NVIDIA has today introduced the GeForce Go 7950GTX GPU. This mobile GPU encompasses 24 pixel shader engines, a core speed of 575MHz, 512MB of video memory onboard, and more. This NVIDIA notebook GPU was introduced at DigitalLife 2006 in New York, NY. NVIDIA's press release for the Go 7950GTX can be found here.
NVIDIA has issued a public Beta display driver for Linux (x86 and x86_64), Solaris, and FreeBSD. The changes have been highlighted here. Most notably is support for GLX_EXT_texture_from_pixmap. The NVIDIA 1.0-9625 display drivers can be found at nZone.
NVIDIA has today unveiled the GeForce 7900GS and GeForce 7950GT. The 7900GS uses 256MB of video memory while the 7950GT uses 512MB; both of these new NVIDIA cards are SLI supportive. The prices for these cards start at $199 USD and $299 respectively. NVIDIA's press release can be found here. After NVIDIA delivers Linux-supportive drivers for these new cards (1.0-9XXX), we will likely be delivering GNU/Linux results for the NVIDIA GeForce 7900GS 256MB and 7950GT 512MB.
Roderick has passed along word that the second Beta is now out for NVClock. There are quite some hefty improvements from added GeForce 7 support to rewriting the core back-end and more. Be sure to check it out over at Linux Hardware. This third party NVIDIA Linux utility can be discussed in our Forums.
A thread has been opened up on the Phoronix Forums for speculating over the release date and possible features in NVIDIA's next Linux display drivers (1.0-9XXX series). It looks as if it will indeed be an August or September launch, and will contain X.Org v7.1 support, but we are still working on nailing down other likely additions. Feel free to share your thoughts or ideas.
NVIDIA has provided their first Linux legacy driver release. This driver is designed solely for older cards where support has been dropped from their official latest Linux display drivers. New in this release is installation problems on newer distributions, compatibility with the latest 2.6 kernels, improved 64-bit 2.6 stability, improved power management on newer kernels, and install NVIDIA OpenGL headers by default. More on this legacy release is at NVIDIA.
There has been some discussion in the community that NVIDIA SLI is capable of running on Intel's i5000XL Greencreek Chipset, and that ATI CrossFire could even run on the Intel Chipset with its dual PCI Express x16 slots. The Greencreek is designed to be a premiere workstation Chipset for use with Intel Xeon Dempsey and Woodcrest processors. However, using the Tyan Tempest i5000XL, we threw in dual PCI Express x16 NVIDIA GeForce graphics cards to see if Scalable Link Interface was really possible. Using nvidia-xconfig --sli=AFR, the SLI option was added to the xorg.conf, however, upon rebooting we were unsuccessful in our attempts of using SLI. The NVIDIA x86_64 1.0-8762 Linux display drivers were unable to enable SLI on the Greencreek. Though the next NVIDIA Linux driver release (1.0-9XXX) may bridge support for this Chipset, at this time we were unsuccessful in our attempts. If we learn any additional information on this matter, we will be sure to pass it along. Below is the error from the X.Org log. (**) NVIDIA(0): Option "SLI" "AFR"
In just a few hours from now NVIDIA Corporation will be doing a hard-launch of several new GeForce 7 products. These new products will include the NVIDIA GeForce 7300GT, 7900GS, and 7950GX2. As we had exposed yesterday, it does raise the question as to whether we will see new NVIDIA Linux display drivers this morning. While we will not be delivering any launch previews this morning or reviews of the new products, we do anticipate that some will arrive at our testing facilities shortly. In the mean time before the launch of the products, we have begun to receive advance press releases from some of NVIDIA's partners. Below are two Biostar GeForce 7300GT products. Throughout the days, and in the coming days, we will be sharing more information, as well as our usual Linux coverage.
Announced today by NVIDIA is EPP memory, or officially known as Enhanced Performance Profiles. Corsair has already issued a press release stating the immediate adoption of Enhanced Performance Profiles. More on this is at NVIDIA. SANTA CLARA, CA—MAY 15, 2006—NVIDIA Corporation (Nasdaq: NVDA), the worldwide leader in programmable graphics processor technologies, today announced the results of a collaborative memory development effort called Enhanced Performance Profiles, or EPP, which allows consumers to easily expose new, advanced performance memory settings built into high performance memory DIMMs for even higher levels of overall PC system performance. In addition, NVIDIA also announced today that Corsair Memory, the worldwide leader in design and manufacture of high performance memory, is the first memory vendor to announce support for the new EPP specification, and is expected to have new EPP-based memory DIMMs available in the channel later this month.
Tomorrow, May 15, NVIDIA is expected to launch the GeForce 7900GS, 7300GT, and 7950GX2. However, it does raise the question as to whether we can expect new Linux drivers to accompany this green hardware. Up to this point, our sources at NVIDIA have remained mute on the matter, however, if there were to be a driver release this week from the green it would partially align with their usual 4 month/3~4 week release cycle. In the past NVIDIA has provided some same-day Linux support, primarily with the NVIDIA GeForce 7800GTX 256MB G70 part. At this point, however, a driver release this week from NVIDIA is nothing more than a professional speculation. The next NVIDIA Linux display drivers out aren't expected to contain anything more than new hardware support and bug fixes. ATI is also quickly approaching with their monthly drivers. Upon seeing any signs of an impending NVIDIA Linux driver release, we will be certain to pass along the details.
With NVIDIA's 1.0-8756 Linux display driver having come out exactly one month ago, it should be about time for another driver release. For a while now, NVIDIA has been on the approximate 4 month/3~4 week release cycle for their Linux and Solaris (occasionally FreeBSD will be added to the mix) display drivers. Of course, the 4 week release cycle in between the major releases generally only contains an improvement or two, and then bug fixes for problems introduced in the last major release. However, this time around it seems that NVIDIA's Linux display driver launch may be pushed back a bit. According to a source, the next Linux driver installment may not be introduced until Computex Taipei of this year. The buzz this year at Computex Taipei will certainly be AMD's Socket AM2 among other things, and it just so happens that new Linux drivers from the green may occur during this Taiwan show. The Phoronix coverage from Computex Taipei 2005 is available here. ATI of course has yet to come out with its May refresh from the Linux and Windows display drivers. We hope to soon acquire more information on details related to the upcoming NVIDIA Linux drivers.
For those pondering whether to use Forcedeth or Nvnet for use as the ethernet driver within Linux on NVIDIA-based systems, there is an active discussion going on over at NvNews Forums. The points brought up so far include NVIDIA engineers working on the open-source Forcedeth project, netio benchmarks, and various other real-world comments.
Not only has ATI Technologies come to the table with their new FireGL workstation cards today, but NVIDIA has also unleashed a new professional graphics solution. The Quadro FX4500 X2, as NVIDIA has called it, utilizes dual GPUs each with 512MB of video memory. Its memory interface is 256-bit and the memory bandwidth is 33.6GB/s. Other features include two dual-link DVI-I connectors, a stereo connector, and SLI, SDI and G-Sync capabilities. If NVIDIA is continuing in their approximate 4 month/3 week release cycle, we should expect to see new NVIDIA Linux drivers within a couple of weeks from this point -- and we certainly do expect support for the NVIDIA Quadro FX4500 X2 in the upcoming release. More details for the Quadro FX4500 X2 are available at NVIDIA.
SANTA CLARA, CA—APRIL 18, 2006—NVIDIA Corporation (Nasdaq: NVDA), the worldwide leader in programmable graphics processor technologies, today unveiled the new NVIDIA GeForce Go 7900 Series of graphics processing units (GPUs), which enable gamers to experience vivid, extreme high-definition (XHD) games on notebook PCs—a dream come true for gamers that demand performance and portability. The entire NVIDIA GeForce Go 7900GTX and GeForce Go 7900GS press release can be viewed here.
Today NVIDIA has unleashed the Linux and Solaris 1.0-8756 display drivers. No FreeBSD drivers to compliment this release are out at this time. Our initial impressions with the NVIDIA 1.0-8756 drivers are available here.
With the new Linux (and Solaris) driver release just hours ago, this serves as the inaugural release for 2006 (unless of course, counting the 1.0-8183 display drivers). This NVIDIA 1.0-8756 release fits in line with NVIDIA's Linux release cycle that has become pretty much standard -- being an approximate 4 months/3 weeks/4 months, with the previous 1.0-8178 release falling towards the end of the December. This also means we should anticipate another driver release towards the end of April or early May. The next minor release may coincide with the X.Org 7.1 launch. After that, we are likely to see no major releases (at least those made available from nvidia.com) until September or October of this year. Of course, this is simply a general overview and the release cycle is always capable of changing. The upcoming minor release should also mark the end of the 1.0-8XXX series. Presently the NVIDIA Linux drivers are at 1.0-8756 while the Windows ForceWare are officially at 84.21 (WHQL). In this time until the next major release, we are anticipating that the 1.0-9XXX series will make its premiere. When the Rel80 1.0-8XXX launched, there were a great deal of attention grabbing features implemented for the Windows drivers, and a handful for Linux and Solaris. Some of the highlights included Scalable Link Interface (SLI) support as well as the nvidia-xconfig utility. When the NVIDIA 1.0-7XXX drivers were introduced, they were also home to a great deal of changes. Of course, NVIDIA isn't in the mood to squeal all the details early, however, we have a few conservative speculations as to what the 1.0-9XXX series may hold in store for alternative OS users. For one, the NVIDIA installer is long over-due for an overhaul. If you will recall, the installer in its current form was initially delivered with the 1.0-4349 drivers on March 31, 2003. While ATI began offering their display drivers as RPMs, with their v8.14.13 release last year they turned to a new graphical installer option. This new installer not only made it unnecessary to drop to stop X in order to run the installer but it also features the capabilities of generating distribution specific drivers. In fact, almost with each new release additional distributions are added -- ATI has done any EXCELLENT job with getting distribution vendors to join their Beta program and to provide installation compatibility. With that said, we imagine NVIDIA will finally take on ATI with a graphical (likely GTK based) installer in the Rel90 drivers. NVIDIA also has room to make improvements with its NVIDIA Settings utility. While nvidia-settings may be more robust than fireglcontrolpanel (at this present time), the Linux drivers are of course a long way from catching up with the Windows options. Both companies do, however, have viable text-based configuration utilities. On top of that, the OpenGL GLX_EXT_texture_from_pixmap extension will be finalized shortly to allow for Fedora Rendering Project AIGLX compatibility. While NVIDIA appears to have no large incentive for improving the quality of its SLI support, the 1.0-9XXX series may be home to a couple of multi-GPU related changes. If NVIDIA does stand by their four month escapades, the 1.0-9XXX drivers should also contain support for new GPUs and products to be released later this summer. Today's 1.0-8756 release also contained improved power management support on recent kernels, while ATI has largely been leading the way with Linux mobile advances. ATI had begun by porting PowerPlay to Linux, as well as other power-saving features, and they will likely continue to implement additional advancements for mobile users. One feature that many users had also hoped to make their way into 1.0-8756 drivers was H.264 hardware support and other NVIDIA PureVideo features. Depending upon how ATI plays their Linux cards with Avivo, additional video and display options could likely find their way into the 9X series. Finally, the NVIDIA Linux 1.0-9XXX series will likely bring to the table a few other features, more information is to come. It is important to reiterate that the information today has not been officially confirmed by NVIDIA, however, these are relatively safe speculations as to what features will make their way into NVIDIA's next Linux software installment due out later this year.
Over at NvNews Linux Forums (the hang-out for NVIDIA representatives attempting to help its users) netllama -- or better known as Lonni J Friedman -- of NVIDIA's Linux department has finally decided to state a better release time-frame than their soon statements. According to the message, it is safe to say the driver will be available by April 15. As we had mentioned numerous times, they have been aiming for an early April release -- and this latest public word seems to definitively say the driver will be out this week, or next week. Next week of course is the second week of the month, or ATI's usual monthly driver release period for both Windows and Linux. For reference, netllama's post is available here.
With our continual investigation of NVIDIA's private Beta drivers (1.0-8751) for Linux graphics solutions, we have yet another small detail to share. Back on August 1, 2005 we had reported to NVIDIA a bug that would cause CoolBits from operating on Xinerama-enabled systems. When one NVIDIA graphics solution is installed and Xinerama is enabled to drive two display heads, and then attempting to alter the X configuration with Option "Coolbits" "1" would result in CoolBits not activating. When NVIDIA reproduced our bug, they opened up the issue as bug 181974 so their development team could look into such a problem. Well, precisely 35 weeks later or 245 days and three public driver releases, this bug has yet to have been addressed. With NVIDIA's TwinView enabled for a multi-headed display, CoolBits is known to work fine but we have just confirmed this morning that the 1.0-8751 drivers still are unable to run CoolBits when Xinerama is activated. While this issue isn't a problem for most consumers, it is another bug that has yet to be addressed almost a year later. The next Linux/FreeBSD/Solaris public driver release is expected soon.
As we have been sharing time and time again in our news posts and articles that NVIDIA's inaugural Linux driver release for 2006 is just around the corner -– or should we say the first release geared towards consumers --- seeing as there was the unmentioned 1.0-8183 release earlier this year that were simply targeted for Hewlett-Packard's workstations until Phoronix had made public this discovery. While we had these NVIDIA Beta drivers for some time now, and have had the privilege of sharing the discovered information since the GeForce 7900 launch, we still have yet to come across the official change-log. However, if you will dig back into some of our previous 1.0-8751 related articles, many of the changes and new features can be spotted. While we have yet to be officially addressed as to the public launch date for NVIDIA's upcoming Linux (and likely Solaris/FreeBSD) drivers, we have a few remaining thoughts to share today. Our last non-confidential briefing on this matter with NVIDIA has yielded that the drivers are still anticipated for an early April release and that the development process for these drivers should have ended at the end of March. What does this mean? Likely we will see these new NVIDIA drivers come about within this upcoming week. Or rather the week of April 2nd to the 9th. With last week's launch of the NVIDIA Quadro FX 350M, 1500M, and 2500M it simply reinforces that new drivers will be imminent in order to officially support these new workstation GPUs. Tracing back the origins of NVIDIA's alternative OS drivers all the way back to 1.0-1251 -- seeing as NVIDIA has no clear strategy unlike ATI's monthly period -- the Linux drivers are most often released on Thursday, followed by Wednesday and Monday. Upon hearing or discovering additional information we will be sure to share it with our readers. For those wearing red, the April emancipation will be the first driver this Spring -- which may prove to be beneficial if you have been tuning into some of our past articles.
While we have been covering the NVIDIA and NVClock updates for almost the past month now when it comes to primarily the GeForce 7900 support, we have a few details to pass along today when it comes to new code merged into the NVClock SourceForge CVS public server in roughly the past week. To begin, a few init script table tokens were added for some GeForce 7 series GPUs and the performance table parsing was re-done. Other changes include adding several new GeForce 7 device IDs, basic support, and F73373S support. Additional NVClock code changes are expected to occur soon. No word yet when other features may be added. Well, with it being the start of April -- NVIDIA's Linux display drivers can be anticipated for a launch at anytime now (come on, just a little bit more waiting :), for real). NVClock CVS information is at SourceForge.
With the discovery yesterday of the NVIDIA Linux 1.0-8183 display drivers on HP servers, we have now had some time to complete a few tests that are designed to distinguish any performance differences between these drivers in just a few of our traditional array of Linux gaming and workstation benchmarks. The NVIDIA Linux drivers used were the latest official ones supported by the green folks at this time -– 1.0-8178 -– and then these newly discovered 1.0-8183 drivers that were built late February and finally the 1.0-8751 drivers. The 1.0-8751 drivers at this time are not publicly accessible, however, we at Phoronix have managed to obtain these drivers for media coverage. These 1.0-8751 Beta drivers were built in early March and contain support for the new GeForce 7 solutions as well as other enhancements, which we have been documenting at Phoronix since we got our hands on them. Look forward to NVIDIA's next official driver step to come in early April –- unless of course a delay ends up taking place. The system configuration used for these miniature tests was a Fedora Core 4 installation with the 2.6.15 x86_64 SMP Linux kernel. The hardware components consisted of an Intel Pentium D 820 processor running at 2.80GHz, 2 x 512MB DDR2-667, Western Digital 160GB SATA2 hard drive, eVGA e-GeForce 6800GT 256MB PCI Express, Swiftech CPU water cooling, and a Tyan Tomcat i7230A S5160. The Tyan i7230A is based upon Intel's E7230 MCH + ICH-7R Mukilteo Chipset. The tests simply used for this short testing extravaganza was Enemy Territory (traditional Railgun demo) and SPECViewPerf v8.1 with the 3dsmax-03, catia-01, and ensight-01 sets. Additional testing will be performed upon the official release of these updated NVIDIA drivers. ================================ Enemy Territory v2.60 (Railgun) 800 x 600 ================================ 1.0-8178: 66.9 1.0-8183: 66.8 1.0-8751: 66.5 ================================ Enemy Territory v2.60 (Railgun) 1280 x 1024 ================================ 1.0-8178: 63.2 1.0-8183: 63.2 1.0-8751: 63.2 ================================ SPECViewPerf v8.1 3dsmax-03 ================================ 1.0-8178: 4.328 1.0-8183: 4.326 1.0-8751: 4.336 ================================ SPECViewPerf v8.1 catia-01 ================================ 1.0-8178: 3.300 1.0-8183: 3.301 1.0-8751: 3.309 ================================ SPECViewPerf v8.1 ensight-01 ================================ 1.0-8178: 7.189 1.0-8183: 7.210 1.0-8751: 7.196 It looks as though any performance benefits have taken a back seat to appending new features and largely addressing existing bugs. Of course, these benchmarks today only stress a small portion of the available Linux testing areas and only on a single system setup. Upon these new official drivers, as always, we will continue our investigation to see how the new drivers shall fair overall and in different environments. If any readers encounter any other changes in these 1.0-8183 drivers, feel free to report them to us using our contact page.
As we had delivered this news in our NVIDIA 1.0-8183 Display Drivers article, it seems as though Hewlett-Packard is presently offering NVIDIA display drivers with a version of 1.0-8183. This version is in contrast to the presently available 1.0-8178 drivers that have been floating around the Internet since last year. These Linux display drivers can be found from navigating Hewlett-Packard's driver section and searching such a workstation as the xw9300. There are x86 and x86_64 driver pages available. The downloads for x86 and x86_64 drivers are in RPMs. However, following the above-linked article, will go through the extraction process to get an end product of the 1.0-8183 Linux display drivers with the universal installer. As these x86/x86_64 1.0-8183 Linux display drivers are freely and publicly available from Hewlett-Packard's website, we have mirrored the .run files here at Phoronix. These files can be obtained from the above hp.com links and then extracting the RHEL 3/4 package. Phoronix.com provides absolutely no form of support or warranty on these hosted files and are intended to run at YOUR OWN RISK, and are not officially supported by NVIDIA Corporation. The download links are listed below as well as the MD5 sum of the compressed file. NVIDIA-Linux-x86-1.0-8183-pkg1.run 3fd345b4517cddb7c7f137e7620e01f3 NVIDIA-Linux-x86_64-1.0-8183-pkg2.run c576a5ad17e7c9822db4033ee653fe14
Here at Phoronix we have been covering the Scalable Link Interface support under Linux since its launch with the inception of the 1.0-8174 display drivers back on December 5, 2005. While this NVIDIA SLI support can still be considered very much rudimentary compared against the Microsoft Windows support with the ForceWare drivers, which were introduced back on November 9 of 2004, there is no clear sight for how it will ultimately fair in the world of Linux. According to some information we have obtained from our sources and research, NVIDIA's motives for Linux SLI may largely dissent from the public opinion. In this article today, there are a few comments we would like to share about the big green manufacturer and their outlook on alternative operating systems. This article in its entirety can be read HERE.
