother is a wing designer for Boeing on the Joint Strike Fighter project. Boeing is doing all their airframe design in the digital domain now. This means preparing 3d models showing the actaul placement of every component in the airframe and determining it's mechanical performance.
Obviously this is important stuff - it's where the action is in the transferrence of science to every day life. I suppose the NVidia card may fit in the low end of some of these applications.
Sigh. winmodems are already there, and they are _NOT_ going anywhere, so it is better to get them working in all OSes instead of only on that horrible winblows.
And if Linux for example would have working drivers for winmodems, what would be wrong in them anymore? Yeah, they use cpu power, but so do passive ISDN-cards so why don't we throw those to trashcan as well?
...the companies that fork out big bucks for this sort of high-end hardware aren't looking to play Q3.
:-) which uses Intergraph (among other) hardware for development. Quadro is placed to compete with the likes of Intergraph and SGI. Wait a moment, wasn't NVIDIA working with SGI?!? Maybe we will see SGI Quadro Reality or something...
Well, unless they are id Software
-
__
Comment submitted. There will be a delay before you understand what you posted.
Ok - workstation-class graphics is what you're using when you're doing real-time intensive CAD/CAM and graphics modeling - you can do some of that stuff on your consumer card, but the really-photorealistic-realtime stuff requires a more expensive card. Basically, this is nVidia's first entry into this market. I suggest you pick up a copy of Computer Graphics World to find out more - it's actually quite interesting. Also, see the E&S ad in there - they differentiate quite nicely between 'cards for play' and 'cards for work'. It's my worst nightmare to have to do intensive 3d stuff and be stuck with a consumer card. For games, OK, but if you want the real tomato, the workstation-class stuff is what you want.
Visit
FYI... Although the press release mentions Linux as a supported OS, I don't see it listed in the preliminary specs on Elsa's page.
Hopefully this is an oversight or I just missed it.
Moderation in everything, including moderation.
Okay. Thanks for the quick education. I appreciate it. I'll pick up a copy of Computer Graphics magazine this evening.
-- Guges --
-- kwashiorkor --
Leaps in Logic
should not be confused with
Jumping to Conclusions.
He can't legally prevent you even if she's not GPLed; people are not property, like software is. Of course, the point is moot, sicne no woman would sleep with a dork like you.
That is an interesting concept. I don't want anybody that is 'not free' for a wife. Did you buy yours?
JM
Unfortunately the only high-end 3D app available (in beta) for Linux is Houdini, which from what I understand requires overlay plane support. I've seen no mention of any overlay planes on the GeForce or Quadro.
On the other hand, Maya and Soft seem to be able to manage without overlays nowadays, so maybe the Houdini people need to follow suit.
Also, hopefully the Linux version of Maya will be released soon.
> Ok - workstation-class graphics is what you're using when you're doing real-time intensive CAD/CAM and graphics modeling - you can do some of that stuff on your consumer card, but the really-photorealistic-realtime stuff requires a more expensive card.
But judging from the specs of what most 'workstation 3D' video cards from those manufacturers (before NVIDIA Quadro), it looks like they are intentionally crippling fill rates just to sucker professionals into buying more expensive hardware that has built-in T&L. If they can shell out a thousand dollar or two for T&L video, why not a decent rasterizer? Simply put, what professional video cards makers did sounded exactly like what Intel had done to Pentium II and Celeron A.
And since there is GeForce now (and probably something better 6 month later), this kind of industry-wide delusion should disappear soon, if not already does so.
> The GeForce was never a high-end vid card. High-end cards cost over $1,000. The GeForce was a high-end gamer's card, but not a high-end video card.
Really? Actually, what GeForce has done is to break the once high-margin, proprietary sector of industry, causing the price of 'high-end' region to come down. No offense, but it sounds like you have paid too much for being an early adopter of technology.
Ah, because we have decided to take an example from the French and declare that the Prime Meridian in fact runs through the Geek Complex. SlashDot time is in fact the world standard (SMT!).
If I recall, SGI was sponsering Debian or something.
If SGI is sponsering it, they want a kick-ass graphics card for serious workstations.
In addition, SGI allied with Nvidia.
NVIDIA already did. Get it at https://www.nvidia.com/nv/nvarch.nsf/Home?OpenView
Actually, no. That title belonged to S3. Granted, the Diamond Edge cards (Which were based on the NV1) sucked, but for its day they were pretty good. Certainly not slower than software. The "3D" at that time was NV1, Matrox Millenium, and Virge. 3Dfx had not yet shown its face. (BTW I have lost all respect for 3Dfx (sorry 3dfx) due to its recent moves. Long live nVidia!)
A deep unwavering belief is a sure sign you're missing something...
Go Be!
A deep unwavering belief is a sure sign you're missing something...
Actually you should learn before you speak. Winmodems do not have 16550s no modems do. That is on the Motherboard chipset. Winmodems however, do not have the modem controller on board. (Actually some of the latest winmodems have the controller on board but use the CPU to do compression, error control, etc.)
A deep unwavering belief is a sure sign you're missing something...
Actually you should learn before you speak. Winmodems do not have 16550s no modems do. That is on the Motherboard chipset. Winmodems however, do not have the modem controller on board. (Actually some of the latest winmodems have the controller on board but use the CPU to do compression, error control, etc.)But true Winmodems do suck. I have one. Should have paid the $20 extra when I was ordering my computer. Dammit, just one little dropdown box away from being able to use my Modem under BeOS. But I didn't understand the concept of software modem. Doh. And on top of that Bell Atlantic is taking forever to get DSL out. Its been in my neighborhood from a month and I still can't get it. I'll stop bitching now.
A deep unwavering belief is a sure sign you're missing something...
People will continue to use NT and its derevitave until Linux performs better. 3D people don't care about stability, NT can easily stay up the few days it take to render most movies (and if it takes more you are probably offloading rendering to an SGI box anyway.) NT and its whole kernel graphics paradigm just performs better.
A deep unwavering belief is a sure sign you're missing something...
Not necessarily. The GeForce still has nowhere near the transform power of a REAL high end card. Plus high end modelers need things like anti-aliasing, etc which the GeForce does not have. As nVidia's something Kirk put it, "you expect a workstation to have anti-aliasing, it a given" or something along those lines.
A deep unwavering belief is a sure sign you're missing something...
There have been a lot of posts so far on whether nVidia's code is open-source. nVidia created a hardware-enabled GLX driver that integrates with XFree86 3.3.x, and source is available (you can compile it yourself). However, the source is obfuscated to protect what they consider proprietary details about their cards.
XFree86 4 will be the thing to watch for GLX with integrated 3D hardware support; it looks to me like this is where nVidia is putting a lot of effort. Should be sweet!
--
Brent J. Nordquist N0BJN
Yea, I know what you are saying. My TNT2 may smoke my G200 in speed - but NVIDIA Trilinear Filtering is more like Trilinear Noise Addition(TM)... and the colors from the G200 allways seemed more vivid.
;-)
However, note that the Quadro is a separate silicon die from the GeForce. To begin with, I think (but I'm not sure) NVIDIA fixed some quality problems when they made the GeForce. They could (and should) have done even more for the Quadro.
Not like any of us mere mortals will ever get to see one in action...
-
__
Comment submitted. There will be a delay before you understand what you posted.
well just cause it is 3:30a for you doesn't mean it isn't something different in another part of the world (thank you timezones)
As I like to say for my 10am beer, it's midnight somewhere
Why doesn't slashdot post GMT, and maybe allow for you to set your timezone.
I really miss being in school where I could have a beer at 10a, instead of at work where that is my ~5th cup of coffee
Actually the GeForce has a lower clockspeed than the TNT2 Ultra.
Indeed, and by quite a bit. BUT, it also has four pixel pipelines, where the TNT* has two. And yes, these are used even if the scene is not quad-textured - they can render multiple pixels at once.
