Linux 3.12 Brings Big AMD Radeon Improvements

[Alejandro Nova]

Please, include E-350 into the testing mix.

I've seen many testimonials of AMD Radeon 4850 owners who were unhappy with Linux 3.11. Couldn't this be simply a bug fix for Radeon 4850?

[bridgman]

See Michael's post #37...

I've been able to confirm this so far on a HD 6000 and HD 4000 different GPUs and a completely different PC.

[agd5f]

The only thing I can think of that may have improved things is changing the default gart size on r7xx+ asics from 512M to 1024M. If anyone want to test different gart sizes, you can change the gart size with the gartsize radeon kernel module parameter. E.g., add radeon.gartsize=1024 to the kernel command line in grub. Default is 512M for r1xx-r6xx and 1024M for r7xx+.

[mcgreg]

my experience

AMD 6850 / Nexuiz

Kernel 3.11 old git ( HEAD is now at f7217b9 r600g: enable SB backend by default )

NORMAL: 98,5 fps
ULTRA: 56,8fps


Kernel 3.12 / git commit 453ea2d309c0314bea8a209f536e2b3c2c4d92c6 / Date: Sun Oct 13 19:53:54 2013 +0400

NORMAL: 206 fps
ULTRA: 108 fps

Pretty nice speed up for me.

[Hamish Wilson]

Something strange and potentially wonderful is afoot, methinks.

[AJSB]

A10-6800 4,1 GHz - Gigabyte AMD Radeon HD 6870 1024MB (915/1050MHz)

Kernel: 3.12.0-rc4:

Xonotic 0.7:
    pts/xonotic-1.3.1 [Resolution: 1920 x 1080 - Effects Quality: Ultra]
    Test 1 of 1
    Estimated Trial Run Count:    3
    Estimated Time To Completion: 11 Minutes
        Started Run 1 @ 21:07:15
        Started Run 2 @ 21:08:54
        Started Run 3 @ 21:10:34  [Std. Dev: 1.61%]

    Test Results:
        119.2862578
        116.5561488
        115.6666134

    Average: 117.17 Frames Per Second
    Minimum: 56
    Maximum: 264

Kernel: 3.11.4:

Xonotic 0.7:
    pts/xonotic-1.3.1 [Resolution: 1920 x 1080 - Effects Quality: Ultra]
    Test 1 of 1
    Estimated Trial Run Count:    3
    Estimated Time To Completion: 11 Minutes
        Started Run 1 @ 21:15:28
        Started Run 2 @ 21:17:04
        Started Run 3 @ 21:18:39  [Std. Dev: 0.11%]

    Test Results:
        122.3697861
        122.3688075
        122.1429572

    Average: 122.29 Frames Per Second
    Minimum: 59
    Maximum: 264

Thanks very much for the info !
(I also would love to see some results in a E-350 and a A8-3870K...anyone ? I wanted this to have an idea what could be the theoretical performance of underclocked and undervolted A6-5400K and a A6-5200 (Kabini) )

Edited...

Wait, your A10-6800K results are with a dGPU, what we need is pure iGPU results ...

[borsook]

Lovely. But even if performance wise the open driver would be the same as Catalyst I would still use the latter for two reasons:

1. Power saving
2. Configuration capabilities.

While there is some work being done on 1 there is absolutely none on 2...

[pingufunkybeat]

Lovely. But even if performance wise the open driver would be the same as Catalyst I would still use the latter for two reasons:

1. Power saving
2. Configuration capabilities.

While there is some work being done on 1 there is absolutely none on 2...

Open drivers should have power saving comparable to Catalyst now.

[przemoli]

Lovely. But even if performance wise the open driver would be the same as Catalyst I would still use the latter for two reasons:

1. Power saving
2. Configuration capabilities.

While there is some work being done on 1 there is absolutely none on 2...

What config cap You need?
(I guess that its not multimonitor setup nor power saving options as that is either via gui, or via etc files)

[borsook]

What config cap You need?
(I guess that its not multimonitor setup nor power saving options as that is either via gui, or via etc files)

Antialiansing settings, texture filtering settings, vsync etc etc. Plus using open drivers still is bothersome due to the situation with S3TC...