Announcement

Collapse
No announcement yet.

AMD "Hawaii" Open-Source GPU Acceleration Still Not Working Right

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • _SXX_
    replied
    Originally posted by drSeehas View Post
    Why another name for the same chip?
    I suppose for same purposes why 7XXX turn into R9 / R7.

    Leave a comment:


  • drSeehas
    replied
    SI

    Originally posted by agd5f View Post
    ... Oland is SI-based.
    SI = Southern Islands (R1000, GCN 1.0)
    or
    SI = Sea Islands (R1100, GCN 1.1)
    ?
    Curacao = pitcairn.
    Thx!
    Why another name for the same chip?

    Leave a comment:


  • agd5f
    replied
    Originally posted by drSeehas View Post
    Sure?
    AFAIR Oland was released after Bonaire (GCN 1.1).

    Curacao = Bonaire?
    Yes, I'm sure. Oland is SI-based. Curacao = pitcairn.

    Leave a comment:


  • drSeehas
    replied
    Oland GCN 1.0?

    Originally posted by agd5f View Post
    ... Southern Islands GPUs (GCN 1.0):
    ... Oland (8xxx, R7 240), ...
    Sure?
    AFAIR Oland was released after Bonaire (GCN 1.1).

    Curacao = Bonaire?

    Leave a comment:


  • drSeehas
    replied
    HD7790 is GCN 1.1

    Originally posted by metalliax View Post
    ...
    All the HD7xxx cards were GCN 1.0.
    ...
    No!
    The HD7790 is Bonaire as the 260/260X are.

    Leave a comment:


  • smitty3268
    replied
    Originally posted by whitecat View Post
    Note that the 7790 is supposed to be GCN 1.0 while 260X is supposed to be GCN 1.1 which is quite contradictory with the above information... Maybe the 7790 was like a GCN 1.1 prototype... The 7790 was released 14 months after the first 7000 series cards and only 7 months before the first 200 series cards.
    7790 was always GCN 1.1, and yes, it was the prototype they used to test yields of the architecture.

    Leave a comment:


  • swiftgeek
    replied
    Code:
    [Tue Jul 22 22:17:52 2014] nouveau  [  PTHERM][0000:01:00.0] temperature (90 C) hit the 'fanboost' threshold
    [Tue Jul 22 22:17:55 2014] nouveau  [  PTHERM][0000:01:00.0] temperature (86 C) went below the 'fanboost' threshold
    [Tue Jul 22 17:09:01 2014] nouveau  [  PTHERM][0000:01:00.0] temperature (96 C) hit the 'downclock' threshold
    [Tue Jul 22 17:16:16 2014] nouveau  [  PTHERM][0000:01:00.0] temperature (92 C) went below the 'downclock' threshold
    PTHERM has other means to downclock GPU, and AFAIK it has the same settings as nvidia blob
    And nvidia blob manages to get temperatures over 110?C (no dust, fresh paste, compal JHL90) . Idle temperature is just 15?C higher on nouveau, power consumption is probably much bigger issue
    Also i have noticed that nvidia somehow has memory leaks and uses way more cpu/ram… I just hope to reach 14 days uptime again
    (About my portal2 testing: i forgot to test any valve maps … and fps drops noticeably there to almost unplayable level)
    Last edited by swiftgeek; 22 July 2014, 06:33 PM.

    Leave a comment:


  • Rakot
    replied
    Originally posted by swiftgeek View Post
    I just tested nouveau again solely because of this comment, same GPU, without blobbed FW this time
    My performance in portal2 is now comparable to nvidia's driver (slightly worse), due to default high clocks
    Maybe it's time for modifying VBIOS before attempting honest benchmark? At least for the NV50
    I would definitely switch to open drivers but there are certain problems. First, this is a laptop. Second, the GPU I have (GT 9600M) is really very-very hot even while using nvidia driver. So I'm not sure how safe it would be to use nouveau since it has obvious problems with power management.

    Leave a comment:


  • akincer
    replied
    Originally posted by Espionage724 View Post
    Meanwhile, some people will swear up and down AMD's drivers caused their GPU to explode or some bs whenever they touched the GPU with a screwdriver or intentionally overvolted/clocked it.

    Umm -- the power management on the OS drivers still causes at least some chipsets to get WAAAAY too hot. I mean as in uncomfortable to touch your laptop near your heat exhaust. It turns your laptop essentially into a mini-space heater and/or noisemaker from the sound of the system fans. This happens even if you have a USB cooling platform hooked up BTW.

    I deeply applaud AMD for their OS efforts and focus on buying their products because of it. But lets not pretend that things are hunky dory and that there aren't real issues with the drivers that can necessarily damage your system.

    Leave a comment:


  • swiftgeek
    replied
    Originally posted by Rakot View Post
    Say thanks to nvidia with their "the way it's meant to be played" stuff. Fglrx follows specs more closely than nvidia but developers just don't care (see for example recent commits to mesa where Marek made quirks for Unigine software which do not follow specs but "just works" on nvidia).

    Use fglrx. It just works. I've been using fglrx for more than 5 years and it worked most of the time. Yes, there are certain problems but at least it never crashed my kernel at random times.

    Tell me moar about it. I have 9600GT m in a laptop and you know what? Nvidia has only legacy support for it. Moreover there are kernel crashes from time to time which appeared some time ago and which nvidia is not going to fix anyways (there was a lengthy thread in nvidia's forum).

    So all your facts are the same as MS's "get the facts".
    I just tested nouveau again solely because of this comment, same GPU, without blobbed FW this time
    My performance in portal2 is now comparable to nvidia's driver (slightly worse), due to default high clocks
    Maybe it's time for modifying VBIOS before attempting honest benchmark? At least for the NV50

    Leave a comment:

Working...
X