But, as you say, the real big deal with NV10 is the transform engine.
-
__
Comment submitted. There will be a delay before you understand what you posted.
True, but somewhat misleading to a newbie.
Real OpenGL has to be licensed, as your post suggests. however, just in case some reader doesn't know this yet, Linux 3D is generally based on an OpenGL "workalike" called Mesa. This provides an _uncertified_ OpenGL API clone.
Once you've ln -s libMesaGL.o libGL.o
and ln -s libMesaGLU.o libGLU.o
you might as well be using a real OpenGL.
Mesa is to OpenGL as Linux is to Unix, really.
The current (GPL'd) drivers for the TNT2 use Mesa, which is OpenGL in all but name, so they'd probably go down that route. Alternatively, they could actually pump enough momey into mesa to get it OpenGL certified...
If you have a Voodoo card under linux, Mesa is used to wrap around Glide, meaning you have better OpenGL compatibility under linux than with the crummy windows MiniGL 3dfx provide!
Another thing a lot of people don't know is that Mesa can be compiled for windows to run on top of DirectX/Direct3D, so _any_ Direct3D supported card can run OpenGL games, not just the ones that include OpenGL drivers! This is useful for running eg. Quake3 Test on an older 3D card.
why would anyone ever want their cpu time to be spend doing the hardware's work? my guess is that people simply do mot realize that this is what winmodems do. maybe someday people will realize this, and they will buy hardware which actually does something, rather than pushing it onto the cpu. its like having a video card that uses the main cpu to render graphics.
"The importance of using technology in the right way has never been more clear."
Very true. SGI owns the source of the reference implementation ('driver'). I can't seem to remember any announcements from SGI saying that they were going to open that souce. They get money from licencing that source to hardware makers. But then again, who knows, SGI has been releasing some other cool stuff lately. (Or promising to release...)
:-)
There is allways Mesa... and any OpenGL system, be it closed or open, is better that Direct3D. (Noooo, not like anybody on this site would agree with me on that one
-
__
Comment submitted. There will be a delay before you understand what you posted.
Let me spell it out for you one more time: Linux != Intel What's your point? I run FreeBSD, my processor is AMD. Or is that your .sig?
Don't agree. The glx-driver found on their ftp site performs very nicely with q3test (and my PII 450 :) and has not caused a single problem during the 2 months I've used it. /jarek
Even though they conspicuously didn't mention, Quadro is a member of the GeForce / TNT family, so existing GeForce / TNT drivers should WORK, just not as well as Quadro drivers. . .
Are they releasing the specs/source as well so that drivers may be written for other platforms? I would *really* like to see BeOS drivers for this. It seems like the kind of hardware BeOS was made to run.
--
grappler
Vidi, Vici, Veni
> overlay planes are needed by many high end apps,
> and if these cards dont support overlay they may
> not get far
Overlay planes are not really essential if you can grab the image and display it fast enough. I know that its a waste of memory and processor cycles and bandwidth, but for interactive applications you have those cycles and bandwidth to burn.
The nice thing about throwing out the overlay-plane hack (perhaps you can detect some bias here) is that you can do much better rendering of the foreground elements that you are interacting with. Overlay planes were typically used to draw things over static backgrounds, and were limited to just a few bits. If you just load the whole background image in every frame, then you can draw nice antialised, colored, even shadowed lines and objects over the background, and get a much richer interactive experience.
I've ported a few of my SGI-based visual effects tools to Linux, and had to give up on overlay planes, and while it was difficult at first -- I don't miss them any more. And this is using extremely slow refreshes; once there is good hardware accelerating for OpenGL glDrawPixels commands then I will not miss overlay planes at all.
One thing that these programs do is they only redraw the dirty parts of the screen. As you're dragging a rubber-band line across the screen, only a sub-rectangle of the image needs to be refreshed, and this can be substantially faster than refreshing the whole screen.
thad
I love Mondays. On a Monday, anything is possible.
