RadeonHD Weekly Progress Report

As there's been a RadeonHD article at least once a week (from the Ubuntu installation to the Conntest utility) on Phoronix since the RadeonHD driver was introduced, it's almost becoming a weekly progress report here for this official AMD open-source R500 and R600 driver.

Last week when talking about the RadeonHD driver there were about 90 commits to its xf86-video-radeonhd git tree the entire week and just today from the Novell developers we have seen 26 commits in the past day. As opposed to commits just adding new ATI R500 and R600 product IDs, most of the committed work today involves fixing bugs that pertain to the identification of bogus table headers, error messages, mismatches, crashes, and other fixes. Aside from the fixes, there was some additional work done with the AtomBIOS (specifically the connector device tags) there was much work done with adding a framework for the connector table parser along with more consistent debug message printing, and object header parsing.

Aside from today's work, there was a fix earlier in the week so the AtomBIOS would build on OpenBSD, adding new product IDs, and fixing other broken areas. With the latest git code this afternoon we had tried it on an ATI Radeon HD 2600PRO 256MB graphics card, but it presented a gloomy situation. We hadn't run into identification problems or the common TMDS connector issues resulting in X to fail, but no display was being detected by the LCD monitor.

Within 24 hours of the RadeonHD driver being introduced, we were able to successfully get the open-source driver working on the Radeon HD 2900XT 512MB, which was certainly a beautiful sight to see. Though over the course of the driver's lifespan thus far we've run into a few quirks along the way with the Radeon HD 2900XT and in a few situations led to the driver not working correctly. With the latest xf86-video-radeonhd driver code, the HD 2900XT is (again) working great. The open-source HD 2900XT had no problems running on Ubuntu 7.10 Gutsy Gibbon Beta. Right now though the resolution was maxed at 800 x 600 as opposed to the monitor's native 1680 x 1050 resolution.

The Radeon HD 2400PRO had also worked like a charm, albeit at 800 x 600 instead of 1280 x 1024 in this case.

As we shared in the last RadeonHD article, at any time there are at least two full-time Novell developers working on this driver and it is certainly showing in a good way. To the public, this open-source AMD driver isn't even a month old but it's accelerating at a phenomenal rate. This progress is in fact quicker than we expected even after we learned internally that Novell would be doing much of the development work with the open NDA-free specifications. As we shared in that same article, by the end of the year these developers hope to complete the 2D work and then move onto the 3D work. We will be back with more RadeonHD driver coverage as the developments continue.