Page 3 of 3 FirstFirst 123
Results 21 to 29 of 29

Thread: AMD Radeon R9 290 Is Still A Wreck On Linux

  1. #21
    Join Date
    Oct 2008
    Posts
    3,035

    Default

    Quote Originally Posted by deanjo View Post
    Not sure how you could paint the 290's support in Linux as anything but a wreck.
    When i read the word "wreck", i assume it doesn't work at all. Because when my car is in a wreck, it's completely stopped. 290 support seems to work just fine, it's just way slower than it should be.

    But i agree things are really bad in terms of performance.

  2. #22

    Default Precisely!

    Quote Originally Posted by smitty3268 View Post
    When i read the word "wreck", i assume it doesn't work at all. Because when my car is in a wreck, it's completely stopped. 290 support seems to work just fine, it's just way slower than it should be.

    But i agree things are really bad in terms of performance.
    I agree completely!

    A "Wreck" would mean I would not have been able to play Metro LL maxed out, nor Serious Sam BFE maxed out, nor Wargame EE maxed out, nor smooth graphics with Valley, Heaven, Tropics, Sanctuary!
    Well, for the games above, I played them all MAXED out! Even overriding AA and Anisotropic filtering and dialing up the settings to MAX with Edge-Detect AA on CCC.

    I agree the Windows benchmarks and their Nvidia counterparts show us how far the R9-290(X) can go still. But it isn't a hardware problem/limitation, NOR a problem (or wreck) at all since all the CURRENT hardest hitting games on Linux can be MAXED out with smooth playing; with consumers getting what they want out of the games currently available to play from Steam or Desura!

    As for the "FPS" difference between Windows benchmarks and Nvida counterparts; it is just a matter of time as Valve and AMD have already said that AMD-SteamMachines and their specs will be out next year AND that won't happen without AMD sorting out the driver to make significant improvements. By the latter half of next year, it should be close to parity if not better (*better for Mantle enabled games).

  3. #23
    Join Date
    Jan 2010
    Location
    Somewhere in Kansas.
    Posts
    247

    Default

    All the games tested though (on high settings) played at higher than 60fps... so.....

    I don't see why you would consider it a wreck. Since the lower than expected performance isn't a problem for current games on linux. As long as they get the performance improvments in before more demanding games show up, I'm ok.

  4. #24
    Join Date
    May 2013
    Posts
    38

    Default Wreck is too harsh

    Quote Originally Posted by ua=42 View Post
    All the games tested though (on high settings) played at higher than 60fps... so.....

    I don't see why you would consider it a wreck. Since the lower than expected performance isn't a problem for current games on linux. As long as they get the performance improvments in before more demanding games show up, I'm ok.
    Agree.

    "Wreck" normally means that it's completely dead. Apparently it's not the case here.
    True is that R9 290 on Linux doesn't run as well as it is on Win, it's still capable to deliver 60+FPS in all games in benchmark.

    In the other hand, I quite understand Michael's feeling about his near $400 investment on this card.

  5. #25
    Join Date
    Jul 2008
    Location
    Berlin, Germany
    Posts
    821

    Default

    Quote Originally Posted by Sdar View Post
    I didn't know how people still thought that AMD has better OpenCL performance...

    Yeah it has better performance when is a simple calculation.... if you try to make something bigger then the performance drops drastically and eventually just stop working.
    Which people thought that? Anybody with minimal knowledge about GPU compute can tell it more precisely than that. For one, don't forget that NVidia consumer cards are intentionally crippled in DP (a Radeon HD7970 is more than four times as fast as a Geforce GTX 680 in Linpack). AMD usually does better for workloads which are ALU limited. NVidia often does better for workloads that are limited by something else (e.g. I/O).

    A nice explanation can be found here: https://en.bitcoin.it/wiki/Why_a_GPU...Nvidia_GPUs.3F

  6. #26

    Default

    I hope AMD improve the drivers before Rome II Total War comes out on Linux!

  7. #27
    Join Date
    Aug 2007
    Posts
    6,607

    Default

    Every time i update my repository with the latest fglrx drivers i think later that it was more or less useless effort. Ok, with the latest 13.11 beta 9.4 it gained kernel 3.12 support, but thats it. fglrx driver is pure crap with my hd 5670 and l4d2 on linux. even a gt630 kepler is feeling better in that game and that card is slower due to the stupid lag with the autodetected settings. ok, you could use lowest settings but why does the game detect higher ones... the fps jumps from 50 to 299 and back all the time, if it would not jump that often it would be much better. I think a highend card would not be that critical as the raw performance is higher but i still would only buy nvidia for linux. Now killing floor got some nvidia bugfixes as well (the only game with nvidia rendering errors). Most likely the oss drivers would work better with source engine than catalyst - maybe i get mesa 10 running with wheezy, lets see...

  8. #28
    Join Date
    Mar 2012
    Posts
    280

    Default mesa 10

    if you use a ppa or repository its not hard to get it running - and with that card you definitely go oss. You will have regressions in games that use openGL 4 like serious sam but almost all other games should work smoother now than with catalyst, from all I have heard so far, cause the fps is more steady AND on the same absolute level.

    compare
    http://www.phoronix.com/scan.php?pag...ing_gpus&num=3
    http://www.phoronix.com/scan.php?pag...nux_gpus&num=5

    first is mesa, second fglrx
    counter strike source, nexuiz and openarena 0.8.8 fps is even much better with mesa using a HD 6870 while an HD 6950 is a bit behing fglrx. So it depends on the card, very...

  9. #29
    Join Date
    Sep 2007
    Location
    Europe
    Posts
    10

    Default updated test possible?

    any chance someone can do a quick rerun of the test with the 290 and the 14.1 beta 1.3 driver?
    Lockheed

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •