Here is a test concerning NVIDIA SLI on Linux (in this case, OpenSUSE 10.0 OSS). “With our previous article that we published moments ago, demonstrating the performance of the GeForce 7800GTX 256MB under Linux with the 1.0-8174 Rel80 drivers that were finally released today, there’s no disputing that the Windows XP NVIDIA ForceWare users can generally see a significantly higher frame-rate with the same hardware components, in addition to other features that aren’t yet supported by the proprietary NVIDIA Linux drivers. However, how do NVIDIA’s initial Rel80 Linux drivers (1.0-8174) fair in the world of Scalable Link Interface?”
NVIDIA Linux GPL Results
lol.. is this your feeble attempt at trolling? If my only choice is nvidia’s proprietary drivers vs. ati’s OSS or proprietary drivers, I’ll take nvidia. ATI can’t even get their Windows drivers right.
The OSS drivers for ATi’s cards are not made by ATi. They are made by http://dri.freedesktop.org/wiki/
They’re pretty good actually. Not perfect, no driver is.
I understand their want/need to keep things secret, but at the very least they should be willing to open up the documentation for chip families that go back a few generations.
I highly doubt there are many secrets left in the TNT2 or Geforce1 families that they would be too worried about at this point. Even the Geforce2 family. I’d think that ATI has surpassed any of the advances in this family with the R5xx series, considering that ATI works with the OSS community on drivers for chips all the way up to R2xx chips.
This is the reason why I own a FireGL 8800. It’s the fastest card that I know of that has OSS drivers. When ATI or someone else releases specs on a family of chips with greater performance, I’ll upgrade.
Edited 2005-12-06 21:32
nVidia will never release driver code, as all cards share the same unified driver source.
As to documentation, that’s only going to create more hassles than it’s worth for them. Very few people would go out and buy a GeForce 1 if nVidia released documentation for it today. nVidia stopped making them long ago, and they aren’t going to start producing more now for the couple people that might want one. They probably don’t even have documentation in a suitable form to give to people trying to write a driver from scratch. To make the docs useful to people, they would have to assign engineers to the task of providing support, which isn’t very likely to happen considering it won’t bring in more money than it would cost.
I don’t expect them to release code, and didn’t even mention that.
———that’s only going to create more hassles than it’s worth for them.————-
Supporting more families of cards = less hassles? It seems as if they were supporting less families that would mean less hassles. Let the OSS community have the hassles if it wants them that bad.
———nVidia stopped making them long ago, and they aren’t going to start producing more now for the couple people that might want one.————
That’s the point. It’s not about them producing more. I doubt anybody thinks they will. It’s about those who already have them getting support from the community. You know as well as I do that nvidia isn’t gonna support these cards forever, but it’s in their best interest for some sort of support platform to be available in case people decide to use these cards.
———-They probably don’t even have documentation in a suitable form to give to people trying to write a driver from scratch.————
Unlikely. They can make a tremendous driver for these cards but don’t have good documentation? Sorry, but these two just do not go together.
———–To make the docs useful to people, they would have to assign engineers to the task———-
No they wouldn’t. They could do this to increase the effectiveness and timely manner in which these drivers could/would be created, but the lack of nvidia engineers would not result in zero driver development.
I understand their want/need to keep things secret, but at the very least they should be willing to open up the documentation for chip families that go back a few generations.
Well, very, very few people are ever going to consider buying chips that are several years old because they have specs, documentation, and potentially, open source drivers, and companies like nVidia will never provide support, engineers and resources in doing that. From their point of view it is a loss-leading exercise and it just isn’t economically viable. They won’t do anything like that simply out of the goodness of their hearts.
nVidia’s drivers have a unified base as well, so it would be pretty difficult to crowbar all of the relevant stuff out of there for people to use without giving stuff away about all their hardware.
This is the reason why I own a FireGL 8800. It’s the fastest card that I know of that has OSS drivers. When ATI or someone else releases specs on a family of chips with greater performance, I’ll upgrade.
Well, I admire your stand on that but I hate to point out that you’re the exception rather than the rule.
