Download & Discuss AMD Catalyst 15.4 Beta Driver for Windows OS

Discussion in 'Videocards - AMD Radeon Drivers Section' started by LtMatt81, Apr 13, 2015.

  1. dirtula

    dirtula Member

    Messages:
    16
    Likes Received:
    1
    GPU:
    Asus RX580 8gb
    BSOD, after a little time with firefox and youtube. I do have the bioses modded on my 7870's, been so long i don't really remember how they are configured, but this never happened with any other driver.
     
  2. realmadrid

    realmadrid Member Guru

    Messages:
    177
    Likes Received:
    0
    GPU:
    GTX 1050 TI MSI OC
    hi i have purche new card have hd 6770 go1 now hd 7970 omg is amzing card thank you amd now what the best driver this card ? sorry my bad english
     
  3. Romulus_ut3

    Romulus_ut3 Master Guru

    Messages:
    708
    Likes Received:
    170
    GPU:
    AMD RX 570 4GB
    I don't get it, if users are reporting TDRs with the latest Beta driver, something which didn't occur on an earlier version of the drivers, and the problem is fixed when reverted back to the earlier version of the driver, suddenly the card's BIOS or the card itself gets deemed faulty? The amount of fanboyism and effort put into defending AMD's half baked drivers is simply astonishing.
     
    Last edited: Apr 30, 2015
  4. Espionage724

    Espionage724 Guest

    UVD clocks has been a well known issue, and has shown itself in quite a few driver releases, while being unnoticed in others.

    It's not about "fanboyism"; it's about placing the blame where it's supposed to go.

    If you have problems with TDR/BSOD when dealing with Flash content or even videos in-general, check your GPU memory clocks before and after starting such content. If my GPU memory is 1225MHz, and flash content takes it to 1220MHz, that's an issue.

    Your GPU has a reference speed. You card either comes at said reference speed, or was overclocked either by the GPU vendor or yourself.

    (AMD) GPUs have a few clock states that determine what clockseed the GPU runs at and the voltage, and switch based on usage and content being ran; Boot (used at boot), 2D/Idle, 3D/High-performance, and UVD. UVD is used for video decoding. 2D/Idle might run at 150MHz/300MHz (core/mem), while 3D might be 1000MHz/1200MHz. Factory-OC cards and user-overclocking usually only affect 3D/High-performance clocks, and nothing else.

    Changing GPU memory clock speeds under-load or under other conditions (using a certain chip-type, single/multi-monitor, etc) can cause instability. This is mostly noticed under multi-monitor setups (when the GPU memory changes, you'll see brief artifacting on one or more monitors). High-refresh rate single-monitors and multi-monitor setups typically run both 2D and 3D clock states with the 3D GPU memory speed.

    UVD is a clock state that is slightly higher than 2D clocks, but below 3D clocks, and as mentioned, is used mainly for video decoding (so that video decodes smoothly, but without wasted power). Anything using hardware-accelerated video decoding (DXVA from something like MPC-HC or Flash content) will switch the GPU to this clock state, and it will take priority over anything else (hence if you start flash content while a game is open, you'll see your clock speeds drop).

    So with all that in-mind:

    - UVD clocks are higher than 2D, but under 3D
    - Changing GPU memory speeds in certain scenarios can cause instability
    - User or Factory OC only touches 3D clocks
    - When UVD content is started, your GPU will go from whatever clock state it is in, and go to UVD
    - If your GPU memory is at all different between 3D and UVD (which it likely is if your GPU memory is above reference), you can have instability.

    As for options on how to fix it (and TLDR; just want fixes):

    - Disable PowerPlay (can use MSI Afterburner)
    - Downclock your GPU memory to reference speeds (you need to lookup reference specs for your card)
    - Request an updated VBIOS from your vendor
    - Fix the VBIOS yourself (on applicable cards compatible with VBIOS modding/flashing; you simply need UVD GPU memory to match 3D memory speeds)
    - Don't use HW-accelerated video decoding (on a per-app basis; disable HW acceleration on Flash, DXVA on MPC-HC, etc)
    - Downgrade your driver (which is silly imo considering one of the above is better)

    My 7850 card is factory OC, but only on the core clock. This leads me to believe that MSI knows about this issue, and chose not to mess with the memory (i'm unsure of other cards though).

    As for why instability occurs on some drivers/setups but not others; it's just like overclocking. My 4.5GHz on my FX-8350 may work fine at 1.48V, but someone else might have instability. My 4.5GHz might also prove stable with prime95, but not with Guild Wars 2.

    In an ideal world; UVD would either copy the 3D clock speed for memory, user overclocking would affect UVD memory clocks, or vendors putting out factory OC cards would modify UVD memory clocks. "Maybe" AMD can do something about this on a driver-level (given there isn't some specific reason (like HDCP) for the clock speed on UVD anyway).
     
    Last edited by a moderator: Apr 30, 2015

  5. Romulus_ut3

    Romulus_ut3 Master Guru

    Messages:
    708
    Likes Received:
    170
    GPU:
    AMD RX 570 4GB
    I've been editing and playing with VBIOS flashing since the AGP era and I appreciate your concerns on trying to educate me on different clock states of video cards, but I'm very well aware of what they are. I do not have any experience with multimonitor setups so that is an unknown territory for me, however it is highly unlikely for device drivers to manipulate the memory clocks to something different than what it is set in it's BIOS. If it's happening, then there's something else at play.
     
