Hello guys, I guess most of you know already about the bug GCN 1.0 cards have. I mean UVD bug. I think it is safe to assume that AMD will not bother to fix it. That's why in order to get rid of the bug as much as possible I decided to use custom bios. I modified clocks of UVD state. The clocks of the first entry I made similar to 2D state and the clocks of the second entry I made same as closk of the fully loaded state. Thats how it looks: The questions I have to you is the following: Is it okay to set clocks of the UVD(#5 entry from the image) same as clocks of the 2D(#4 entry) considering the fact that the voltages can not be changed? Thanks in advance.
I already said that is the fix some time ago. Still to this day, even using latest 16.12.2 i dont have UVD bug. Instead of listening some people choose to only to report issue to AMD over and over and expecting a fix (which never came). Spoiler http://forums.guru3d.com/showpost.php?p=5329225&postcount=7
Yes, I have seen your post. That's why I decided to edit bios. But how about them? Is it okay to keep them this low? I lowered them on purpose, decided to match these values to the default 2D ones. In default version they were 501 and 1500. Now if I am stuck in UDV I will not even notice, neither game performance nor idle clocks will be affected... I think.
Vbe7 tutorial : "There are some limitations though... You cant change 2D / UVD voltages You cant set 2D / UVD clocks higher than the values stored in BIOS (these limitations are intentional, without these limits powerplay would go nuts...) Idk, only way is to try it yourself. I never did try changing 2d.
I actually believe that a fix will come. AMD is in a big rewrite of their 2D driver currently, they want it to be cross-platform compatible. Add to that the introduction of two new architectures and Wattman, and you can get into a mess. That doesn't mean they won't fix it. Keep reporting it. You are correct about VBE. It can't change the 2D voltage/memory values at all, for example. I have an unlocked version from the author, but he was right, it only messes up the way Catalyst (back then) was working. It was the time of the 2D bug and I was trying to put the 2D memory clock to something like 900, but it wasn't working. I'm not sure how the card will behave if someone changes the 500/1500 clock. You have to have some kind of reliable monitoring software to check, and a clean reinstall (using DDU), is a MUST in order to properly check it.
I actually believe that a fix will come. AMD is in a big rewrite of their 2D driver currently, they want it to be cross-platform compatible. Add to that the introduction of two new architectures and Wattman, and you can get into a mess. That doesn't mean they won't fix it. Keep reporting it. If you see the new driver thread, there is a general issue with 2D clocks and voltages, it's not only us now. This Phoronix article is a good starter on what they are trying to achieve. Here's the direct link to the PDF presentation. For all intents and purposes, the display driver is being rewritten from scratch and open sourced. You are correct about VBE. It can't change the 2D voltage/memory values at all, for example. I have an unlocked version from the author, but he was right, it only messes up the way Catalyst (back then) was working. It was the time of the 2D bug and I was trying to put the 2D memory clock to something like 900, but it wasn't working. I'm not sure how the card will behave if someone changes the 500/1500 clock. You have to have some kind of reliable monitoring software to check, and a clean reinstall (using DDU), is a MUST in order to properly check it.
I have reinstalled drivers (using DDU) and so far nothing weird has happened. Both gaming and benchmark tool performance is as it is supposed to be. Video player performance is also not affected despite changed 500/1500 values. But I think time is needed to be sure that everything is stable. If nothing weird will occur I will use this bios until/if amd fixes the bug.
