AMD: It's time to open up the GPU [LWN.net]

AMD: It's time to open up the GPU

Posted Jan 27, 2016 14:14 UTC (Wed) by flussence (guest, #85566)
In reply to: AMD: It's time to open up the GPU by tshow
Parent article: AMD: It's time to open up the GPU

The choice of "open" as opposed to "libre" seems deliberate, unfortunately.

AMD remains the only one of the big three GPU vendors whose entire hardware range is completely non-functional on a deblobbed kernel.

to post comments

AMD: It's time to open up the GPU

Posted Jan 27, 2016 14:46 UTC (Wed) by brunowolff (guest, #71160) [Link] (2 responses)

Recent Intel GPUs require blobs. For older nVidia GPUs someone has reverse engineered enough to replace the blobs, but for new GPUs the blobs need to be signed by nVidia. I don't think AMD is worse than the others here.

AMD: It's time to open up the GPU

Posted Jan 27, 2016 15:29 UTC (Wed) by flussence (guest, #85566) [Link] (1 responses)

You seem to be violently agreeing with what I said; that's why I specified entire hardware range.

(Somewhat depressingly, AMD's penchant for binary blobs extends *beyond* the scope of their own hardware — I bought an AGP Radeon a dozen years ago which mysteriously sprouted a firmware dependency only after ATI ceased to be...)

AMD: It's time to open up the GPU

Posted Feb 12, 2016 1:05 UTC (Fri) by bridgman (guest, #50408) [Link]

>>(Somewhat depressingly, AMD's penchant for binary blobs extends *beyond* the scope of their own hardware — I bought an AGP Radeon a dozen years ago which mysteriously sprouted a firmware dependency only after ATI ceased to be...)

IIRC the early drivers managed the acceleration hardware by writing directly to registers then waiting for completion, but later versions used PM4 packets (processed by the microcoded CP block) which allowed the driver to run ahead of the hardware for higher performance.