18 comments

[ 7.6 ms ] story [ 53.7 ms ] thread
If you have an AMD GPU the fix mentioned above is already released. Check AMD's driver page for the "15.11.1 Beta". But my card (an R9 290) was unaffected.
This is why I don't try out beta drivers (until some brave canaries have tried them first)!
(comment deleted)
but the crimson drivers (that broke fan speeds) were not beta
You have it backwards. The GPU fan bug is in the stable driver, and the fix is in the beta driver.

While a lot of teams are moving to a more frequent release schedule like Chrome, AMD for some reason has decided to go for the Apple model of one big release per year. The 2016 big stable release is the one with the horrible bug.

One big release a year? I have a hard time wrapping my head around why that would be a good idea.

Also, I'm not too impressed with the UI change. It looks like they sourced their graphic design from 2008.

I can't think of very many hardware vendors (especially in the enthusiast space) who seem to have done much UI design work past 2008 or so... AMD, Realtek, Gigabyte et al, even Intel's tweak tools are all awful. Gaudy, nonscalable, and bitmap-heavy is the name of the game.
One big feature release in addition to 6 WHQL driver updates in addition to random beta drops.
> AMD for some reason has decided to go for the Apple model of one big release per year

That's not really true.[1] They release "as needed", which varies in cadence. This is normal for a lot of software.

[1] http://www.anandtech.com/show/5880/amd-discontinues-monthly-...

That's outdated info. A newer article, also from AnandTech, nicely goes into detail. "in essence putting major feature updates on a yearly cadence" [1]

[1] http://www.anandtech.com/show/9811/amd-crimson-driver-overvi...

From that same link it's 6 WHQL drivers a year, at least that's the plan.

Major features on a yearly cadence is not really that big of a change. Major feature here would be things like Eyefinity, FreeSync, etc... This is very different from the monthly/bimonthly updates for performance improvements and bugfixes which is the real reason people update GPU drivers frequently.

AMD used to do monthly updates, but they were too buggy so they moved to updating less often so they could spend more time testing, but the drivers are still buggy. Rapid releases can help but you can't really solve testing problems with release schedules.
(comment deleted)
How could this have slipped through testing? Did their new version result in substantial enough code changes that the tests needed to be reimplemented but weren't? Of course, testing that kind of software/hardware interaction seems difficult.
Probably not enough test cases, from what I read this seemed to effect non-reference cards, overclocked from the OEM.
This is why you have hardware failsafes, people.

The driver software shouldn't get a vote about fan speed if the temperature is at failure-level extremes.