![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Unknown-One
[H]F Junkie
- Joined
- Mar 5, 2005
- Messages
- 8,909
In that case, I'm left with very few options as to what is really going on at AMD.
Option 1: AMD's engineers are so utterly incompetent that they designed UVD mode to function this way on purpose (when no other manufacturer uses the described bug-like behavior).
Option 2: This is an unintended flaw that cannot be corrected without a new hardware revision, and they're trying to cover themselves to avoid a massive recall.
Either way, not good. I did ask if there were any chance of changing the "expected behavior," we'll see what they have to say.
Also, it's not just videos that are actively playing that causes the problem. If you're using a hardware accelerated codec (like the default h.264 codec that comes with Windows 7, or hardware accelerated Flash on YouTube), and you decide to pause the video and minimize it to play a game, clockspeed scaling still won't work. You'll be stuck at 400 / 900 even though the video isn't even playing (you have to close the player entirely). Obviously, 400 / 900 is quite a bit slower than 725 / 1000, and leads to a considerable performance hit.
If I make the drivers glitch so that they jump to single-display clocks (even though I'm still using an extended desktop), I see flickering on secondary display heads when it jumps from 157/300 up to 725/1000, just like I see when it drops from 400/1000 to 400/900.
You will see flickering on secondary display heads when memclock changes, which is why the memclock is ALWAYS supposed to be 1000MHz on an HD 5850, and ALWAYS supposed to be 1200MHz on an HD 5870 when multiple monitors at attached (which is how the drivers behave now, except for the glitched UVD mode). Note, the flickering does not appear on the display defined in Windows as the "Primary Monitor," you need to have multiple displays attached with the desktop "extended" to them (not in an Eyefinity group).
For reference:
- 2D idle clocks on an HD 5850 with multiple monitors in extended desktop mode is 400 / 1000, which means the glitched UVD mode (400 / 900) is actually slower than what my 2D idle speed is supposed to be.
- 2D idle clocks on an HD 5850 with only a single display attached are 157 / 300. You'll never see flickering because the one attached display is always defined as "primary," so memclock is allowed to scale.
Option 1: AMD's engineers are so utterly incompetent that they designed UVD mode to function this way on purpose (when no other manufacturer uses the described bug-like behavior).
Option 2: This is an unintended flaw that cannot be corrected without a new hardware revision, and they're trying to cover themselves to avoid a massive recall.
Either way, not good. I did ask if there were any chance of changing the "expected behavior," we'll see what they have to say.
Yes, I expect that to work, because it works on all other manufacturer's video cards.I'm not sure why you are speechless, do you really watch youtube while you play a game? at least not in full screen, or run any other flash / uvd app.
Also, it's not just videos that are actively playing that causes the problem. If you're using a hardware accelerated codec (like the default h.264 codec that comes with Windows 7, or hardware accelerated Flash on YouTube), and you decide to pause the video and minimize it to play a game, clockspeed scaling still won't work. You'll be stuck at 400 / 900 even though the video isn't even playing (you have to close the player entirely). Obviously, 400 / 900 is quite a bit slower than 725 / 1000, and leads to a considerable performance hit.
As I've described previously in this thread, it does apply to all clock states.
If I make the drivers glitch so that they jump to single-display clocks (even though I'm still using an extended desktop), I see flickering on secondary display heads when it jumps from 157/300 up to 725/1000, just like I see when it drops from 400/1000 to 400/900.
You will see flickering on secondary display heads when memclock changes, which is why the memclock is ALWAYS supposed to be 1000MHz on an HD 5850, and ALWAYS supposed to be 1200MHz on an HD 5870 when multiple monitors at attached (which is how the drivers behave now, except for the glitched UVD mode). Note, the flickering does not appear on the display defined in Windows as the "Primary Monitor," you need to have multiple displays attached with the desktop "extended" to them (not in an Eyefinity group).
For reference:
- 2D idle clocks on an HD 5850 with multiple monitors in extended desktop mode is 400 / 1000, which means the glitched UVD mode (400 / 900) is actually slower than what my 2D idle speed is supposed to be.
- 2D idle clocks on an HD 5850 with only a single display attached are 157 / 300. You'll never see flickering because the one attached display is always defined as "primary," so memclock is allowed to scale.
Last edited:
