Weekly Edition
Return to the Press page
| |||||||||||||
Intel to speed lagging Linux support (ZDNet)
ZDNet has an
article on Intel's plan to improve its Linux support. "Theo de
Raadt, head of another open-source operating system, OpenBSD, steers people
away from Intel and toward the Prism wireless networking chips from
GlobespanVirata. 'Everyone in the open-source community is buying Prism
chip based (wireless network) cards,' de Raadt said. 'The chipset is fully
documented, and open-source drivers exist on all operating
systems.'"
(Log in to post comments)
Intel to speed lagging Linux support (ZDNet) Posted Feb 18, 2004 21:34 UTC (Wed) by alspnost (guest, #2763) [Link] About time. But it's not exactly going to make me warm to Intel. They will release closed-source drivers, and we all know the problems those cause. Forget it Intel, thousands of Linux users still won't be buying a Centrino any time soon.I've never been a fan of Intel, and I must admit I'm laughing now that they've made one of the biggest tech U-turns in history, by announcing a 64-bit extended Xeon. AMD can proudly claim the moral and technological victory this time, though sadly, not the financial one, now that Intel (and thus Dell) are on the bandwagon. Anyway, enough ranting....
Intel to speed lagging Linux support (ZDNet) Posted Feb 18, 2004 22:09 UTC (Wed) by log2 (guest, #10024) [Link] I don't buy Nvidia because of their closed-source drivers, but Idon't have a lot of company. If Intel does a good job with closed source drivers, people will probably whine, but not enough to get anything open-sourced. I don't like it, but that's my observation.
Intel to speed lagging Linux support (ZDNet) Posted Feb 18, 2004 22:20 UTC (Wed) by piman (subscriber, #8957) [Link] As someone else who doesn't buy nVidia, I find that's the case. A lot of people complain about Intel, but go right out and buy AMD-based nForce boards, which are just as bad (or worse).
Intel to speed lagging Linux support (ZDNet) Posted Feb 19, 2004 2:46 UTC (Thu) by ImpTech (guest, #2937) [Link] Whats wrong with nForce? The soundstorm support in ALSA is abysmal, but otherwise everything works as it should. Okay, okay, the ethernet driver had to be reverse engineered, but big deal, its done.
An aside re: nvidia Posted Feb 18, 2004 22:46 UTC (Wed) by einstein (subscriber, #2052) [Link] Someone said "I don't buy Nvidia because of their closed-source drivers"When I look for the best possible bang for the buck in linux friendly video cards, nvidia stands head and shoulders above the rest. Their cards are fast, and affordable, and they provide high quality, well-maintained drivers specifically for linux, and even support the 2.6 kernel at this point. I know I can go buy an nvidia card, download & install the nvidia 3D drivers and get good performance and stability, whether it's an old tnt/2, a geforce 5200, or a quadra card. It's hard to argue with a vendor who is obviously working so hard to win my business as a linux user who wants fast, affordable graphics performance. People complain that the drivers aren't GPL'd, but to me it's more a matter of quality and support - and the example of nvidia ought to give some of the more zealous nvidia opponents food for thought.
An aside re: nvidia Posted Feb 18, 2004 23:07 UTC (Wed) by dmaxwell (guest, #14010) [Link] I've had to disable SMT (hyperthreading) because the nvidia drivers apparantly aren't SMP safe. It would lock the keyboard and X11 at intervals ranging from daily to every five minutes. In any case, my uptime never exceeds 3 or 4 days with these drivers. Yes, I have kickass 3D but these drivers are by far the worst drivers I have ever dealt with on a Linux machine. I'll give Nvidia credit for trying but there is a quality argument to be made for opensource drivers.I can fully understand why they can't opensource their drivers. I don't see any good reason for them to hold back the register specs. At least the community could take a crack at producing its own 3D drivers.
