While NVIDIA's NVISION 3D conference is taking place this week, we have yet to hear anything in regards to Linux announcements. However, NVIDIA has updated its open-source driver. Sadly though, their open-source driver is still the xf86-video-nv driver with its 2D limitations and obfuscated code. The xf86-video-nv 2.1.11 release has its hardware cursor initialization code rearranged, ROP bug-fixes for the G80 series, logging the i2c G80 port number, fix some warnings, a pixel clock setting fix, and adding in more and missing chip names. The release announcement for xf86-video-nv 2.1.11 can be read on the X.Org mailing list.
Last week OpenGL 3.0 and GLSL 1.30 were released during SIGGRAPH 2008. In that article we expected NVIDIA would be first to provide an OpenGL 3.0 driver due to NVIDIA's involvement with the Khronos Group and their much rumored "Big Bang II" project, but we had no official comment from them at the time of publishing. Though we now have a response from NVIDIA.
While NVIDIA's newest GPU family is the GTX 200 series, they aren't yet introducing the GTX low-end models but are continuing to mature the GeForce 9 arsenal. Introduced this morning by NVIDIA Corporation was the GeForce 9500GT, which is a sub-$100 product.
If you're using an older NVIDIA graphics card and haven't yet turned to the reverse-engineered Nouveau driver, you may want to check out the latest NVIDIA legacy drivers. NVIDIA has updated its two oldest legacy drivers -- the 71.xx.xx and 96.xx.xx series -- for Linux. The new 71.86.06 driver for Linux fixes a bug that resulted in black-and-white video output on secondary TVs. The other new driver, 96.43.07, has this TV fix too along with improved nvidia-xconfig parsing when using a xorg.conf file that lacks a module section. Both of these new drivers also have support for the Linux 2.6.26 kernel. The NVIDIA 71.86.06 and 96.43.07 Linux drivers can be downloaded from here and here, respectively.
It's been seven weeks since the obfuscated xf86-video-nv driver was last updated, so this afternoon NVIDIA's Aaron Plattner decided to release version 2.1.10, which will be in time for X.Org 7.4. However, before getting excited that they may actually be doing something good with this open-source driver, it continues to be limited to 2D acceleration and of limited use to someone that just laid down several hundred dollars for a graphics card or two.
Yesterday we reported on the Linux Foundation's message they have issued on the behalf of more than 140 kernel developers: Binary-only kernel modules are harmful and undesirable. While no vendor was singled out in this message, the biggest hardware manufacturer that has yet to provide any real level of open-source support is NVIDIA Corporation.
NVIDIA has released the 173.14.09 binary display driver. However, support for the GeForce GTX 200 series is not part of the change-log, unfortunately. This driver is just a bug-fix release for the recently released 173.14.05. Though this driver does have the following fixes: aliased font rendering corruption on X.Org Server 1.5, display corruption problem driving two dual-link DFPs with the Quadro FX 1700, a regression that prevented the X driver from starting on some GeForce FX/6/7 series, and fixing a locale-interaction issue in the nvidia-settings parser. In addition, this release also has preliminary support for the Linux 2.6.26 kernel. To download the x86 and x86_64 NVIDIA Linux drivers, click here and here, respectively.
This morning NVIDIA had unveiled the GeForce GTX 200 series and already NVIDIA's Aaron Plattner has committed support for these cards to the open-source xf86-video-nv driver. This support just involved adding in two new PCI IDs for the GeForce GTX 260 and GTX 280. No other work was needed, since of course this driver is limited to mode-setting and 2D acceleration. Though if you've just shelled out $399 or $649 on a graphics card, changes are you will want to stay clear of this obfuscated driver. If you want to check out this driver, you'll need to obtain it from git on FreeDesktop.org.
As expected, today NVIDIA unveiled their GeForce GTX 200 family of graphics processors. At this time their family is a bit small with only the GeForce GTX 260 and GTX 280 being shown off, but the GTX family should grown soon. These GTX GPUs support CUDA (with PhysX support), second generation NVIDIA unified architecture, 3-way SLI Technology, and PureVideo HD. The NVIDIA GeForce GTX 260 is made up of 192 processing cores, 576MHz core clock, 999MHz memory clock for its 896MB GDDR3 memory on a 448-bit interface. The fastest GTX processor right now, the GTX 280, has 240 processing cores, 602MHz core clock, 1107MHz memory clock for its 1GB GDDR3 memory on a 512-bit interface. Right now there is no NVIDIA Linux driver to support these next-generation GPUs, but once there is we'll let you know along with providing a performance run-down and other analysis. The GTX 260 costs $399 USD while the GeForce GTX 280 will set you back $649.
This past week NVIDIA had unveiled the GeForce 9 Mobile GPUs at the Computex Taipei trade-show. The GeForce 9M GPUs were announced just before AMD had rolled out its Puma platform with the fastest ATI mobile graphics ever and the introduction of XGP, which is a PCI Express 2.0 technology for allowing external graphics cards to be used with this new notebook platform.