Anybody got an reference to the official word on Maya for Linux? I haven't seen it mentioned in a while...
Yeah, they're GOING TO HAVE Linux drivers... I'll believe it when I see it. Lots of bile gets spewed in 3dfx'es direction, but their commitment to Linux is long-standing. Nvidia has made *ONE* release of 3d drivers so far. Since then, 3dfx (or more properly, Darryl Strauss has released several revisions of the voodoo's drivers). If anyone asked me "I want a 3D card now...what's the best one to get that'll work w/ Linux?" I'd say, "If I were getting one today, it'd be a 3dFX based card...probably a V3 3000" (Tho, from what I'm reading, the Matrox G200/400 drivers have come a long way...but still the frame rates are lower, and the drivers less mature than the 3dfx Mesa driver)
The Nvidia drivers are actually under the XFree licence, not the GPL.
--
Business. Numbers. Money. People. Computer World.
Well, 3dfx may release the V3 specs sometime soon. (I expect to see them around Q1). Real specs (Hello nVidia!!)
I think Matrox and ATI are currently ahead of the curve as far as being good to us open sourcers.
I currently have the nVidia TNT, Permedia2, and the VooDoo3. The V3 rocks em all, 2D. I'm waiting for documents from 3dfx so I can start coding (Hello, DRI anybody?) for XFree86 version 4.0.
If I can find the documentation, I'm going to look into the Rage and see if it works on the Alpha.
Pan
I said no... but I missed and it came out yes.
yes, but... I was hoping to get away from binary-only stuff, and I also would like windowed 3D. Will I see much difference between my Voodoo Banshee and a Voodoo 3 ?
3dfx'es glide is binary only (so far), but their 2D part of the card is open source. Even so, glide/linux is mature and stable. V3's get much higher frame rates in Linux/Mesa than TNT's, TNT2's and Matrox can at present. And no, that situation will not -instantly- change with the public release of DRI. The latest #'s I've seen show G400's giving frame rates (yes, using Q3 as a benchmark) in the 40-55 range w/ a resonably fast computer (Cel 400+). V3's are providing frame rates of 70+ on equivalent hardware. And no, the V3's dont "look awful". For the most part, Linux apps dont use enormous textures. Comparing the drivers on Windows, Mesa's/windows/3dfx output looks better than 3dfx'es own OpenGL driver.
Geforce is a gaming card, this card is for a totally different performance so polys/sec is fairly irrelevant way of measuring performance
I know you might think, its a hardware driver, if it works, why would you need open source drivers.
But hardware drivers for 3d cards are almost never perfectly, and it is common for a company to release several drivers before they are usable for most games. Especially opengl drivers, which seem to only be made for the sake of Quake*.
Think how half-ass the drivers will be for linux. And do you really think that they will fix bugs in a timely manner, for an alternative OS. Hell no.
For an alternative OS, you need programming specs, and maybe some open source drivers to accompany it. Even without programming specs, you can get a few bugs fixed (like the NVidia OGL driver from the glx cvs server does not have that XScreenSaver bug, but it is slower than the released version). However, without specs, something that is just open source like the NVidia tnt drivers, can't really be updated. That is why the TNT is slower than the G200 on linux. That is why open source drivers don't matter, but specs do.
