AMD Is Disabling DPM Support For RV770 GPUs

[cutterjohn]

har har har ATI(AMD) is teh shit with GPUz! NOT!

[I'm going to go offtopic kinda here, since I've had nothing but shitzor experience w/ATI(AMD) software EVAR]

Hell, I have a notebook with their mobile version of the 4850, the 4850 Radeon Mobility HD which was the same as their desktop excepting slightly lower core/memory clocks.

I suffered through YEARS of UTTER CRAP driver support, theoretical bugfixes(lulz), while dealing with copious new bugs before being dumped out of their mainline quite quickly(i.e. they had no f'ing idea wtf they were doing).

So, I'd say that they still have no f'ing cluez, their sw boyz are just godawful, as I learnt last night attempting to up a Winders, mind you, based system w/get this an FX-9590 + R9 280X to their "beta"(alpha is a better description as their "WHQL" is beta at best) 14.3 drivers(mind you after using a selection in succession of builtin old driver uninstall, then their own driver removal utility) only to get well display driver installs(thank all the f'ing gods) then either it craps out immediately(rinse and repeat with their own driver removal tools), or it get's to install crapalyst control center then craps out(repeat and rinse), or it installs crapti experience(or WTF they call their halfassed version of geforce experience) and then craps outs with a lengthy error log w/o installing crapalyst control center. (I gave up at this point and am storing the box until I have time to f'around w/it using guru3d's ddu at least the display driver installs every time, but I'z gotz no controlz overz itz, so it just runs at defaults... mind you this is under their supposedly better supported winderz... linux was just a cluster fuck of even worse than dealing with nvidia optimus proportions and I'll leave it at that...)

Christ! I'd swear that ATI(oops sorry AMD) has a million monkies ATTEMPTING to write drivers for their lackluster, spaceheating GPUs and have, well hell, since I can recall since I've NEVER had a "good" experience with their drivers.

Meanwhile, beta nvidia drivers only throw errors on RARE occasions and MOST of those are spurious, and I'm talking ATI/nVidia experience dating back to c. 2000, or Rage 128 Pro v. GF2(yeah linux support for both sucked at the time, but by the gods ati drivers were still crap back then and they STILL HAVE NOT LEARNT THE ERROR OF THEIR CRAPPY WAYZ! IM SURPRISED THAT PUNTERS STILL BUY THEIR CRAP AS let's face it their new GCN ain't even up to kepler level unless their crap drivers are dragging it down so much... sorry to say... (the 4850 looked good and mostly kinda worked, but borked out from drivers WAY too often)).

hmmm... mebbe delete me but I just had to vent about ATI/AMD, their crap recent hw, and continously crap sw, and am sick and f'ing tired of hearing about wah OSS, which would be nice if it actually worked any bettar than the official sw, but nope, just crap(crapalyst) or crappier(OSS radeonsi or whatevar teh f' it's called, just a PR campaign by ATI/AMD, and a phailwhale one in my book since it's pretty well halfassed at best!)

...so, I'll be sauntering back to my nice perfectly functional nv blobs, unapologetically, after all they just WORK. (and yes I'm using their "betas" on some of my systems both windows and linux, only problem is on notebooks optimus is poorly supported from what I can see of freetardedness, but it works a treat under winderz...)

[zanny]

[I'm going to go offtopic kinda here, since I've had nothing but shitzor experience w/ATI(AMD) software EVAR]

...

...so, I'll be sauntering back to my nice perfectly functional nv blobs, unapologetically, after all they just WORK. (and yes I'm using their "betas" on some of my systems both windows and linux, only problem is on notebooks optimus is poorly supported from what I can see of freetardedness, but it works a treat under winderz...)

This trolling is fantastic. Keep fighting the good fight John.

Just to make sure everyone is aware, AMD employs 6 developers on Mesa. They employ at least a hundred on Catalyst. Those 6 are absolutely fantastic, so criticize the company for not putting more hands on Mesa (which, really, they should be doing for their long term viability as a company - moving into the ARM / mobile space with tightly integrated FOSS drivers auditable by enterprise customers on next generation Linux based convergence devices (Ubuntu Phone, Plasma Active, etc) will be huge, I think).

[agd5f]

For the majority of users, dpm on rv770 is stable, that's why we enabled it by default originally. Unfortunately, you don't get quite the same testing coverage until you enable something by default. Once you do, all the corner cases suddenly appear.

[smitty3268]

I think he is referring to HyperZ.

I personally have had no problems with DPM myself, but I am using an RV730.

Maybe. It's just that he specifically mentioned older cards, and hyperz is disabled for everything.

[GreatEmerald]

RV770 includes RadeonHD 4890 cards too as far as I know.

Really, it's disabled for the HD 4890 as well? Argh. I'm using DPM on it and it seems to be working fine (although still runs a bit hotter than I'd like, but then it's a beast to begin with). Well, at least it's using a pre-3.13 kernel at the moment, so I'll just have to not change the GRUB settings.

[dfx.]

damn it

Odd. As far as I know, only Radeon HD 6800s were affected by DPM on kernel 3.13, as I had reported it several RCs before it went stable and it was finally disabled by default in 3.14. My Radeon HD 4870 machine runs perfectly fine with kernel 3.13 and DPM enabled.

"finally disabled by default", seriously ? you saying it as if it's a good thing.
my 6870 is a goddamn beast when on full-speed, noisy and hot. which is completely unpractical default state on Linux, since there are almost no usage for its power. without DPM there is no reason to even use open drivers other than pure testing. hard-setting it to run on minimal settings might help, but you may as well just switch to shitty embedded GPU. 4730 (RV770CE) is also not well on constant full-speed (~70℃ doing nothing, just spinning fan and heating the chip).

this approach may work fine for shit-tier cards, but is really bad for something more serious. especially if you account for delays in updating kernel and graphic stack in distributions: it may be from half to a year between major updates, unless you're using third-party repos ! i guess it's for clueless people, so their distribution installer at least wouldn't hang and crap on screen with default settings during fresh installation.

[moilami]

Get Intel OS graphics.

[zanny]

Get Intel OS graphics.

On FOSS drivers, my 7870 benchmarks as almost 15x faster than my integrated 4600. AMD is the only choice for high end free driver graphics cards.

It sucks, though. We have a chicken and egg problem, and I don't think anyone is really to blame. Honestly, it may just be in AMD's favor they have salaried Mesa developers at all, because money talks, and there is this pervasive overriding ideology that AMD stuff is bad on Linux so don't use it. Which means nobody buys their stuff or uses the free drivers and pressures AMD to improve them, so they don't hire more developers. So the driver improves slowly, and everyone keeps saying it sucks, and when things like this happen everyone just hates on AMD, and the casual viewer who is looking for a new GPU sees this and just buys Nvidia stuff.

But Nvidia doesn't even try to be open. They are probably the most antithetical company to FOSS out there, even beyond Microsoft at this point, who at least open sourced a few trite things over the years. And the Linux ecosystem is running headlong into their embrace, surrendering their freedom like nobodies business for convenience, while AMD, who obviously don't have enough Linux foss developers, at least have some.

It is a vicious cycle, and I know that in buying AMD stuff I'm getting a less reliable experience (at best), but in the end, they are doing the (mostly) right thing, and I just have to incentivize that. And it might just be forward thinking, because with AMD stuff I can reliably expect long term improvement and the eventual eclipsing of the proprietary drivers as they stay stagnant in their closed isolated bubbles and Mesa can integrate with all the best and newest features in the ecosystem. It just might take a long time to get there.

[whitecat]

@agd5f
RV770 are now quite old but in fact still quite widespread I guess (that was a popular card) and moreover able to run most of the native "Steam" Linux games.
I hope you will not give up with these trickies cards.

@cutterjohn
Why do you use the proprietary driver? Use the free one, "after all they just WORK".

Really, it's disabled for the HD 4890 as well? Argh.

RV770 includes RadeonHD 4890 cards too as far as I know.

The Radeon HD 4890 has the RV790 chipset codename but is in fact overclocked RV770 from the Mesa and Kernel perspective: it is handled like a RV770. That said, looking on Google, it looks like the RV790 has been slightly reworked in order to be more energy efficient, so this is not just a super-clocked RV770 from the engineering perspective.

[Dukenukemx]

Maybe. It's just that he specifically mentioned older cards, and hyperz is disabled for everything.

http://www.phoronix.com/scan.php?page=news_item&px=MTY2NjI