From Computex Taipei, NVIDIA has announced the GeForce 9 Mobile GPUs. NVIDIA claims these next-generation mobile GPUs are 40% faster than their current GeForce 8 mobile processors and 10x faster than IGPs. These low-power GPUs also support PureVideo HD with full support for Blu-ray. When it comes to connectors, these GPUs are designed for NVIDIA's new MXM v3.0 module specification and from the outside they can support DVI, HDMI 1.3, DisplayPort 1.1, and analog VGA connections.
Following the introduction of Intel's Atom MID (Mobile Internet Device) processor family as part of the Menlow platform, VIA had introduced the Nano Processor Family just last week. Today at Computex Taipei, NVIDIA has announced their own mobile processor. NVIDIA's family of mobile processors is called Tegra and is currently made up of the Tegra 650 and Tegra APX 2500, which are planned for use with Windows smart-phones. The Tegra computer-on-a-chip architecture is made up of an 800MHz ARM CPU, HD video processor, and ultra low-power (and low-end) GeForce GPU. More information can be found in the NVIDIA Tegra Press Release.
In addition to NVIDIA releasing the 173.14.05 Linux/FreeBSD/Solaris driver this past week, they've made an additional software move. NVIDIA has decided to release Gelato Pro, which previously costed $1,500 USD per node, as now a free download. Gelato Pro is rendering software developed (originally the Blue Moon Rendering Tools and Entropy software before a 2002 acquisition by NVIDIA) that allows for advanced acceleration on NVIDIA (specifically the Quadro series) GPUs. There has been a Gelato non-Pro edition of this software capable of rendering film-quality images that has been available for a free download, but now the professional edition is also free. The caveat for making this free, however, is that NVIDIA no longer plans to maintain the Gelato software. They have discontinued all work on the Gelato software and will be focusing their resources on mental ray software.
As a late Friday night release, NVIDIA's Aaron Plattner has announced xf86-video-nv 2.1.9. Back in March the xf86-video-nv 2.1.8 driver was released with initial support for the GeForce 9600GT and today's release improves the G80 support as well as fixing some startup bugs, sorting the supported devices table, adding an option to allow validation of dual-link DVI modes, and a few other minor changes. This announcement can be read on the xorg mailing list.
When OpenTheBlob.com started in late February, within one week of its launch it already had 5,000 signatures and days later it passed the 6,000 and 7,000 marks too. This letter was an open letter to NVIDIA looking for more information on their open-source strategy. While things have slowed down, this past week it crossed the point of having 9,000 signatures!
Last week NVIDIA had released the 173.08 display driver for Linux and Solaris operating systems. This driver update had introduced support for new GeForce and Quadro GPUs, experimental X Server 1.5 support, NVIDIA mobile improvements, and a few bug-fixes.
Back in January AMD had introduced the Radeon HD 3870 X2 and this morning NVIDIA has counter-attacked them by unleashing the NVIDIA GeForce 9800 GX2. The GeForce 9800 GX2 graphics card is made up of two GeForce G9x 65nm GPUs with each one being clocked at 600MHz and each having 128 stream processors and 512MB of 1GHz GDDR3 memory. These two GPUs are on the same PCB and connected via Scalable Link Interface. Early Windows benchmarks show the NVIDIA GeForce 9800 GX2 slaughtering the Radeon HD 3870 X2, but AMD's dual-GPU offering is priced much lower.
Back during CES 2008 we reported that NVIDIA may be plotting an open-source strategy (according to a NVIDIA partner) as the ATI/AMD camp has been on a spree with releasing NDA-free specifications to the OSS community and supporting two open source drivers (Radeon and RadeonHD). Since then, Intel has also joined this open bandwagon by releasing the G965/G35 programming documentation in full (2D, 3D, video, everything) even though they have already have a reliable open-source driver.
Following the open letter to NVIDIA at OpenTheBlob.com that takes aim at NVIDIA's lack of a reliable open-source driver, now out is a letter geared for NVIDIA's board partners (ASUS, Dell, BFG Tech, etc). This happens to be based off of a strategy I discussed before for frustrated ATI customers prior to the new driver code-base. If you're interested in taking a stand for an open NVIDIA, the page with links and a sample letter can be found here.
Late last month NVIDIA had unveiled the GeForce 9600GT graphics card as direct competition to the AMD Radeon HD 3850 / 3870. This new PCI Express graphics card still isn't supported by the latest official driver (169.12), but it's been reported to work (at least partially) when using the Tesla 171.05 driver. There is now, however, limited open-source support for this card.
Just four days ago the OpenTheBlob.com letter to NVIDIA requesting open-source support was published and it already has in excess of 5,000 signatures. Most of these signatures are also accompanied by comments, after hitting the front-pages of Digg and Reddit. Congratulations to the community with 5,000+ signatures in just four days, and hopefully the rumor pans out and NVIDIA will see the mutual benefit in an open-source strategy.