———very few people are ever going to consider buying chips that are several years old because they have specs, documentation———–
It’s not about people who are potential buyers. These cards have been bought and sold and are nearly completely out of the marketplace.(from nvidia’s standpoint) The only ones left are either in production machines or collecting dust at some mom-n-pop-shop awaiting a home.
———-and companies like nVidia will never provide support, engineers and resources in doing that.———-
I’m not calling for any resources of any kind besides the docs. The OSS community has proven many times it has the resources. Just give up the docs, we’ll be happy to do the rest. Being as these families are a few gens back, there shouldn’t be any secrets left for nvidia or whoever to hide.
———–Well, I admire your stand on that but I hate to point out that you’re the exception rather than the rule.————–
Yeah, I know. And I also realize that the DRI has only recieved a majority of docs, not full docs from ATi. Which is fine. A mostly complete OSS driver is still better than the closed source ones provided.
Given XGI’s announcement, my next card may very well be a volari.
How is that card anyhow? I thought that in the ATI family, it went up to the Radeon 9250 in terms of OSS drivers. I’m maybe like you in that, it does actually matter to me. Recently I’ve been looking around for what components I’d want to put in a new box, and the 9250 looks like it may be the sweet point in terms of price, performance, noise level, and freedom’ness (to invent a word)
Funny though, ’cause till recently I would have strongly told anyone building a linux box to go Nvidia. Having dealt with both nvidia and ati in their proprietary drivers, easily nvidia takes the cake in my book. That said, not having to mess with outside drivers _at all_ and go open source all the way, at least on that level, sounds increasingly like a “Good Thing” to have.
I am also curious to know about the performance of the FireGL 8800 with open drivers. Whenever I buy components (or suggest brands to others) I definitely keep products in mind that are well-supported in Free-Software land.
If only graphic card specifications weren’t all but impossible to obtain in order to write drivers. [sigh] On that note, I *really* hope that The Open Graphics Project [ http://lists.duskglow.com/mailman/listinfo/open-graphics ] succeeds and produces a card soon. I am more than ready to put my money where my mouth is by purchasing the first truly open graphics card. If that card happens to come from Nvidia, ATI, or Intel–great! I’m not holding my breath, though.
——–I thought that in the ATI family, it went up to the Radeon 9250 in terms of OSS drivers.———-
It does. In the ATi families, Radeon 9250 is an r2xx chip. So is the FireGL 8800. The 9280 is specifically a r280.
This is explained at the DRI wiki…..
http://dri.freedesktop.org/wiki/ATIRadeon#head-7528b1c267d9e6f91e26…
I also have a Radeon 9000, another nice card. It is a R250, but still based on the R2xx familily.
Let’s put it this way. I got my 8800 off ebay for around $60. There’s no way the 9250 could touch this in terms of performance. And the fan is tiny, it doesn’t make much noise.
You can find these on pricewatch for about $75.(which IMHO is worth it)
This is what the card looks like.
http://www.lostcircuits.com/video/ati_firegl8800/2.shtml
(Scroll down a tiny bit)
Regarding your title, a GPL policy wouldn’t do much good. To my knowledge, X.org is released on a BSD-like licence. Of course, X11R7 is going to be modular, but reinventing the wheel once again when the documentation could be useful for many platforms would be a waste of time.
As for not releasing the specs for their older chips… Why would they? They want you to buy their current chips. Sure, it would be useful for many people, including myself. After all, I still use a GF4 Ti4200. Still, they are a corporation with shareholders, not a charity. It’s all about money and keeping control on their products while avoiding potential lawsuits for stepping on some patents or exposing a workaround to a copy protection (TV out with Macrovision comes to mind)… Same for ATI, by the way. (Un)fortunately (depending from which point of view you are looking), it’s not engineers that are running enterprises.
———Regarding your title, a GPL policy wouldn’t do much good. To my knowledge, X.org is released on a BSD-like licence.——–
Ok. They need a work-with-the-OSS-community policy.
———As for not releasing the specs for their older chips… Why would they?————
Because smaller companies such as XGI are talking about it. Even if they don’t do it for another year this is a good way to gain marketshare.
The last thing nVidia(or ATi, whoever) wants is more competition.
——–They want you to buy their current chips.———
I know. I know I’m not a majority, but I know there are alot more linuxheads like myself who would very easily switch to XGI for the OSS driver. And nVidia knows this. They’ve been happy to have *the* card of the linux community, thanks in part to the amount of effort they’ve put into their driver. If I were recommending a card to a friend, I’d definately say nvidia first. But let XGI take the OSS route and they’ll become the first choice for many.
Linux might not have much marketshare now, but it keeps growing.
———–Still, they are a corporation with shareholders, not a charity. It’s all about money and keeping control on their products while avoiding potential lawsuits for stepping on some patents or exposing a workaround to a copy protection———-
I know. That’s why I don’t expect them to open a new card family’s docs. They have legitimate reasons to protect their shiny new IP. And if XGI became a bigger player I wouldn’t expect them bother with it either.
Apparently it’s just me.
Because smaller companies such as XGI are talking about it. Even if they don’t do it for another year this is a good way to gain marketshare.
Rather mindshare. I don’t know for you, but I would personally wait until they fulfill their promises before buying one of their cards!
I know. I know I’m not a majority, but I know there are alot more linuxheads like myself […] Linux might not have much marketshare now, but it keeps growing.
You are not alone. If XGI does deliver the goods, I will buy one of their cards (as long as the 2D output is crisp like ATI; I hate my NV for that). Still, I don’t think the user base for desktop Linux is significant enough to change the way of major corporations to doing business, especially if their existing customers don’t really care of the nature of their driver as long as it works.
I know. That’s why I don’t expect them to open a new card family’s docs. They have legitimate reasons to protect their shiny new IP. And if XGI became a bigger player I wouldn’t expect them bother with it either.
It’s not only about protecting their new IPs, but also complying with laws in some countries while avoiding legal disputes with rabid IP companies. Of course, there is probably a question of control, too.
By the way, I am not endorsing what they are doing, I am merely stating what I am observing.
———-Rather mindshare. I don’t know for you, but I would personally wait until they fulfill their promises before buying one of their cards!———–
Well heh…. we should still ultimately be realistic about things. In 2009 when XGI hasn’t made good on it’s promise I won’t be holding out hope. But if they are a few months late for various reasons they will still be held just as high in my opinion.
—————Still, I don’t think the user base for desktop Linux is significant enough to change the way of major corporations to doing business, especially if their existing customers don’t really care of the nature of their driver as long as it works.———
I agree with that. However, linux is enough of a growing threat that most corporations are paying attention.
If XGI actually does this, it’ll be a huge thing for the OSS crowd.
———It’s not only about protecting their new IPs, but also complying with laws in some countries while avoiding legal disputes with rabid IP companies. Of course, there is probably a question of control, too.————
Well… yeah, I know there are multiple reasons. IP is still the main reason, as I doubt there is any law in any country that states the illegality of showing the world how you constructed your chip and how to create drivers.
As IP ages, it becomes less important. That’s why/when documentation would/could and ultimately should be opened.
And as much as it bothers me I’ll still stop to say that ultimately if they don’t release the docs, it’s still their right to do so. The docs are theirs and that’s really the bottom line.
Still no stable composite support it seems
not exactly stable, but the older drivers that used to crash and burn really hard on me are flawless
[ubuntu 5.10]
Well, I have a slightly older card then what they’re testing (Geforce FX 5600 Ultra). Obviously it isn’t really old per say, but in relation to there test. At any rate, UT2004 happens to get around 80-110 FPS (it’s not very steady) now as opposed to the 35-60 it used to get just before on the other drivers. This may be the chance for some older cards to start really showing what they’re actually capable of. I can imagine newer cards just simply having far too much muscle to see a phenominal difference.
One other odd thing I found, though, is that with this that fonts are rendered much differently with it. I’m not really sure why that is exactly, but my fonts have by far changed. The size of my size 8px font (on the older driver) is the same size as 11px font now. Kind of odd, and Firefox doesn’t seem to be handling it well for whatever reasons. Overall, a nice change by far though.