An aside re: nvidia Posted Feb 18, 2004 23:24 UTC (Wed) by einstein (subscriber, #2052) [Link] You said "I've had to disable SMT (hyperthreading) because the nvidia drivers apparantly aren't SMP safe"That's bizzare, I have nvidia cards installed in about half a dozen linux boxes, and I've given them all a heavy workout with quake 3 arena, ut2003, lots of 3D screensavers, and I've never seen a crash - uptimes are up to a few months, basically the boxes all stay up until whenever I decide upgrade the kernel. I'd be curious to know what hardware you have, there are known issues with certain chipsets, and you most likley have one of the pathological cases. There's a wealth of info about this in the readme distributed with the drivers. The worst linux drivers? That prize belongs to the open source DRI drivers for ATI cards. I got rid of my ATI card because of a well known problem, I could get the whole system to instantly lock up hard, every time, by starting a game of Return to Castle Wolfenstein - OK, granted that was a couple years ago, so I recently installed linux on a new system with ATI Rage 128 video, and sure enough, as soon as the "atlantis" screensaver activated, the whole system locked up hard. I tried some BIOS options, but ended up just disabling DRI, which finally yielded stability. I told the customer that if he wanted high performance 3D, he should buy an nvidia card.
An aside re: nvidia Posted Feb 19, 2004 0:04 UTC (Thu) by ballombe (subscriber, #9523) [Link] > That's bizzare, I have nvidia cards installed in about half a dozen> linux boxes, and I've given them all a heavy workout with quake 3 arena, > ut2003, lots of 3D screensavers, and I've never seen a crash - uptimes > are up to a few months, basically the boxes all stay up until whenever I > decide upgrade the kernel. Did any of them ran with hyperthreading enabled?
An aside re: nvidia Posted Feb 19, 2004 0:31 UTC (Thu) by einstein (subscriber, #2052) [Link] You wrote "Did any of them ran with hyperthreading enabled?"Just this weekend I built a box for a customer, new hyperthreading P4 - FC1, running smp kernel, nvidia card and nvidia accelerated drivers - admittedly, that one has been shipped off to a customer, so I didn't run it for months, just a few days, but it was solid during that time. I am about to upgrade my main workstation with a HT-enabled P4-2800, and nvidia 5200, with, the nvidia drivers of course. I will report back on the success of that -
An aside re: nvidia Posted Feb 19, 2004 0:31 UTC (Thu) by CuZnDragon (subscriber, #5097) [Link] I haven't had a Rage card so can't comment on them but I have several Radeon cards from the Radeon AIW to the 8500 and haven't had a problem with the DRI. *shrug*
An aside re: nvidia Posted Feb 19, 2004 0:38 UTC (Thu) by einstein (subscriber, #2052) [Link] You wrote "I have several Radeon cards from the Radeon AIW to the 8500 and haven't had a problem with the DRI. *shrug*"That may well be, if you avoid using 3D screensavers, don't play any 3D FPS games, etc - but the kind of load I put on a system seems to invariably lock it up in short order if I use the combination of ATI card and accelerated 3D. I don't have that problem with nvidia cards, and come to think of it, the old Voodoo3s were solid too. I have friends running the newer ATIs with binary drivers, and they tell me the cards are stable (shrug) Of course, it's hard to justify spending $400 for a decent ATI card when I can get a hella fast nvidia card for under $100.
An aside re: nvidia Posted Feb 19, 2004 2:50 UTC (Thu) by ImpTech (guest, #2937) [Link] Nvidia's binary drivers are way more stable than ATI's for sure... but they work well enough once you get them working. I dunno about your $400/$100 comparison though. Everything I've seen says dollar-for-dollar you get more performance and better image quality from ATI... of course, since FSAA and S3TC don't work with ATI's Linux drivers, that changes the picture a bit.
An aside re: nvidia Posted Feb 19, 2004 1:34 UTC (Thu) by fLameDogg (guest, #11305) [Link] I have a Rage 128 Pro. I used to get lockups from time to time when doing certain video-related things in Slackware--mostly from trying to use MPlayer. I *think* this was Slackware 8.0, and I don't recall the version of MPlayer. But that was moons ago and newer versions of MPlayer have behaved nicely in newer versions of both Slackware and Mandrake. I did run a couple of Loki games and Tuxracer on that card with never an issue.
An aside re: nvidia Posted Feb 19, 2004 20:27 UTC (Thu) by allesfresser (subscriber, #216) [Link] I've always bought ATI cards because of NVidia's closed-source approach, and haven't had any regrets. I first had a Rage128, then a Rage128 Pro (different machines, one an Athlon 600, the other an Athlon XP 2200), and on my present machine I've run the Rage128 Pro, then a Radeon 7500, and now a Radeon 9200 variant (which is unfortunately only really supported in the XFree86 4.4 release candidates, but that's another story.)On all these machines and cards I've never had a problem with lockups when doing 3D work (in Blender, in Quake3 Arena, and, yes, with the atlantis screensaver. It's one of my favorites.) Now, I don't run atlantis all the time because frequently I have something going on in the background that I want to have using all the available CPU time and cache space (a POV-Ray render, for example.) But I've consistently got good results from ATI cards across the board. The single exception to this was when, as someone else reported, my power supply was going bad. Then I would get lockups fairly regularly when doing something that probably pulled a lot of power (like POV-Ray or heavy 3D usage or building from a large source tree such as Mozilla or XFree86.) Replacing the power supply solved the problem. Other flaky components (RAM, insufficient heat sink, etc.) could cause the same thing. <rant>And also, overclocking. Components are built with extra padding in their recommended specs *for a reason*: that is, they're over-engineered for reliability's sake. (I'm speaking of good components here, of course... cheap components are anyone's guess.) If you take away that margin of error by overclocking, by definition you're pushing the component to the edge of its capabilities and specs, and you certainly shouldn't be surprised if it (and anything it touches, such as the video card) starts being flaky. So you can say I'm not a fan of overclocking.</rant> :-) I also run vanilla kernel.org kernels, with everything I need compiled in (not as modules) and everything else left out. I don't run hotplugging (though it's a nice piece of work--I just don't need it) and I feel this gives a little extra bit of stability as well. And, I use Slackware, which also keeps the stability factor nicely elevated. All this to say, I don't think ATI deserves the rap they get sometimes for instability... if you buy good base components (PS, MB, RAM, CPU) and run them at the recommended speeds, you shouldn't have many problems (at least after the Murphy infant mortality phase.) Just my EUR 0.01571832. ;-)
An aside re: nvidia Posted Feb 19, 2004 20:33 UTC (Thu) by allesfresser (subscriber, #216) [Link] Sorry, forgot this: the latest machine is a P4-2.6, with a Gigabyte 8IK1100 motherboard (Socket 478-800MHz-Intel 875P-DDR), hyperthreading on and functioning with an SMP 2.6.2 kernel with ACPI, IO-APICs and i2c running.
An aside re: nvidia Posted Feb 19, 2004 1:41 UTC (Thu) by stephenjudd (subscriber, #3227) [Link] Hah! I have a Rage 128 card, and that same problem with the "atlantis" screen saver.FWIW, moving to a lower bit depth (15 or 16 bpp) solved the problem. This was almost a year ago, so possibly it's been fixed in the meanwhile anyway.
An aside re: nvidia Posted Feb 19, 2004 1:06 UTC (Thu) by jonabbey (subscriber, #2736) [Link] I use NVidia's 1.0-4496 Linux kernel driver on my Springdale Hyperthreaded P4 2.4ghz system. Q3, UT2003, UT2004, OpenGL screensavers.. no problems whatsoever, and I regularly have months of uptime between kernel updates.Perhaps you're running with flaky hardware? I had a dodgy power supply on my last system and I'd get random lockups whenever I started doing much of anything 3d on the Nvidia card. I also got lockups on sustained longterm hard drive activity. The new power supply fixed all of that right up.
An aside re: nvidia Posted Feb 19, 2004 1:30 UTC (Thu) by xorbe (guest, #3165) [Link] I get this also, sometimes just when the text scrolls in an xterm.It only seems to happen to my main box (XP), which is overclocked. My other machine, p4-2.8/800 seems to never lock up. ie, I think your conclusion is wrong about the cause.
An aside re: nvidia Posted Feb 19, 2004 7:37 UTC (Thu) by einstein (subscriber, #2052) [Link] You wrote " I get this also, sometimes just when the text scrolls in an xterm. It only seems to happen to my main box (XP), which is overclocked. My other machine, p4-2.8/800 seems to never lock up. ie, I think your conclusion is wrong about the cause."Every intel box I used with an ati card would lock up, every intel box I used with an nvidia has been stable. On the unstable intel boxes where I replaced an ati card with an nvidia, the box became stable as soon as the nvidia card was installed. Not much room for doubt IMHO...
An aside re: nvidia Posted Feb 19, 2004 0:36 UTC (Thu) by Eudyptes (guest, #15589) [Link] "People complain that the drivers aren't GPL'd, but to me it's more amatter of quality and support - and the example of nvidia ought to give some of the more zealous nvidia opponents food for thought." Well, IMHO it's a double edged sword. I just switched to an ATI Radeon 9600 XT. The firegl drivers are kludgy, at times very kludgy. But the cards hardware is very nice. I have gotten them to work but it's not for the tame of heart - or the newbie Linux users to say the least. Nvidia drivers in SuSE are much easier to deal with - yes they're closed source but between SuSE creating a specific modules-override so one can upgrade kernels without going nutz trying to reinstall the nvidia drivers, this with an installer - though not completely user-friendly, is much more like many ex-windows users are accustomed to. In otherwords, though the more purist Linux user has difficulty with these closed drivers, that's the mentality of corporate Hardware OEM's in the commercial end-user product space. So one is at the mercy of the companies commitment to a given market sector and competence for general support of it's product line. If Intel provides driver support that gives someone both well written and stable drivers with at least installation support on a cursory level then a fair amount of people will settle into accepting this. The alternative is to continually hack away at backtraces and fail.logs to create a working driver. Anyone having a non-sane supported scanner knows the hit or miss nature of driver development. Those trying to write drivers tend to do so for either/or device they themsleves use and can get some response from the vendor related to accomplishing this task. Sure open drivers are the preferred method by far. On the otherhand, if a closed source driver works as it should then the issue might be considered more academic than practical.
Intel to speed lagging Linux support (ZDNet) Posted Feb 19, 2004 0:12 UTC (Thu) by brouhaha (subscriber, #1698) [Link] I don't buy Nvidia because of their closed-source drivers,Same here. Won't buy their video cards, or motherboards usign their chipsets. I can at least somewhat understand their position on not documenting the video chips, even though I strongly disagree with it. But what's their excuse for refusing to document their ethernet interface? I've been buying ATI Radeon 7000 and 7500 cards. I gather that ATI isn't 100% forthcoming with documentation on their 3D support, but at least they document everything else. For instance, the X Video extension works well with these, so I can play MPEG video. That's a whole lot more useful to me that 3D anyhow. There are some closed-source application programs I run on Linux, but I'm not going to load binary-only kernel modules. That way lies madness.
