-=PESTILENCE=-
Gawd
- Joined
- Jun 10, 2004
- Messages
- 931
Well, the 8.3 Hotfix Driver has taken care of the powerplay issues guys.
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
http://forums.techpowerup.com/showthread.php?p=616941#post616941I'm not quite convinced this is an accurate characterization of what's happening. Think for a minute about some details of the explanation of PowerPlay's functionality. The embedded controller makes its decisions based on an evaluation of the contents of the command buffer. That suggests to me that if the GPU is clocking down, it's because the command buffer is either empty or nearly empty. I can easily envision a case where the GPU has completed a frame and vsync is enabled, causing the queue of pending operations to zero out long enough for the card to decide to downclock. Also, cases where data must be loaded or retrieved from the pagefile could easily trigger a downclock.
What I'm saying here is that you're making the assumption that Powerplay is triggeringing hitches. But what if you have your causality mixed up and it's the hitches that are triggering Powerplay?
That doesn't mean the issue isn't ATI's responsibility. There could be a driver bug leading to performance issues (like the framerate problem with cinematic playback in Crysis with pre-7.12 Catalyst), but that doesn't mean Powerplay is the cause.
While searching for this issue, I found a thread where this guy says it better than I can.
http://forums.techpowerup.com/showthread.php?p=616941#post616941
+1.That's entirely possible as well. I'm still waiting to hear something official from ATI about the whole thing. Even in Crysis, my GPU load is not 100% on the core I am monitoring. I don't think that's right.
This causes intense studdering when I'm going to another application (like Ventrilo or IE) and lasts for 3-5 seconds. The GPU jumps from 775 (I keep at stock) down to 297.
I thought this was a problem. I would put to you that while your playing a game you shouldn't dick around with other programs where you have to leave the game?
If that is the problem everyone is reporting, then its been happening for years....
Nice find! I'm going to try to remote to my desktop and implement this method. Then give it a try in a game. Hopefully tonight.Pestilence - you seem pretty knowledgeable here, have you taken a look at some of the test fixes? I'm definitely not savvy enough to try these, I'm still working on getting CCC 8.3 working :]
http://www.sapphiretech.com/en/forums/showthread.php?t=18798
Pestilence - you seem pretty knowledgeable here, have you taken a look at some of the test fixes? I'm definitely not savvy enough to try these, I'm still working on getting CCC 8.3 working :]
http://www.sapphiretech.com/en/forums/showthread.php?t=18798
Sounds like you have an overheating issue Pestilence (you mentioned the fan issue). I'm guessing that's what's causing your instability.
I noticed there are some fan control options in the XML profile file. I wonder if we could manually tweak the fan speeds from there also.
Went and looked at rivatuner and my temp had spiked to 100 degrees and shut down. Looks like the fan didn't kick in at all. Well, to clarify, it stayed at 22% fan duty cycle. This always happens with the bottom card. The fan profile is not working correctly on this card. The top card in the system however can stay at 40% fan duty cycle, forced through rivatuner and it stays just fine.
I noticed that also. I did change my voltages to the highest value for all 3 modes. That could cause a problem because in the game, it could be dropping to 2d clocks, which you have forced to the high values, but the voltages would be dropping to the low values. So you'd be running at high clocks on low volts.Hmmm...know what I just realized? I didn't change the voltage on the 2 lower power modes inside the XML...maybe that's what was causing the driver crash? I'll have to edit that when I get home and see what happens.
So nevermind looks like it works, does anyone know the default memory voltage ?