With an inquiry from one of our readers (Tako Schotanus), we spent some time today looking at the suspend/hibernation support of NVIDIA's private 1.0-8751 Beta drivers. With the 1.0-8751 drivers, we used Fedora Core 5 and the 2.6.16 kernel. With the tests we have done thus far, it appears that the suspend and resume support may be improved beyond past proprietary NVIDIA driver releases -- but it may partially be due to improvements in Fedora Core 5 Bordeaux. We had called the system into suspend mode several times and in various scenarios. In this testing, we have yet to come across any specific problems pertaining to Linux suspend with NVIDIA's proprietary 1.0-8751 drivers. For those that had missed our latest posting regarding this matter, we are still anticipating NVIDIA will release these drivers to the public towards the beginning of April.
Up to this point there has been very little official information from NVIDIA in regards to their upcoming Linux/Solaris/FreeBSD drivers. Luckily, we have had the privilege from NVIDIA to share our acquired information a bit early in regards to these upcoming drivers, and a few NVIDIA Linux developers frequently post some information on the NvNews forums. We have also had the luxury of handling the Beta 1.0-8751 drivers and deliver its results on multiple occasions. We continually post new information at Phoronix as it is discovered from our in-house investigations with these drivers, as well as tapping additional information from our various sources. For the most part, there has largely been two discussed release dates for these drivers as stated by Phoronix -- March 09, 2006 and March 22, 2006. Prior to NVIDIA's CeBIT GeForce 7600/7900 launch event, we had thought they would provide same-day support for their Linux users when handling this newer hardware. However, this wasn't the case unlike the 7800GTX 256MB launch last year and the 1.0-7667 drivers. The other largely contemplated release date was today, March 22. With that said, today NVIDIA is expected to release a minor Windows ForceWare update (v84.25). These Windows 2000/XP drivers are expected to contain enhancements for the recently released Oblivion game. However, we have been finally notified by NVIDIA that the Linux drivers will not coincide with the ForceWare launch that is expected to occur today. Rather, we have been told by NVIDIA's Sean Cleveland that they are anticipating an early April release. With that said, the public can expect to see these new Linux (and likely Solaris/FreeBSD) drivers appearing sometime during the first or second week of April. The NVIDIA drivers will likely strike in advance of ATI's monthly Windows/Linux CATALYST driver updates. If NVIDIA continues in their roughly 4 month/3 week Linux release cycle, we will also likely see yet another driver set coming in late April or early May. However, much of this planning is still likely up in the air. For those pondering over the Fedora Core 5 support with NVIDIA's drivers, the next NVIDIA Linux driver release should produce the same level of support that can be found with Fedora Core 4. Using the private NVIDIA 1.0-8751 drivers, we had no strongholds preventing the module from operating. As Fedora Core 5 had mistakenly shipped with a kernel which doesn't allow non-GPL modules to load, we had first upgraded to the Linux 2.6.16-1.2064_FC5 (x86_64) unofficial kernel from Dave Jone's Fedora server. We also had to attain xorg-x11-server-sdk (v1.0.1-9.x86_64) for proper NVIDIA compatibility. With the updated kernel and the sever SDK, we dropped to run-level 3 and proceeded with the traditional NVIDIA installer. With no signs of errors during the installation, the X configuration needs to be modified, or simply run nvidia-xconfig. Restarting the system, the NVIDIA GeForce card was accelerated by NVIDIA's proprietary drivers under Fedora Core 5 Linux (2.6.16 kernel) with no apparent signs of problems. The installation process should be similar to the public drivers upon their release. All of these procedures were carried out on a fresh Fedora Core 5 DVD installation. This new process is quite contrary to the 1.0-8178 drivers that require patches or using the drivers available from the Livna repository. More information to come...
Yet some more items to throw out to the Linux community in regards to NVIDIA's upcoming display drivers. Further testing has revealed that these drivers we have tested (1.0-8751 -- Beta) may contain some slight improvements to the Transgaming Cedega performance when it was tried in Half-Life 2. Outside of the Windows gaming emulation, there has not been much else that we have found when it comes to performance improvements in other Linux-native games, nor nothing in the way of noticeable performance drops. With one of the NVIDIA GeForce 7800GTX G70 512MB solutions that we had used for this Beta testing, we had found the ambient temperature no longer being displayed, while other cards with nvidia-settings have faired appropriately. While NVIDIA has yet to come out and officially address the Linux community as to the status of these upcoming display drivers, and we have been stating here at Phoronix that the drivers are a "few weeks away", we believe the release may be quickly approaching. With many of NVIDIA's past driver launches, they have coincided with the release of the Windows ForceWare drivers. These simultaneous releases are similar to ATI and their monthly ritual of delivering out new display drivers for both Windows and Linux. Anyhow, we have begun to receive reports today that this Wednesday, March 22, 2006, NVIDIA will be introducing yet another ForceWare release. This new ForceWare released (dubbed 84.25) will largely address performance issues for a Windows game coming out tomorrow entitled Oblivion. Whether NVIDIA will compliment this new ForceWare release with these updated Linux (and possibly Solaris/FreeBSD) drivers, we do not know for certainty at this time but it's yet another possibility.
We have just been notified by NVIDIA's Aaron Plattner that there are a few corrections to be made when it comes to NVIDIA's upcoming Linux driver release. First off, the GLX_EXT_texture_from_pixmap extension has yet to be finalized, and thus it will not be supported in the upcoming Linux driver release. What this means is no AIGLX under NVIDIA for quite some time, unless things change. As NVIDIA's proprietary drivers do not use DRI, they are waiting upon the GLX_EXT_tfp extension to be finalized and upon the preceding driver release it will support the ability to run Compiz and the AIGLX project's updated version of Metacity. Other than the few clarifications to note, we are still working with the 1.0-8751 Linux display drivers.
Yet some more NVIDIA 1.0-8751 result discussions... Shall we? One of the key concerns for NVIDIA Linux users has been to the support of the GeForce 6100/6150 + 410 MCP. On paper, and perhaps under Windows, the integrated video component may be an excellent choice for HTPC and video enthusiasts, but under Linux (1.0-8178) there has been numerous reports of user problems. We are in the process of examining the NVIDIA GeForce 6100 using a ASRock 939NF4G-SATA2 in conjunction with these 1.0-8751 drivers. We hope to have some performance (and stability) tests completed within a few days and will deliver some of these notes. We will also be looking at the possibility of the XvMC support. Receiving a great deal of user feedback in regards to what they hope to see in NVIDIA's upcoming display drivers, we are running a few other examinations right now in preparation. One of the items we can confirm still exists (which may partially be due to KDE developers) is 2D slowness when using KDE (Qt) applications. With the 1.0-8751 drivers, we used a GeForce 6600GT and KDE v3.4 and when enabling sub-pixel rendering we still encounter a noticeable slowdown when dealing with such KDE applications as KWrite. On a different topic, so far we have not experienced much in the way of any visual artifacts or stability issues when it comes to testing a handful of different Linux systems and various GeForce FX, 6, and 7 graphics cards. Fedora Core 4 with the various updates continues to fair well with these drivers but when it comes to Fedora Core 5 we have noticed a few issues when building the kernel module as well as related problems when running the NVIDIA installer. These issues, however, may be fixed upon the March 20 launch of Fedora Core 5 and we will be sure to check back on its progress at that time. The Fedora Rendering Project has stated NVIDIA will add the GLX_EXT_texture_from_pixmap OpenGL extension in their next binary release, however, in this 1.0-8751 version this doesn't appear to exist, which will append support for AIGLX or Accelerated Indirect GL X.
As we are fortunate with this NVIDIA Beta release to be able to share the relevant information a bit early prior to the public launch (thanks in part to NVIDIA/EVGA as well as the CeBIT launch), we have a few more details to share this afternoon. While we are holding back on delivering our official driver briefing until the official driver launch, we ran a few tests this morning with the 1.0-8751 package and then again with the current 1.0-8178 drivers. The hardware setup remained the same for both of these driver tests and the graphics card used was a normal GeForce 6600GT with 128MB of video memory. The Linux distribution was Fedora Core 4 with the 2.6.15 kernel. Providing a very rough comparison between these two driver releases was RTCW: Enemy Territory v2.60 with the Railgun time-demo that we have been using for years. For this news posting, we simply used the high-quality visual settings with the in-game configuration. We also refrained from using any AA/AF visuals. The resolutions tested were simply 640 x 480 and 1280 x 1024. Using the 1.0-8178 drivers the average frame-rate was 107.8 and 106.1 FPS, respectively. Using the private 1.0-8751 drivers, the results were 108.2 and 105.1 FPS. Interesting, eh? As a very preliminary generalization, the performance aspect of these drivers do not appear to be a miracle frame-rate wise; however, the results right here are simply from a single Linux-native program. Additional tests will come with our official briefing upon the official driver launch. On a side note, the memory issues that have been repeatedly reported since the Rel80 drivers (the ones in which NVIDIA had originally denounced) appear to have been addressed. The investigation still continues...
While the new NVIDIA drivers have yet to surface in the hands of the public, we have been fortunate enough to obtain a pre-release (Beta) copy of these soon-to-be-released drivers. While no official change-log is available, we have begun our exploration of these new NVIDIA Linux display drivers. Luckily NVIDIA has allowed Phoronix to share our results at this time, we simply can't distribute the package. One of the first areas we examined was was the nvidia-settings Linux control panel. While we hadn't visibly seen the changes we were hoping for (such as any Scalable Link options or toggle items for the various other features we were anticipating), there are a few minute changes that can be found in this next release. Below are a few images taken today from the NVIDIA 1.0-8751 Beta Linux display drivers from a GeForce 6600GT SLI setup. The new setting on the X Server XVideo Settings page is the Sync to this display device, and then lists the various monitors. When running Linux SLI, the Enable SLI Heads-Up-Display is still present under the OpenGL area. Unfortunately, this appears to be the lone location in nvidia-settings for toggling any SLI options, unlike the Windows drivers that allow profiling, etc... Another noticeable area that is in need of reworking is the thermal monitor page when running multiple graphics cards, or SLI. The slowdown threshold, core temperature, and ambient temperature are displayed; however, there isn't a separate area for each of the respective GPUs. Moving onto the display device section, the monitor does appear to now be properly identified (i.e. Acer AL1714) rather than previously displaying CRT-0. Opening up the display device screen, image sharpening was the only slider-bar option as it appears the digital vibrance ability has been removed. Keep in mind, these images and text above are based upon the 1.0-8751 display drivers which are private Beta drivers, and some of these options may change between now and the public availability of the next driver set.
607 NVIDIA news articles published on Phoronix.