Intel to speed lagging Linux support (ZDNet) Posted Feb 19, 2004 2:24 UTC (Thu) by elanthis (guest, #6227) [Link] There are now Open Source nvnet drivers, just as a note.So far as using binary only drivers... I guess it depends on what you do with your computer and your needs. I personally couldn't care less - even if I had the source, I wouldn't ever look at it; I probably wouldn't even understand most of it. I do care that I get the full power out of my $300 video card, however, for playing my binary-only games I exercised my Freedom to purchase, so my needs are (at this time) better filled by a the NVIDIA binary-only driver. Everyone has their own standards and needs. The great thing is, both you and I have the Freedom to make the choice based on our needs and desires. :)
Intel to speed lagging Linux support (ZDNet) Posted Feb 19, 2004 16:00 UTC (Thu) by mmarq (guest, #2332) [Link] " About time. But it's not exactly going to make me warm to Intel..."It aint going to "harm" intel neither... or make Intel deliver specs or open source drivers This is a war Open Source cannot win!... It can only hurt Linux/Open Source in a medium to long time.... Its about time for the community to get real! Open Source could have a win if there wasent any other alternative... or if a Open Source OS had near 90% of the market... IMHO a solution would be to harden and "close" the possibility of inserting external modules into the Linux Kernel... and replace that functionality by making it easy, the separate compilation of any single module code that lays inside the kernel... that would be about the same ting as external modules. In Australia conference, there was discussed the possibility of userland drivers!... IMHO not a good security alternative anyway. Create several "well defined and close to the hardware as possible" API/ABI for external modules and firmware, one for each subsystem functionality, that could be placed in Ring 1 of the processor(example).... This would be a split driver scheme( ala I2O) that would facilitate "reverse engineering", debugging and portability for other OSes, because the same binary (therefore source)module could be used in Linux, OpenBSD, Darwin or... ALSA and the Linux USB subsystem are close to a split driver module paradigma,... it only takes the will of the entire community and a "little more code" to make a *HUMAN INTERFACE*(in split driver model) from Video3D to sound and pointing devices that would put windows to shame. Hardware Industry not only aint going to reveal their microcode or VLSI designs,... they want more, like 6 months first time to market protection, and a little more "hide out" fat protection... Open Source could deliver that in a way that dont hurt Reverse Engineering, if need to, but also(IMHO) would facilitate the release of Open Source drivers code (older hardware) from the Hardware Industry...
Closed source drivers and GPL Posted Feb 19, 2004 11:09 UTC (Thu) by simlo (subscriber, #10866) [Link] A company can make closed source drivers and offer them for download. That won't brake the GPL as the driver isn't a derivative work of the Linux kernel.But can you ship Linux along with the closed source driver? Is "Linux + driver" a derivative work of Linux even though they aren't linked and the driver sits as a binary module on the filesystem?
Closed source drivers and GPL Posted Feb 19, 2004 12:41 UTC (Thu) by ekj (guest, #1524) [Link] Noone knows for sure. This hinges on the definition of "derived work". as per the copyrigth law. The details are fuzzy. It is clear that even if US law recognizes "mere aggregation" (i.e. distribution as a unit on CD) as sufficient to create a derived work, the GPL allows this. (more spesifically, it says that the GPL does not consider this to be derivation.)Even more unclear is the cases where someone distributes say a set-top-box of some sort, like the KISS dvd-divx-mp3-ogg-everything ethernet enabled players that actually run uClinux. Here the kernel itself is GPL (and freely avqailable form the vendor as the GPL requires), but they've made additional modules for driving the decoder and output chip, and for driving the LCD on the unit. One could argue that those drivers are in no way a separate creation, but rather an integrated part of a whole product. Afterall, the product as a sum is useless without them. We will see what comes of it.
Closed source drivers and GPL Posted Feb 19, 2004 16:48 UTC (Thu) by mmarq (guest, #2332) [Link] " But can you ship Linux along with the closed source driver? Is "Linux + driver" a derivative work of Linux even though they aren't linked and the driver sits as a binary module on the filesystem? "As long as the closed source driver is not a derivative work of Linux(otherwise it would have to be open sourced because of GPL), the answer is NO... *THOSE* closed source drivers could have licenses that dont permit general, or commercial re-distribution... If generalized, that could make the *Linux kernel owners* not the linux kernel developers but the companys that make the hardware and those closed source drivers modules... So its better to prevent (and harden) the possibility of external binary modules insertions, anyway,...(easly replasable IMHO).
Closed source drivers and GPL Posted Feb 19, 2004 20:08 UTC (Thu) by oak (subscriber, #2786) [Link] If the driver is developed specifically for Linux, then it's derivative work. If the driver exists already or the same driver works on multiple platforms (e.g. through wrappers), then the current concensus is that it's not derivative work. AFAIK.
Closed source drivers and GPL Posted Feb 20, 2004 12:14 UTC (Fri) by simlo (subscriber, #10866) [Link] That sounds a bit like SCO is arguing towards IBM...If I write a driver I must have a the copyright no matter if I intend to use it under Linux or not. Derivative work of Linux? Then it must include actual Linux source code. If it is a derivative work just because it is intented for Linux then RCU is indeed a derivative work of SYSv5 and SCO will win. Thus I can't see why I can't distribute the driver alone under any license I want no matter if I indended it for Linux or not. My original question was: Is "Linux + driver" a derivative work even though the two pieces of code isn't linked on the media I deliver it on?
Intel to speed lagging Linux support (ZDNet) Posted Feb 19, 2004 15:08 UTC (Thu) by josh_stern (guest, #4868) [Link] Thank goodness Linux users will now be able to have binary only support when they participate in Intel's attempt to further decommoditize laptops by confusing users with marketing piffle!!
Intel to speed lagging Linux support (ZDNet) Posted Feb 22, 2004 23:09 UTC (Sun) by iwilcox (guest, #18701) [Link] In mid-2000 I chose one ATI card (some variety of Rage 128) over another (can't remember what it was now) for its TV-out support. I was a MS Windows/Linux dual-booting user back then, and at the time I figured: what the hell, if TV-out doesn't work under Linux it'll still be fine under Windows.Now it's a few years on; I've long since purged all MS stuff (it was unlicensed anyway), and DVDs have really taken off. But I'm unable to use TV-out on my ATI card, despite valiant efforts by the maintainer of the "atitvout" package to reverse-engineer the card's interface - once enabled, the card refuses to yield anything but console mode to my TV or monitor. Attempts at using graphical modes give no signal to either TV or monitor from that point on and I have to reboot the machine from the console before the card will generate a (graphical) signal again. "atitvout" works for some other cards that are more similar to the one used by the package author, but r128 users are out of luck. I concede that mid-2000 was not a great time for vendor awareness of the need for open drivers, but there's nothing stopping them opening up the interface now. As a lone voice using fairly dated hardware, however, I'm not surprised my requests for documentation went unheard. Similarly, I recently bought a "Linux-compliant" PCI ADSL modem; while the drivers are partly open, they're for 2.4.1 (I run 2.4.2x) and aimed squarely at the "Linux means Redhat" market (I run Debian); while they do compile against my .2x kernel (with many warnings) they reliably cause kernel panics. The card works fine in a friend's Windows XP box, of course. I ended up using an OS-independent Ethernet-based modem. Friends tell me I should have chosen an Ethernet ADSL modem all along - but why should I have to? I have no recourse against the retailer that sold me the PCI modem (no distance selling, as it was in person, over a till). It's now just sat here in a box gathering dust. I've learned valuable lessons here: I won't be buying hardware by any manufacturer again unless I have at maintained open source drivers in my hand as I walk to the till, because I don't want to be left in any such situation again. By this point in this common discussion people have usually raised the trade secrets argument: "Vendors can't give out open drivers; that would compromise their IP". However, this sounds like a worthless argument when you talk to any industry insider and they tell you that reverse-engineering of rival vendors' drivers is not just commonplace but *standard practice*. So when I wirelessly network my house next month I know which vendor's name *I'll* be looking for on the box. Just another £0.02 for the pot.
|
|||||||||||||