I know I will be moderated down for this, but . . . Vincent
I am a 3D designer and I've been waiting patiently for a very good 3D card to come out for Linux. I know that currently some of the 3D labs chipsets. But full hardware support of OpenGl isn't that great. This introduction of the new chip from nVida will help boost support for 3D apps for Linux. With the addition of Houdini to the list of 3D apps for linux put's us on the map and with this card coming. This will hopefully convince Avid, makers of SoftImage and SGI, makers of Alias PowerAnimator and the ever so popular Maya, that linux is a viable platform to do 3D work. I think if the price is right i.e. around $700 US or less then I'm gonna buy it. As far as image quality, as one of the previous posters was wondering about, the quality of images displayed may be the same as the Matrox card, but the thing you have to consider is that this new card can open a 3D project in Maya with about 1,000,000 poly's or more and you will have no delays when moving around, whereas in Matrox's case it will be chuggin along and have you waiting till the cows come home. Some of you want to know what makes this board "workstation" class. Well I think that the benchmark numbers speak for themselves. Look at the comparison between the Intergraph Wilcat 4000 and the Oxygen GVX1. The Wildcat has geometry accelerators which means that the processor don't have to transform the geometry, but the board can handle everything. If this is included on the card that Elsa is making for the Quadro, then definitely we will be seeing support for more 3D apps in the near future. I know one thing, I'm gonna get the Quadro as soon as it comes out =)
Well, if you want windowed, then you should probably go with NVidia or Matrox. Note that matrox glx drivers are open source and are being actively developed ...
yeah.. this card will be absolutely useless without hardware overlay plane support. I recently switched from a permedia2 to an older elsa gloria XL. The difference(just with the overlay support) in apps such as maya(artisan), alias studio 8.5, and softimage is assounding. With any luck it'll support overlay planes and hardware alpha simultaneously, unlike the highend 3dlabs cards.
Hey, don't bitch about the DSL... You will eventually get it. It could be worse. If I lived 400 feet down the road, I would have cable modem access for $25 a month instead of the $20 a month for 28.8k (sometimes 31.2k!!! :/ ). I am serviced by a different cable company which outright SUCKS by comparison.
Ouch. That does hurt
A deep unwavering belief is a sure sign you're missing something...
It is great to see that kind of support for Linux from Nvidia. I'm still waiting for my TNT2 to arrive. I hope that there will be good drivers for TNT2 available for Linux/Mesa soon (with Xfree4?)
Greetings,
Project Manager of Crystal Space (http://www.crystalspace3d.org). Support CS at http://tinyurl.com/cb3x4
The specs look good, but at what price and when.
Maybe with all these manufacturers releasing drivers, the Winmodem makers will follow suit and release their drivers, and we can finally get Winmodems working under Linux
"Oh, I got me a helmet - I got a beauty!"
Jack Nicholson, Easy Rider
winmodems have their name for a reason, they are win modems, noone needs another incarnation of windows inspired hardware whether it be in linux, bsd, or anything else. i say go back to the modems with their own 16550s, that way you don't drain the CPU unnecessarily.. Why use modems anymore anyway? ``What's the only other thing that spreads like Microsoft? A virus''
it looks like a great chip, and, as it seems to be using much of the same technology as the GeForce, I am sure it IS a great chip....
/., that is a card I would love to have...
but did you all read the press release? it is going to be placed on some very high-end boards... it is competing in the price-class of $1000 videocards....
even if the chip is not so expensive, the board will be, the older ELSA high-end cards were VERY pricey...
too bad really, because, just like all the other members of
We are all in the gutter, but some of us are looking at the stars --Oscar Wilde
Grrr. my nick is "Forward the Light Brigade"...
They didn't say whether the Linux drivers will be GPLed, or even some Open Source(tm) variant. With all the noise about OS lately, you'd think they could at least TRY to jump in the bandwagon... unless they've already decided not to, which would be a Bad Thing. Maybe they don't know about the /. effect yet...
They're not dropping NT support any time soon, I see...
Hmm... W2K isn't the sure thing it's supposed to be, is it?
--
Peter
overlay planes are needed by many high end apps,
and if these cards dont support overlay they may
not get far.
I hope not, we don't need any more winmodems
What about image quality? Professional 3D cards generate extremely high fidelity images, and thats one of the things you pay for. Just compare the image from a TNT2 and an "identical" one on a 3Dlabs or matrox card.
It seems that this new card is someway related to the 'older' GeoForce256, does it mean that we have GeoForce support under Linux? maybe some kind of specs 'ala' Matrox G200/G400? I have just order a Matrox G400 MAX, just because they really suport OpenSource (or at least they release the specs so other people can support it), but if I were sure that NVidia is going to support some kind of GLX interface for their GeoForce I would buy it instead of the G400. Any info about NVidia current plans on releasing their chip specs?
