AMD 760G (Radeon 3000) May Be Worrisome On Linux
For those looking to utilize the Radeon HD 3000 integrated graphics found with the AMD 760G motherboard chipset, the support may be less than stellar when using the modern open-source Radeon Linux graphics driver.
In needing an extra motherboard for use with an AMD FX-8150 "Bulldozer" processor to use as an extra test system for a new research project, I ended up ordering the MSI 760GM-P34 (FX). The 760GM-P34 is centered around an AMD 760G chipset with the SB710 Southbridge. The 760G+SB710 is by no means new but the motherboard is cost effective at around $50 USD and gets the job done for what I need the system to do in conjunction with the Eight-Core Bulldozer CPU. The AMD 760G chipset features integrated Radeon 3000 graphics.
Being based on an older AMD chipset, the MSI 760GM-P34 had no problems when installing Ubuntu 13.04 with the Linux 3.8 kernel. Everything was working fine in conjunction with the AMD FX-8150 Bulldozer, 4GB of Corsair DDR3-1600MHz memory, and Western Digital HDD. The latest BIOS was also in use for the MSI micro-ATX motherboard.
The problems with this AMD 760G motherboard came when curiosity got the best of me to run some tests of the modern Radeon Gallium3D driver (Linux 3.8 + Mesa Git) on the hardware.
When running any of the automated OpenGL tests on Linux with this hardware configuration via the Phoronix Test Suite, Radeon DRM problems were quick to occur.
The most notable issue were the "*ERROR* Failed to schedule IB" / "couldn't schedule ib" kernel messages. Ultimately, the system would end up rebooting after a while with any of the OpenGL tests. These IB error messages for the Radeon DRM aren't too uncommon, but not recently. I've hit the problem before and there's been many reports of the error previously in numerous bug reports. Seeing the problem emerge with the Linux 3.8 kernel and a mature AMD 760G chipset with Radeon 3000 graphics is a bit of a surprise. It isn't a terrible surprise though since generally I've found the Radeon HD 3000 graphics card support on the open-source Linux graphics driver to be the worst. While I have several HD 3000 series graphics cards, in the common multi-way GPU comparisons on Phoronix, these cards are generally left out since they usually end up tripping and failing on the open-source stack in one way or another.
After some tweaks and trial and error, the problems remained. The only test results from tinkering with the problematic configuration are the 1302257-FO-AMD760G7837 result file.
The performance on the open-source Radeon Gallium3D Linux graphics driver for the Radeon 3000 integrated graphics though is rather low and not really worth it unless you just care about running a compositing window manager. Unfortunately the mainline AMD Catalyst graphics driver no longer carries support for the pre-HD5000 series hardware, so the open-source driver is the only real option with modern Linux distributions.
In needing an extra motherboard for use with an AMD FX-8150 "Bulldozer" processor to use as an extra test system for a new research project, I ended up ordering the MSI 760GM-P34 (FX). The 760GM-P34 is centered around an AMD 760G chipset with the SB710 Southbridge. The 760G+SB710 is by no means new but the motherboard is cost effective at around $50 USD and gets the job done for what I need the system to do in conjunction with the Eight-Core Bulldozer CPU. The AMD 760G chipset features integrated Radeon 3000 graphics.
Being based on an older AMD chipset, the MSI 760GM-P34 had no problems when installing Ubuntu 13.04 with the Linux 3.8 kernel. Everything was working fine in conjunction with the AMD FX-8150 Bulldozer, 4GB of Corsair DDR3-1600MHz memory, and Western Digital HDD. The latest BIOS was also in use for the MSI micro-ATX motherboard.
The problems with this AMD 760G motherboard came when curiosity got the best of me to run some tests of the modern Radeon Gallium3D driver (Linux 3.8 + Mesa Git) on the hardware.
When running any of the automated OpenGL tests on Linux with this hardware configuration via the Phoronix Test Suite, Radeon DRM problems were quick to occur.
The most notable issue were the "*ERROR* Failed to schedule IB" / "couldn't schedule ib" kernel messages. Ultimately, the system would end up rebooting after a while with any of the OpenGL tests. These IB error messages for the Radeon DRM aren't too uncommon, but not recently. I've hit the problem before and there's been many reports of the error previously in numerous bug reports. Seeing the problem emerge with the Linux 3.8 kernel and a mature AMD 760G chipset with Radeon 3000 graphics is a bit of a surprise. It isn't a terrible surprise though since generally I've found the Radeon HD 3000 graphics card support on the open-source Linux graphics driver to be the worst. While I have several HD 3000 series graphics cards, in the common multi-way GPU comparisons on Phoronix, these cards are generally left out since they usually end up tripping and failing on the open-source stack in one way or another.
After some tweaks and trial and error, the problems remained. The only test results from tinkering with the problematic configuration are the 1302257-FO-AMD760G7837 result file.
The performance on the open-source Radeon Gallium3D Linux graphics driver for the Radeon 3000 integrated graphics though is rather low and not really worth it unless you just care about running a compositing window manager. Unfortunately the mainline AMD Catalyst graphics driver no longer carries support for the pre-HD5000 series hardware, so the open-source driver is the only real option with modern Linux distributions.
17 Comments