Composite with xcompmgr has always been stable with me and these drivers are no different. It’s not nvidia’s fault that composite dont work, it’s Xorg and still experimental. If you try the release candidate version of xorg , composite is more stable. I’ve always found GNOME to be MUCH better with xcompmgr, KDE all sorts of crap happens, missing parts of the window, right click stopping.
The window movement is super smooth and no redraw what soever, xcompmgr is with shadows actually speeds up things here (Geforce 5900 ultra), couple of minor bugs other then that it’s great.
IMHO the problem isn’t patents or support, but it’s consumers. If no-one bought a video card unless it had OSS drivers then video card manufacturers wouldn’t have much choice but to release documentation.
Consumers are a bit too thick though. A theoretically fast video card with a crappy driver doesn’t perform as well as a slower card with full 2D/3D accelerator and bit blit support.
Anyway, three cheers for Intel (who do release full documentation for their video, including 2D and 3D accelerators and everything else, for everyone in the world to download)…
Theres are no need for OSS drivers. My Windows Nvidia and Ati drivers work very well and I really don’t feel like reading the source.
That’s because you use Windows, a commercially well supported propietary platform. You really just don’t understand the problem with propietary drivers, but that’s because you never had to deal with them.
————–Theres are no need for OSS drivers. My Windows Nvidia and Ati drivers work very well———–
You said the key word. Windows.
Linux/OSS progresses and evolves in a different way, and also evolves/progresses faster than windows.
Having OSS drivers means that as the OS progresses, the drivers progress with them, and when your shiny new linux comes out your hardware just works. With the binaries it requires a wait time from the companies to add in compatibility first.
Whatever driver version you are using right now, try using it in vista. Chances are it won’t work.
You get new windows versions every few years, I get a new linux every 6 months.
i know it’s a benchmark but who will play with the setting beyond 4xaa/8xaf… sigh
I’d like to see the performance of nvidia drivers on FreeBSD x64-oops, they don’t make drivers for FreeBSD x64. They have never said why.
@halfmanhalfamazing
You’re not even able to get the specs under NDA, nVidia is completely sealed.
That is one of the main reasons for me not to use nVidia driven technology.
Can you compile this driver without kernel headers (they are GPL’ed)? If no, someone should sue nVidiia, because this driver is illegal and breaks GNU GPL license (it’s derived work of kernel).
I don’t want to live in binary only world -> http://thread.gmane.org/gmane.linux.kernel/354704
The kernel interface to the binary core is under the GPL. It’s debatable whether the licence should extend to the binary or not. And honestly, I would rather have some drivers than no driver at all, considering they are really not willing to give any documentation.
Problem is clear – nVidia has no clear commercial iniciative to create good working drivers for Linux. Because it just spends their money and nothing good comes out of it anyway.
But they can’t open source their specs clearly for agreements with thirty parties about their copyrighted code and info, as they use lot of different chipsets/code/solutions in their cards (my pick is that NVidia code simply exposes inner design of card, so open it would be violation of some NDA)
And there is a problem. It is easy to create piece of hardware who follows AGP, PCI Extreme, whatever specifications and easily is manipulatable trough API interface, leaving inner details or card hidden. What is HARD is to create it cheap. And create so cheap that it is profitable.
Actually drivers who tries to do work their way in ANY operational system is cause of crashes, freezes, etc. so it is not only Linux problem. In contrary, Linux and BSD is only systems where you can try to workaround any binary only driver bugs. If someone’s motherboard gliches with Nvidia card in Windows, it is no brainer that you simply has to change a card – you CAN’T find a solution which could work there.
Don’t know…I have left Nvidia for good, ATI have some good supported open source driver driven cards, so I choose them. For now.
Edited 2005-12-07 13:23
nVidia has no clear commercial iniciative to create good working drivers for Linux
On the contrary, they want their cards to be used in the workstations used in movie and animation studios, etc.
Nvidia has explained in the past why they *can’t* release their drivers as opensource: the drivers contain code they licensed from third parties, over which they have no authority.
And the third parties will not open their code. Simple as that.
Nvidia has never in the past explained why it can’t release docs to the community so that we could create drivers ourselves.
On their newer families of cards, the answer is obvious. It’s an answer that I’ve already stated I understand and can respect. IP.