The AMD Driver Bug Thread

One more email from AMD. I am not pleased...

They still can't seem to figure out the idle speed problem. If their card is working the way they say it's working, then their screens would be flashing constantly because their memory clock isn't staying at 1000 MHz... hopefully they don't consider that to be "correct"
It's also worth noting that the tables I defined were dumped from their video BIOS, they aren't just base on observation.
No actually I have tested this behavior while two monitors where attached to the graphics card (1x NEC EA231 WMi DP and 1x Samsung SyncMaster T173 DVI) in extended mode and I see 157MHz of idle value.
Click to expand...

Worse still, we have a "WILL NOT FIX" on the list. They are outright refusing to fix their broken UVD mode. They seem to think your screen(s) flickering every time you open or close a video is acceptable (not to mention a severe drop in 3D performance due to being stuck at the wrong clockspeed).
The behavior I explained under item number 5 does not exhibit any problem, as such there will be no fix. AMD has no plans to change the priority of UVD state over Performance state at this point..
Click to expand...

I'm so fuming mad I haven't responded yet. Need to cool off first...
 
It's things like this that make me want to go (back) to the green team after 4 ATI cards.
 
Bah... at this point it would be interesting to do a comparison of this issue with a nvidia card, sending results to AMD. I suppose, with a GeForce GTX260 or similar, this issue never happened:)
 
Decided to actually call in to a manager rather than continuing to deal with email (which isn't getting us anywhere). He's going to check in on the support tickets and see what the deal is.

He'll call back tomorrow...guess we'll wait and see what he has to say.

Edit: I've got a conference call scheduled for Monday at noon, we'll try and figure out how best to go forward. Stand by!
Edit2: Make that Wednesday, someone caught the flue...

weah said:
Bah... at this point it would be interesting to do a comparison of this issue with a nvidia card, sending results to AMD. I suppose, with a GeForce GTX260 or similar, this issue never happened:)
Click to expand...
I actually have a GTX260 in this box along with my HD 5850. The intention was to use it as a PhysX card, but now it acts as a crutch for the HD 5850 and handles all hardware accelerated video decoding (CoreAVC with CUDA enabled). It's pretty ridiculous that you have to use a competitors card simultaneously to avoid bugs and glitches like this...
 
Last edited:
regarding the opengl issues after 10.4, there seem to be 2 major things i'm seeing & i have no idea how far into single cards or newer cards the issues reach

first, i ran into games that do NOT get fixed by using atioglxx, but by using atigktxx http://www.kn00tcn.net/site/cities-in-motion-startup-crash-fix/
single 5800 card users have claimed this fix even works for them
the atom zombie smasher dev even put atigktxx into the game folder, i wonder what implications that has on cards that dont support 10.4

then only recently did i find something that gets fixed with atioglxx, & that's cityengine http://www.kn00tcn.net/site/fix-for-cityengine-startup-crash/

then riddick dark athena was broken vsync+cf from 11.1 to maybe 11.3 (i uninstalled before the driver that had riddick in the notes came out)


as for UVD stuff, i kinda like how my 4870x2 is not as 'smart' as these new cards, simple 2d = desktop, 3d = fullscreen, no uvd or jumping between 2d/3d mid game (& i render videos on the cpu anyway so...)
 
Enabling Morphological filtering (MLAA) in 3D Application Settings on 11.8 preview causes Dirt 3 to look terrible and run at a very low framerate in Eyefinity/CrossFireX @4800x1200 w/ bezel management. I have not tried other games. Unchecking it returns performance to expected levels.

I tried playing around with different settings in Dirt 3 to see if something was conflicting and the MLAA was fighting with other AA settings, to no avail.
 
Alright, had that conference call, here's where we stand.

- The issues with UVD on the HD 5000 series of cards will not be resolved. You will continue to experience flickering and slowdowns unless you mod your video BIOS.
- I have AMD's blessing to distribute pre-modified HD 5000 series video BIOS that alleviate the symptoms of the UVD bug.
- According to AMD, UVD behaves properly on the HD 6000 series (I'll test that for myself...)
- They are still investigating the issue involving AMD Overdrive and incorrect idle clockspeeds being set.