I know that there is always "they will copy our technology" thought, but they will spend a lot of time in doing so, and current 3D scene goes too fast to make it a risk.
Of course, just my two cents.
Athlon-optimized OpenGL drivers for Linux... How could life possibly get any better? Can you say "Quake III"?
Oh well, I hope I get a video card that's half as nice as this in a year or so...
Maybe one day this sort of card will inspire a Tom's Hardware comparison that includes using NT as compared to Linux.
---
pb Reply rather than vaguely moderate me.
pb Reply or e-mail; don't vaguely moderate.
A little bit of info at Thresh's FiringSquad
:)
It has some information about Quadro vs. NV10, and even some CAD-related benchmarks against cards like the GVX1.
Guess who won..
-Warren
I saw preliminary pricing for the board; the 64MB version is expected to go for ~$800 if I recall correctly.
:)
I see some people complaining about prices here. While it's not the $250 price point of the GeForce, it's still very reasonable/pretty inexpensive for workstation-class graphics.
Nvidia has now expressed an amazing commitment to Linux across the board with their products. Let's hope SGI would do the same with their 3d Software, Maya on Linux would be oh so cool
My friend just ordered his 2 weeks ago and is awaiting its arrival. Will we be able to dl a driver for this bad boy and get it working?
Natas
Natas of
-=Pedophagia=-
http://www.mp3.com/pedophagia
Also Admin of
http://loki.linuxgames.com
Hey, I'm all for throwing ISDN to the curb. It had it's brief 15 minutes of fame and the telcos ruined it by overpricing. Now, ADSL takes it's place for a cheaper price and a huge increase in performance. Good riddance ISDN.
Considering how abysmal their previous attempts at drivers are, it might actually be working on Linux in, oh, 2001. LATE 2001. They're a marketting driven company. Don't believe the hype.
Actually, the TNT's were NV4's. TNT2 is NV5, and Riva128 is NV3. The NV1 was a terrible failure, and could actually render 3D graphics *slower* than the top end software solution of its day...
Just goes to show, if at first you don't succeed...
Dan
From what I've seen on the lists, the TNT cards are generally regared as faster than the Matrox cards when it comes to 3D.
But - and this is a big ol' but - Nvidia hasn't played quite as nicely as Matrox when it comes to releasing specs. So, the GLX guys have been able to optimize the heck out of the Matrox driver, and the Nvidia driver hasn't gone as far.
In fact, John Carmack has more or less stated that he's personally focusing on development for the G400 because the specs are there and he likes to program the *hardware.* This is kind of a bummer, because at this point, the TNT cards are 2 generations old - how many super-secret secrets can be left in it?
This will hopefully convince Avid, makers of SoftImage and SGI, makers of Alias PowerAnimator and the ever so popular Maya, that linux is a viable platform to do 3D work. I think if the price is right i.e. around $700 US or less then I'm gonna buy it.
As far as image quality, as one of the previous posters was wondering about, the quality of images displayed may be the same as the Matrox card, but the thing you have to consider is that this new card can open a 3D project in Maya with about 1,000,000 poly's or more and you will have no delays when moving around, where as in Matrox's case it will be chuggin along and have you waiting till the cows come home.
Some of you want to know what makes this board "workstation" class. Well I think that the benchmark numbers speak for themselves. Look at the comparison between the Intergraph Wilcat 4000 and the Oxygen GVX1. The Wildcat has geometry accelerators which means that the processor don't have to transform the geometry, but the board can handle everything. If this is included on the card that Elsa is making for the Quadro, then definitely we will be seeing support for more 3D apps in the near future.
Who could forget about gaming? Well the board should support fully OpenGl games, such as Quake 3. Most games say the support OpenGl, but that's not full OpenGl, but use the OpenGl implementation like the one 3DFX makes. So if you plan on playing games that have full support for OpenGl then your frame rates should be extremely good.
I know one thing, I'm gonna get the Quadro as soon as it comes out, and first thing I'm gonna do is fire up q3 and let the fraggin begin =)
Query -- what exactly does ''workstation graphics'' mean? This is an honest question, not a troll. The sort of work I do doesn't require much in the ways of graphics, and I don't know what sorts of things other engineers are using computers for -- I'm a little sheltered in my Unix-programming-only microcosm. What is it that people do with graphics that there's a significant market demand for ''workstation-class graphics''? I'm not saying that there is no need, just admitting to my own ignorance.
-- Guges --
If you want to play 3D games NOW and don't want to wait for months and months for XFree4.0 or promissed drivers, use 3DFx (Voodoo3?).
Yup! SGI partnered with Nvidia last year, and had engineers work on the GeForce256 and the Quadro. There are major implications downstream for SGI and Linux in the graphics workstation market, using Nvidia-based graphics cards.
Over the past few months:
SGI has openly licensed its XFS journaling file system to linux, paving the way for Linux integration on SGI hardware.
Along with Redhat, SGI is funding Precision-Insight. Precision-Insight hired Brian Paul, the author of Mesa OpenGL port to Linux. Precision Insight's Multipipe Direct Rendering Infrastructure (DRI) within the upcoming XFree86 4.0 X Server. SGI is also providing extensive technical help and other resources to benefit the project. The DRI will include additions and modifications to GLX, Mesa 3.1, and XFree86 4.0, as well as any required modifications to the Linux kernel. Both Red Hat and SGI have agreed to allow full source for the entire project to be donated by Precision Insight to the open source community.
SGI is shipping servers with Linux, adapted from Redhat.
SGI has had a rocky time recently in terms of profits, but the technology is first-rate, and they are leveraging a strong Linux future, probably replacing IRIX, for x86-based workstations. Think about boxes with 1GHz+ Athlon, Coppermine, and Merced processors, and video cards like the Quadro, outperforming graphics workstations that cost 10 times as much. All of this is great for Linux.
---
It's unfortunate that while 3D hardware is nice and cheap, the nice-and-cheap kind is designed to throw many frames/second of low-polygon models at the screen. The benchmark is Quake III, and that means the most important features on a cheap 3D card are fill rate and texture-map speed.
Higher-end cards, those designed with more advanced features like geometry setup and anti-aliasing, are much more suitable targets for whatever 3D-like user interfaces eventually arrive. You can count on such interfaces to make use of high-precision models, high polygon counts, and almost no dependence on texture-mapping (or even fill-rate, for that matter).
The key is _detail_, and that will require very high resolution rendering of anti-aliased models in very large memory spaces. Hopefully, NVidia's entry signals a new era for high-end 3D graphics pipelines, one of increasing affordability.
MJP
Don't try that "protecting the children" shit you people use to keep the tits and bad words off my TV. --Seanbaby
Just wondering - The voodoo cards have binary only support, how are the Matrox G400 and NVIDIA TNT 2 ?
I've currently got a Voodoo Banshee, and it's showing it's age, so I'm thinking of getting a new card.
so how is this "high-end" card different from GeForce? Is GeForce not high-end any more? And what exactly is meant by "workstation-class"?
___
If you think big enough, you'll never have to do it.
who cares if the drivers aren't open source? at least they're supporting the linux platform.
Look, I'm not tearing my garments: this is an area in which 'closed source' is relaitvely harmless, but Free is Better. If you're interested in hardware drivers, having the source code means you can
a) learn something,
b) contribute: maybe users can perfect a driver.
c) Port it to another platform( say *BSD).
That, and good PR, is why they should release the drivers free.
That they support the platform is good, but only as a sign of the recognition Linux is getting as a widely used OS. With OS drivers, they would be supporting the concept behind it, which would make me way happier.
Well, based on past performance from high-end 3D graphics cards, you'd probably be better off with a card designed with games in mind. Remember the Diamond Fire GL Pro 4000, or whatever it's name was? It was based on the 3DLabs Glint chip I believe and was around $3000. Good for 3D modelling and such, but horrible for 3D games. This card is probably optimized the same way, games are not their target market by any means.
Unfortunately, I've heard that NVidia hasn't released many specs on their cards other than a single version of a video driver. I think it's been updated slightly, but it still doesn't have any of the fun features, I guess..
--
...how similar the names 'GeForce' and 'G400' are to 'G4'? I wonder if nVidea and Matrox are trying to capitalize on Apple's hype about the G4 by giving their products very similar names. Call me paranoid, but I'm sure those marketing types consider such subliminal associations very seriously. Sorry if this seems off-topic.
As the title mentions, they don't. The only way to know how to program anything in NVIDIA's chips is to reverse engineer the driver code they released for Linux thus far. This way, unless NVIDIA do the Linux drivers themselves noone will, and if and XFree86 code changes noone but them will know how to change it. C'mon NVIDIA, Matrox released the specs for their cards and so did ATI, so just do the same!
I tolerate Windows, love Linux, and adore BeOS, but the hardware issues are a major pain in the ass for my favorite OS. If you only want to have a single 2D/3D card and want hardware 3D in BeOS, your only option for now is a Voodoo3. BTW, does anyone have any idea how much of the work in hardware-accelerated OpenGL driver development is OS-specific? In other words, once a company develops a good Windows OpenGL driver, do they have to start from scratch to develop Linux and BeOS drivers, or is a lot of the code portable? (sorry for being such an ignorant newbie).
The Quadro has 388 pins.
The Quadro claims 17 million polygons/second.
The GeForce claims 15 million polygons/second.
How large is the performance gap between the two?
penguinicide... when jumping out a window just won't do.
that's correct. That's why the drivers are optimized for athlon too :)
---
Still, I wonder if the nVidia coding for OpenGL will be applied to any of their earlier boards. I have an NV1 (Diamond Viper RIVA TNT V770D) board that required a driver download before I could get anything other than 320*240 on.
--
My office has been taken over by iPod people.
I have a TNT2 (Diamond v770 Ultra) and it's
fantastic under Linux. I love the 32m of
texture memory, but 64 would sure be nice...
I do OpenGL development and it suits all my
needs.
I highly recommend it.
--- witty signature
I have a G400 in one box and a Leadtek TNT2 Ultra in another. Like most cards these days, they both are very good for 2D work, easily capable of exceeding my monitor's 95KHz horizontal sync limit.
:-)
What's more, there are GLX libraries available for both chipsets -- still binary-only (I think) for nVidia, open source for the G400 -- so you'll have hardware-accelerated Mesa in either case. In my experience, the nVidia GLX library is more stable than the one for the G400. The G400's also has some problems with texturing -- the 'superquadrics' mode for xlockmore will demonstrate this. On the other hand, the G400's implementation feels slightly faster than nVidia's. I haven't measured frame rates with the G400 but the TNT2 generally posts scores of 60-80 frames per second for the 'ssystem' OpenGL solar system program if you turn off the on-screen HUD (seems having text in the window slows things down to single-digits). I would expect the G400 to post numbers in the same range.
Next time I have the G400 in the Linux box, I'll measure it. Right now, the G400 is sitting in a Windows box where I can watch an occasional DVD -- the DVD player that ships with the G400 is better than the PowerDVD software that shipped with the TNT2 Ultra.
One thing I noticed with the GLX drivers, and the one for the G400 in particular, is that if the screenblanker kicks in while displaying 3D, you're in trouble. The machine is still usable if you login remotely, but the local console/keyboard is hosed until you reboot. Moral of the story? "xset s off"
The kicker, though, is in the price. You can pick up a 16MB OEM G400 (what I have) for around $100. The best price I've seen for a TNT2 Ultra is around $170. For $100, the G400 is hard to beat.