For those interested in seeing official open-source support from NVIDIA or open specifications, there is an open letter for open drivers to NVIDIA at OpenTheBlob.com. With both AMD and Intel now supporting open-source X.Org drivers and releasing specifications/documentation, the community is looking for the same (if not more) from NVIDIA. Keep in mind, last month we reported at Phoronix that NVIDIA may be developing an open-source strategy. This open letter to NVIDIA (and where you can leave your signature/comments) can be found here.
This morning NVIDIA has introduced their first GeForce 9 graphics card, with the introduction of the GeForce 9600GT. The NVIDIA GeForce 9600GT isn't the new high-end graphics card, but rather their next-generation mainstream graphics card that is designed to compete with AMD's recent Radeon HD 3850 and 3870. The NVIDIA 9600GT (G94) GPU is built on a 65nm process, 64 stream processors, 650MHz reference core clock, 1800MHz GDDR3 reference memory clock with a 256-bit interface, two dual-link DVI connectors, and is a PCI Express 2.0 part.
Earlier this month it was revealed that NVIDIA Corporation would be buying up AGEIA Technologies, which is the maker of the PhysX SDK and the PhysX PPU (Physics Processing Unit) hardware. That same day we had then asked the question whether is NVIDIA buying AGEIA good for Linux? (The responses.) AGEIA had produced a PhysX software SDK binary for Linux but have never released a Linux driver to enable the offloading of these physics calculations to their PPU hardware.
It was announced this afternoon in a laconic press release that NVIDIA will be acquiring AGEIA Technologies. AGEIA is the company behind the PhysX SDK and their Physics Processing Unit (PPU). NVIDIA's hopes for this acquisition is to offer GeForce graphics cards in the future that are packed with PhysX technology for in-game physics rendering and is a complement technology to NVIDIA's CUDA. CUDA is NVIDIA's Compute Unified Device Architecture for writing algorithmic code to be executed on the GPU with its massively parallel capabilities.
We're still working on finding out more details on NVIDIA's open-source strategy, but in minor open-source news, there are two updates now available for the xf86-video-nv (a.k.a. "nv") driver. The xf86-video-nv 2.1.7 update is for those using X.Org 7.2 or later, while xf86-video-nv 2.0.3 is for pre-X11R72. The nv 2.1.7 driver update adds support for the GeForce 8800GT and Quadro FX 3700, improved load detection, and a couple of fixes. The 2.0.3 release has a few more changes back-ported from the xf86-video-nv 2.1 series, including new GeForce 8 desktop/mobile product support. The release announcement made by Aaron Plattner can be read on the X.Org mailing list.
According to an AIB partner, NVIDIA is planning an open-source counterattack against ATI/AMD. Since this past September, AMD has been increasingly open-source friendly with their Novell partnership to deliver the RadeonHD driver and releasing open specifications. We have received information that NVIDIA is reportedly planning an increased open-source presence. Does this mean a cleaner "nv" driver? Open specifications? Jointly working with the Nouveau developers? Open sourcing part(s) of their blob? It's not known just yet.
For those impacted by the 100% fan speed bug present in the NVIDIA 169.07 Linux driver, there is now a community fix for this problem. NVClock 0.8 Beta 3 has been released, which (among other changes) addresses this driver bug. The major changes in NVClock 0.8 Beta 3, since it's Beta 2 release in July 2006 is GeForce 8 support, rewritten low-level GeForce 6/7 overclocking back-end, added BIOS PLL table parsing for the GeForce 6/7/8 generations, GeForce 7 AGP support, NV-CONTROL OpenGL settings, and GeForce 6 bug-fixes for pipeline modding and faking the Quadro. NVClock 0.8 Beta 3 can be downloaded from its project web-page.
Prior to NVIDIA porting CoolBits over to Linux back in 2005, the only way to overclock your NVIDIA graphics card was using NVClock. NVClock has been developed as a third-party open-source utility by Roderick Colenbrander and hosted at SourceForge and LinuxHardware.org. NVClock is accessible via the command-line as well as Qt and GTK interfaces. In addition to just overclocking the core and memory frequencies on NVIDIA graphics cards, NVClock also allows for some graphics cards to do pipeline soft-modding, enabling temperature sensors that have been disabled, OpenGL tweaks, and fan-speed adjustment. However, it looks like this project has faded away and that we may never see the final release of NVClock v0.8.
In addition to sharing that we are approaching a point in the Nouveau development where a stable 2D release with EXA and X-Video support is in sight, the Nouveau Companion 30 also mentioned that the NV50 work is "seriously understaffed." Fortunately though, today there were nine git commits for the xf86-video-nouveau driver that improve the state of open-source NV50 support. These commits include code cleanups for the NV50, a new wrapper, and a few renamed functions. You can checkout the latest Nouveau source-code from the git repository at FreeDesktop.org.
574 NVIDIA news articles published on Phoronix.