    Last edited: Apr 30, 2015
  6. Espionage724

    Espionage724 Guest

    You can simply check if what I mentioned is happening by opening GPU-Z up when there isn't flash content playing, and then see if the clocks (both core and mem) change when flash content is started.

    The short part is; if you overclock your GPU or it comes factory OC'd, it's likely only on the 3D clocks, and not UVD. If the GPU memory clock is different between 2D/3D and UVD either in the VBIOS or userspace, you'll likely have issues when anything uses UVD.

    Here is my Idle (multi-monitor; note the 3D-level memory clock of 1250MHz):
    [​IMG]

    Here is my 3D clocks (at 1050MHz/1250MHz as expected):
    [​IMG]

    Here is my UVD clocks with 3D content going on in the background (aka going from 3D clocks to UVD):
    [​IMG]

    And finally, my UVD clocks going from 2D/Idle:
    [​IMG]

    So without 3D content in the background, UVD changes the core clock up to what I assume is the min UVD clock in the VBIOS, but forces my memory from my user OC levels down to the reference clock of 1200MHz (which is what you don't want happening).

    With 3D content, UVD takes priority and shifts the core clock down to the factory OC level (900MHz) and drops the memory to reference levels (1200MHz; once again you don't want this happening).
     
    Last edited by a moderator: Apr 30, 2015
  7. xxela

    xxela Master Guru

    Messages:
    231
    Likes Received:
    8
    GPU:
    RX6800 XT Red Devil
    @Espionage724 I always have the memory clock working at full speed when I play hardware accelerated content regardless of what profile I use ( reference or overclock speed); and even on non hardware accelerated stuff I see the memory clock bouncing from idle to full speed all the time. I see no UVD state, for memory clock just idle and full speed. With that said the Omega and beyond drivers (15.3,15.4) have proved to be rock stable for me (ASUS R9 290 reference model). Maybe this clock states are specific for 7000 series?
     
  8. Di0g0

    Di0g0 Active Member

    Messages:
    62
    Likes Received:
    0
    GPU:
    Sapphire HD7850 Dual-X
    "Downgrade your driver (which is silly imo considering one of the above is better)"

    As I said, this problem only happens with THIS beta driver (15.4), so the only silly thing is to do the other things instead of downgrading to a previous version. Btw, thank you for the explanation ;)
     
    Last edited: Apr 30, 2015
  9. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    You are correct.

    290+ dont have discrete UVD levels, they are dynamic.

    However, the individual having the problem is on a 7xxx
     
    Last edited: Apr 30, 2015
  10. Valerys

    Valerys Master Guru

    Messages:
    393
    Likes Received:
    18
    GPU:
    Gigabyte RTX2080S
    I was just about to post that I have no issues on my R9 290 with Flash acceleration and never did. Oh well
     

  11. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    neither do i
     
  12. Espionage724

    Espionage724 Guest

    If you want to (or need to for one of the improvement/fixes) use this driver, using one of the other fixes would be a better option considering you get to keep the driver and resolve the issue (given that it's the cause for the instability anyway).

    Ah, thanks for the info; I just assumed it affected all cards.
     
  13. xxela

    xxela Master Guru

    Messages:
    231
    Likes Received:
    8
    GPU:
    RX6800 XT Red Devil
    So, then what was the cause for the instability of early 14 driver branch encountered by many R.200 series users with flash content?
     
  14. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    That was a flash issue, Adobe supposedly patched it.
     
  15. Noisiv

    Noisiv Ancient Guru

    Messages:
    8,143
    Likes Received:
    1,413
    GPU:
    2070 Super
    Annoying as ***
    although "half baked" is not how I would describe my 290 experience.

    Flash acc. used to cause issues for me pre 14.4; seems fixed now, but...

    I would certainly hope that AMD's threshold for an issue to be considered a bug is lower than all people having an issue all the time :infinity:
    Why some people think that their (rock-solid) system is more relevant than the other guy's is beyond me.
     

  16. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    i dont think its a revency issue i think its more to mitigate the fact that when most people post having a problem, they tend to take the tone that "all the things are broke"

    Pointing out that an individual doesnt have a problem is just trying to tone it down a bit, and that not everyone has that particular issue.
     
    Last edited: Apr 30, 2015
  17. Di0g0

    Di0g0 Active Member

    Messages:
    62
    Likes Received:
    0
    GPU:
    Sapphire HD7850 Dual-X
    I understand, but in my case I will wait for a new driver and see what happens. In my case there is now reason to use this new beta driver, so I will wait for a new one.

    Btw, sapphire website doesn't have any vbios on their support page, is it normal? (I don't want to flash a vbios, is too risky :( ).

    Thanks ;)
     
  18. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
  19. Di0g0

    Di0g0 Active Member

    Messages:
    62
    Likes Received:
    0
    GPU:
    Sapphire HD7850 Dual-X
    Yeah, I know that page, but i can't find any changelog between vbios versions :p
     
  20. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    you arent going to.

    they are directly downloaded from people's actual cards.

    as long as you get one that matches your card, flashing should be safe.

    use GPU-Z to see your bios version, you may have the latest for your card anyway.
     
    Last edited: Apr 30, 2015

Share This Page