They're also sending me an HD 6970 (and I get to keep this one). It's nice of them, and I really do appreciate it... but it doesn't help everyone else who's stuck on an HD 5000 series card. They're basically giving early adopters the finger here :(

Edit: Some good news. As of Catalyst 11.7, the driver nolonger spams your System logfile with "UVD Information" while playing videos. Yay!

kn00tcn said:
regarding the opengl issues after 10.4, there seem to be 2 major things i'm seeing & i have no idea how far into single cards or newer cards the issues reach

first, i ran into games that do NOT get fixed by using atioglxx, but by using atigktxx http://www.kn00tcn.net/site/cities-in-motion-startup-crash-fix/
single 5800 card users have claimed this fix even works for them
the atom zombie smasher dev even put atigktxx into the game folder, i wonder what implications that has on cards that dont support 10.4

then only recently did i find something that gets fixed with atioglxx, & that's cityengine http://www.kn00tcn.net/site/fix-for-cityengine-startup-crash/

then riddick dark athena was broken vsync+cf from 11.1 to maybe 11.3 (i uninstalled before the driver that had riddick in the notes came out)
Click to expand...
I mentioned this to them over the phone. They'll probably have to implement some game-specific fixes (like those added for games that use the Quake III engine).

Can you confirm that the fixes in those blog posts are still required on Catalyst 11.6?
 
Last edited:
@Unknown-One: System spam errors are not completely fixed. AMD fixed only 1 spam event error (Event ID 62464 is related only to Amdkmdag issue):

* Log event (Event ID 62464) issues when playing video content have been resolved
Click to expand...


So, this is the actual status:
-Amdkmdag spamming officially fixed on 11.7
-Dcom errors still not fixed on 11.7/11.8 preview: http://forums.guru3d.com/showthread.php?t=343572
--ACEEventLog 'Information' errors Not fixed until 11.6, and not considered in their 11.7 changelog. I don't want install CCC anymore until this crap is officially fixed:
http://www.blindnero.com/forums/phpBB3/viewtopic.php?f=15&t=2273&p=7835#p7837


Unknown-One said:
Can you confirm that the fixes in those blog posts are still required on Catalyst 11.6?
Click to expand...

I think you can read directly the answer done in this topic (11.8p no change):
http://www.rage3d.com/board/showpost.php?p=1336654661&postcount=22
 
Last edited:
Unknown-one - regarding the statement "It's pretty ridiculous that you have to use a competitors card simultaneously to avoid bugs and glitches like this..." While it sucks you are having issues, you can only speak for yourself, not everyone, you should of said "It's pretty ridiculous that I have to use a competitors card simultaneously to avoid bugs and glitches like this..." Generalizations do nothing to help your cause.
 
weah said:
@Unknown-One: System spam errors are not completely fixed. AMD fixed only 1 spam event error (Event ID 62464 is related only to Amdkmdag issue)
Click to expand...
Thanks, I'll just split that into two separate bugs.

darkstarcow said:
Unknown-one - regarding the statement "It's pretty ridiculous that you have to use a competitors card simultaneously to avoid bugs and glitches like this..." While it sucks you are having issues, you can only speak for yourself, not everyone
Click to expand...
No actually, I CAN speak for everyone on this matter. When I made that statement I was talking about issues with UVD mode on an HD 5000 series graphics card:
- AMD has confirmed that the flickering and underclocking is a problem with UVD v2.
- ALL cards with UVD v2 will experience this problem (which means the entire HD 5000 series).
- NONE of the effected cards will be fixed.

This is the last word from AMD themselves, unless something drastic changes and they decide to devote resources to fixing the HD 5000 series again.

darkstarcow said:
you should of said "It's pretty ridiculous that I have to use a competitors card simultaneously to avoid bugs and glitches like this..." Generalizations do nothing to help your cause.
Click to expand...
Incorrect. This is not a generalization, and it's not just me, this is a fact for anyone with an AMD graphics card that uses UVD v2 (the entire HD 5000 series). There is also no cause to be helped, AMD is no-longer devoting resources towards the HD 5000 series of cards, the problem will not be fixed.

If you have an HD 5000 series card, and you want to decode video using hardware acceleration without the glitches inherent in UVD v2, you will have to use an auxiliary Intel or Nvidia card to do your decoding. End of story. Your only other option is to replace the whole card with one from the HD 6000 series (which has UVD v3).
 
Last edited:
Unknown-One said:
- I have AMD's blessing to distribute pre-modified HD 5000 series video BIOS that alleviate the symptoms of the UVD bug.
Click to expand...

I need some confirms about this point.

