Following the RandR 1.2 integration earlier this week, the RadeonHD v0.0.4 driver has been released. In addition to the RandR 1.2 support from master, RadeonHD v0.0.4 also features LVDS support on RS690 GPUs and later. What this means is that this open-source driver should now properly work with notebooks that use the M72, M74, M76, or RS690 ASICs. This release also includes a large number of fixes and simplification of the connector naming scheme. The RadeonHD v0.0.4 announcement can be found on the mailing list and the code can be checked out (as always) from FreeDesktop.org git.
The initial-randr branch for x86-video-radeonhd has been merged into master. What this means is that you no longer need to checkout this separate branch but that all of the RandR 1.2 support is now in the mainline code for this open-source driver. The RadeonHD driver provides full RandR 1.2 support and even some initial work for RandR 1.3. The OpenSuSE mailing list announcement goes into detail about this work and what to do if you run into any RandR problems. A new RadeonHD driver release may be out as soon as early next week.
Since the AtomBIOS branch appeared for the xf86-video-ati driver and the R500 support for DRM (the announcement), we've been trying out this open-source software on a variety of different ATI Radeon graphics cards. As we reported a few hours ago, the xf86-video-ati driver is working with the high-end HD 2900XT 512MB graphics card. We have now tried it out on multiple Radeon X1800 series graphics cards and it was running great there too. Using XAA acceleration was very smooth though with EXA we have ran into a few issues with the DRM R500 branch causing slowdowns. When trying out an ATI Radeon X800XL with the AtomBIOS branch (since the AtomBIOS support goes back to the R400 series) we have ran into TMDS issues. The exploration will continue and we'll report back soon with more findings and possible benchmarks.
While all of the open-source rage for the ATI R500 and R600 series has been about the official RadeonHD driver, it seems that the Radeon (as in xf86-video-ati) driver will support these new ATI GPU families as well. David Airlie has established an atombios-support branch in xf86-video-ati and a r500-support branch in DRM. David Airlie and Alex Deucher have been working on introducing AtomBIOS support to the Radeon driver for the R400/500/600 series and Radeon acceleration support for the R500 (Radeon X1000) family. The Radeon driver now supports XAA and EXA with the R500 graphics cards. 3D is not supported, however, it is being worked on. This AtomBIOS branch does support RandR 1.2 as well as dual-link TMDS outputs, and external TMDS chips on the R400 GPUs. David had mentioned that open-source 3D support on the R500 series through the Radeon driver should be possible in the near future and that Compiz support should come sooner rather than later. Read more on David Airlie's blog.
A Wiki page over at X.Org has been established for the RadeonHD driver. This Wiki page describes the driver along with all supported GPUs, links to the mailing list, git repository, and IRC channel, the general steps to install the xf86-video-radeonhd driver, and common problems during the driver build process. Check out the RadeonHD Wiki.
This morning AMD has introduced their new cutting-edge PC platform that's codenamed "Spider". The Spider platform consists of AMD's Phenom quad-core processors, ATI Radeon HD 3800 graphics, and the AMD 7-series chipsets. The AMD Phenom processors are quad-core and based upon Direct Connection Architecture with an embedded memory controller with DDR2-1066 support and shared L3 cache. More information on the new AMD Phenom processors and chipsets can be found in today's press release. In the near future we hope to share with you how well the AMD Spider platform performs under Linux and whether there's any compatibility troubles with their new AMD 790FX Chipset.
Alex Deucher, a free software developer who lately is responsible for much of the work on the xf86-video-ati driver, has today announced that he has joined AMD. Beginning next month, Alex Deucher will be an AMD employee and will be working on the AMD's open-source initiatives. He will be working on the open-source ATI drivers and working with the open-source community at large. Alex had announced this on his personal blog. Last week at Phoronix we reported that AMD was hiring personnel for their open-source work.
One of the questions that has come up since we reported that AMD is preparing for another GPU documentation release and that R100/200 specifications will be made available, is whether internal information on the All-In-Wonder graphics cards will be published.
On Friday we talked about ATI preparing to release more GPU documentation to the public and without any Non-Disclosure Agreement, and today we have a few new details regarding the specifications for earlier Radeon GPUs. AMD's John Bridgman has posted in this Phoronix Forums thread that he is trying to re-release the GPU specifications for the R100~200 generation of graphics processors. The to be released R300/400 documentation will also assist in further enhancing the Radeon R100/200 driver. John had also mentioned that they will probably be able to assist open-source developers in any questions that they have for even older ATI GPUs as well (ATI Rage era).
Yesterday we reported on the brand new Radeon HD 3850 / 3870 graphics cards and their status under Linux. Later then we reported in AMD Preparing For Another GPU Documentation Release that these new RV670 graphics cards should "just work" with the RadeonHD driver thanks to the use of the AtomBIOS. This morning, Novell's Luc Verhaegen has confirmed in a thread in the Phoronix Forums that these new GPUs should already be supported by xf86-video-radeonhd. The RadeonHD developers haven't actually seen either the HD 3850 or HD 3870 yet, but it's expected that by just adding the correct PCI IDs that these new graphics cards will work with this open-source graphics driver.
This morning ATI/AMD launched the ATI Radeon HD 3800 series graphics cards. The ATI Radeon HD 3800 GPUs contain 320 stream processors, PCI Express 2.0 compliance, CrossFire X technology, and an assortment of other Radeon technologies. The GPU die is also built on a 55nm process. The two cards in this series currently available, the HD 3850 and HD 3870, are mid-range graphics cards priced at $179 and $219 USD, respectively. For Microsoft Windows users, these cards are also DirectX 10.1 compliant. The advantages that these ATI Radeon HD 3800 series cards have over the Radeon HD 2900XT 512MB is PowerPlay support and a Unified Video Decoder (UVD) for high-definition content.
While only a minor change to add the ATI Radeon X1250 support, this GPU is now supported by the open-source RadeonHD driver. There's been inquiries in the forums and elsewhere about the X1250. If you happen to be an ATI X1250 owner, be sure to check out the latest RadeonHD source-code and report your results to the RadeonHD mailing list.
Yesterday we received word from Novell's Matthias Hopf that the 2D portion of the RadeonHD driver will not be stable by the end of the year. After publishing that news article, we heard some new details from Luc Verhaegen, who is also part of Novell's X driver development team.
The key developers at Novell that are behind the open-source RadeonHD driver for the ATI R500/600 graphics processors have been doing a great job (as you can find out by reading a number of our RadeonHD articles). Previously we reported that the Novell developers had hoped to complete the 2D RadeonHD work by the end of the year; however, that will no longer be the case. Novell's Matthias Hopf, one of the RadeonHD lead developers, had commented: "Maybe, MAYBE, some initial stuff (ScreenToScreenCopy or so) will be done this year, but I'd rather doubt it." He had also stated, "The people at AMD have also worked a lot lately." Beyond that, no other details were shed on a revised road-map or when AMD may release their next set of GPU specifications. Wouldn't it be a wonderful present to the free software community if AMD did their next documentation dump in time for Christmas?
Last month Alex Deucher pushed out the xf86-video-ati 6.7.195 driver, which is another release candidate before the 6.8 release for this open-source ATI graphics driver. The ATI 6.7.196 release includes a few Radeon fixes, specifically for cursor-related hangs, LVDS fixes, support for new DDC type, and BIOS/dock/LVDS hardware state fixes. The xf86-video-ati 6.7.196 driver release announcement can be read on the X.Org mailing list. As was mentioned during the last RC release, the 6.8.0 driver will be released once all of the sub-drivers are ported to using PCI rework.
Previously we reported on the Revenge utility being developed by Oliver McFadden for reverse-engineering the ATI fglrx binary display driver. Today marks the first release of the Revenge utility as it has reached version 1.0. This isn't a bug free release but Oliver hopes it will ship with some distributions to help in the reverse-engineering process and turn more Linux users to free software drivers. As you may recall from an earlier news post, Oliver McFadden is also working on developing an open-source video BIOS for at least one ATI graphics card.
It's over 24 hours now since the fglrx 8.42.3 download link first surfaced and we shared the great news with you that 8.42 delivers on AIGLX and other improvements. However, the AMD website still hasn't been updated for this new Linux driver. Though the AMD Proprietary Linux Release Notes for 8.42.3 is now accessible.
Yesterday there was a huge PLL fix for the Mobility R500 and R600 series for the RadeonHD driver along with some other changes and today is another batch of commits. Today's batch of git commits by the Novell developers include adding size checks to the AtomBIOS connector table parser, implementing PCI BIOS access, removing the DDC test, README file update, segmentation fault fix, and enabling LVTMA for panel on the integrated AMD RS690 chipset. As always, check-out xf86-video-radeonhd from FreeDesktop.org for the latest driver.
The ATI/AMD fglrx driver for x86 and x86_64 Linux is now available! This supports AIGLX and everything else we at Phoronix have been talking about. The 8.42.3 download link is here. With that said, be sure to read our just-published AMD 8.42 Display Driver Review. Please Digg to share.
Oliver McFadden, the developer behind the Revenge reverse-engineering utility for ATI Radeon GPUs, is hoping to create a free software (GPL licensed) Video BIOS for at least one ATI Radeon graphics card. This is certainly a much larger project than just reverse engineering a driver and is much more risky, but at the same time is very interesting and holds merit. Mark Shuttleworth would also like to see everything down to the device's firmware being free software. The steps Oliver is taking at this point is to examine the AtomBIOS parser, which is open-source as part of the RadeonHD driver (a partial explanation of AtomBIOS). If this open-source BIOS for ATI Radeon GPUs manages to take shape, we'll be sure to cover it here at Phoronix. More information is available on Oliver's blog.
Luc Verhaegen of Novell has pushed additional PLL fixes into the xf86-video-radeonhd git tree. These PLL fixes should address all outstanding issues involving the setup and generation code, including the monitor not syncing and horizontal steadiness. Specifically, Mobility Radeon X1000 "R500" and all Radeon HD 2000 "R600" owners are encouraged to try the latest code and reporting the results to the RadeonHD mailing list.
Last week the RadeonHD development was rather slow with only 30 commits to their xf86-video-radeonhd git tree all week. However, the Novell developers have been making some good progress this week. Earlier this week there were a few fixes to the RadeonHD driver and AtomBIOS. Today so far there have been fifteen commits. These changes represent a number of fixes, adding additional queries, semantic changes, and one of the most notable commits is adding support for libpciaccess. The open-source R500 Avivo driver has supported libpciaccess for a while, but now it's finally coming to the RadeonHD driver. PCI rework/libpciaccess is the new generic PCI access library and is meant to be platform independent. As libpciaccess is relatively new, the RadeonHD driver hooks into libpciaccess but at the same time maintains backwards compatibility with older versions of X.Org. If XSERVER_LIBPCIACCESS is defined, the RadeonHD driver will take advance of this PCI library. The driver version has also been bumped to v0.0.2.
This week has seen some more improvements to the RadeonHD driver. The week started out with the RadeonHD driver better implementing the AtomBIOS. AtomBIOS is now used for obtaining clock limit data in order to calculate pixel clocks and LVDS parameters for panels. A few memory leak fixes were also fixed. This past week has seen about 30 commits to the RadeonHD driver. This number is in starch contrast to previous weeks where we seen as many as 90 commits. In fact, there hasn't been any commits to xf86-video-radeonhd in three days. The Novell developers though have remained active on the RadeonHD mailing list. Perhaps they are just preparing for the next array of changes? Trying out the Radeon HD 2900XT 512MB graphics card with the latest RadeonHD git on October 13, everything was still running fine.
While most of the open-source ATI/AMD excitement these days is all about the RadeonHD driver, Alex Deucher is working on a few interesting changes for the Radeon (R200/300/400 hardware) driver. The latest Radeon driver available through git now attempts to handle the status of laptop lids to prevent a BIOS problem while the driver is loaded. However, this will only work right now if you boot with the lid closed. Alex has also merged the r128 driver into the Radeon driver, with mode-setting and MMIO acceleration now functioning. DRI and X-Video should also be working for r128 hardware using the Radeon driver in the near future. Some of the benefits of this merger include RandR 1.2 support for the r128, TV-Out support, better BIOS table support, and one less driver to PCI rework. More on the Radeon driver work can be found in Alex Deucher's new blog and all of the recent changes can be found in the FreeDesktop.org git for the xf86-video-ati driver.
Committed to the xf86-video-radeonhd driver this morning has been another series of fixes and other improvements for this open-source AMD R500/600 driver. Some memory leaks in the driver have been fixed, ConnectorTable is now known as ConnectorInfo, the AtomBIOS test code was removed now that the AtomBIOS parser is working, compiler complaint fixes, and other changes are now in the git tree. If you're running the RadeonHD driver now, be sure to update and if you're not but are running R500/600 hardware, be sure to check out the git and let the RadeonHD developers know how it's working for you.
While AtomBIOS, AMD's library for providing effective firmware processes from the BIOS (that we originally detailed in this article), has been present in the RadeonHD driver's tree for a few weeks, it's now starting to be used more. Changes to the open-source R500/600 driver this morning now use AtomBIOS for obtaining clock limit data for calculating pixel clocks and LVDS parameters for panels. The algorithm for searching for valid PLL dividers has also been changed. Though in the mailing list release announcement, the Novell developers are still have a lot of work left with solving all of the clock programming problems. The second change with LVDS panel information from the AtomBIOS will hopefully make the need for the device tables unnecessary.
Alex Deucher has pushed out another release candidate for the xf86-video-ati driver designed for the pre-R500 ATI Radeon GPUs (not to be confused with RadeonHD). In total there are 29 changes for the xf86-video-ati 6.7.195 release candidate -- 19 of them are targeted for the Radeon driver, 9 for the Mach64, and one being for the rc410. The most notable changes for this release include fixing the support for the single CRTC graphics cards, LVDS/RMX is finally sorted out, and pci-rework for ati_msic/radeon. The PowerPC build has also been fixed in this X.Org ATI driver release. Alex had mentioned in the 6.7.195 release announcement that the xf86-video-ati 6.8.0 driver will be released as soon as all of the sub drivers are ported to using pci-rework.
If you're a developer or an experimental user running the RadeonHD Linux driver and have been plagued by cursor problems, check out the latest xf86-video-radeonhd git code. Novell's Matthias Hopf has committed several changes to this open-source AMD R500/600 driver today that pertain to cursor-related fixes and improvements.
While the Novell developers are hoping to have stable 2D working with the RadeonHD driver by the end of the year, much more work will be needed once AMD delivers the rest of the specifications they have promised. Brought up on the RadeonHD mailing list this morning was whether it would be worth it if a user bought an R600 now or not. AMD's next-generation graphics processors will be out before the R500/600 series has an open-source working 3D driver, so is it a better bet waiting for the R700 series (Radeon HD 3000, perhaps)? Granted, the fglrx driver now has R600 support. The R700 is based upon the R600 (Radeon HD 2000) architecture, so it hopefully won't be too much work to implement its functionality -- especially if AMD delivers the R700 specifications to the community in a timely fashion. The ATI R700 is expected in the first half of 2008. While it's been a while waiting for open-source R500 and R600 support, we finally have basic 2D support and 3D should be expected for next year, the waiting for the R700 open-source support will hopefully be dramatically shorter. Once everything arrives, it should be benchmarking heaven here at Phoronix :).
With the availability of the RadeonHD driver, Jerome Glisse (the xf86-video-avivo maintainer) has declared the X.Org Avivo driver dead, but that's not preventing some postmortem changes. There hasn't been any real activity with this open-source R500 driver in about two to three weeks when AMD's open specifications had helped the developers. Earlier this week, however, there was a fix in the Avivo CRTC mode-setting and today support for the Radeon HD 2300 was added. The Avivo driver doesn't support the R600 series, but the Radeon HD 2300 is actually based upon the R500 architecture. The Radeon HD 2300PRO simply worked by adding its PCI ID to the driver. Both of these commits are rather trivial and doesn't ignite any hopes that the Avivo driver may make a comeback (especially with the rapid progress of the RadeonHD driver).
1111 AMD news articles published on Phoronix.