1) I've an ATI HD5850 Sapphire 1GB Toxic. Is this a reference model card or not?
2) I'll lose my warranty if I flash my video bios with your pre-mod BIOS?
 
Okay, I sold my HD5870 CF (XFX - 2 cards) around the 10.4 driver era, but I never experienced any of the mentioned UVD issues with playing bluray and local files (iirc, flash hadn't been GPU accelerated, yet)... I dunno, I had two reference PCB cards bought around the GTX480 launch, one XXX factory OC'd to 875, and one vanilla at 850 stock clocks. I've also never had any DXVA related issue (since I used MPC-HC for local file playback), bar one of the driver or MPC-HC revisions that rendered all vides a sold green screen...


I dunno, maybe I simply didn't have enough types of files and blurays? I used Roxio CinePlayer and TMT back then. Never had an issue to the day I sold my HD5870s...I think I only had my U2311h connected via DVI, since I had a busted DP cable and didn't want to buy another. I guess a later driver revision screwed UVD up?
 
weah said:
1) I've an ATI HD5850 Sapphire 1GB Toxic. Is this a reference model card or not?
2) I'll lose my warranty if I flash my video bios with your pre-mod BIOS?
Click to expand...
1) That is a non-reference card. You can either mod the BIOS yourself, or send me a copy so I can mod it and post it.
2) That will depend upon what Sapphire covers in their warranty.

jeremyshaw said:
Okay, I sold my HD5870 CF (XFX - 2 cards) around the 10.4 driver era, but I never experienced any of the mentioned UVD issues with playing bluray and local files (iirc, flash hadn't been GPU accelerated, yet)... I dunno, I had two reference PCB cards bought around the GTX480 launch, one XXX factory OC'd to 875, and one vanilla at 850 stock clocks. I've also never had any DXVA related issue (since I used MPC-HC for local file playback), bar one of the driver or MPC-HC revisions that rendered all vides a sold green screen...
Click to expand...
I can guarantee you that your cards jammed at 400MHz core / 900MHz memory when UVD mode was engaged by hardware accelerated video playback, and would not return to normal scaling until UVD mode was released by closing the video. That is simply how UVD v2 operates, period (this behavior is hard-coded in the video BIOS, I'm afraid). If you had left a video that was using DXVA paused, your 3D performance would have suffered due to your card not being able to enter 3D mode.

It sounds like you only ever ran a single monitor. The flickering issue does not present itself if you only have a single monitor (if you configure multiple monitors in an eyefinity display group, this also counts as a "single" monitor). Without that warning sign, it sounds like you never really took the opportunity to scrutinize how your cards were behaving.

jeremyshaw said:
Never had an issue to the day I sold my HD5870s...
Click to expand...
I assure you, even if you didn't realize what was happening, you experienced the described UVD mode behavior.

jeremyshaw said:
I guess a later driver revision screwed UVD up?
Click to expand...
Nope, UVD v2 has acted this way since its inception in 2009, well before 10.4 (not that it has anything to do with the drivers, it's a hardware/BIOS issue)
 
The biggest bug in AMD's driver package is that pile of crap Catalyst Install Manager. I've been having to switch between different drivers and SDK's to check performance on each and I've never seen something be so irritating than when the Install Manager gets hung halfway through the install or when I go to uninstall it does the same leaving me to have to reformat Windows just to get the right drivers installed and working again. They need to scrap that installer and start over.
 
Last edited:
Unknown-One said:
I can guarantee you that your cards jammed at 400MHz core / 900MHz memory when UVD mode was engaged by hardware accelerated video playback, and would not return to normal scaling until UVD mode was released by closing the video. That is simply how UVD v2 operates, period (this behavior is hard-coded in the video BIOS, I'm afraid). If you had left a video that was using DXVA paused, your 3D performance would have suffered due to your card not being able to enter 3D mode.

It sounds like you only ever ran a single monitor. The flickering issue does not present itself if you only have a single monitor (if you configure multiple monitors in an eyefinity display group, this also counts as a "single" monitor). Without that warning sign, it sounds like you never really took the opportunity to scrutinize how your cards were behaving.


I assure you, even if you didn't realize what was happening, you experienced the described UVD mode behavior.

Nope, UVD v2 has acted this way since its inception in 2009, well before 10.4 (not that it has anything to do with the drivers, it's a hardware/BIOS issue)
Click to expand...

Ahh....I see, thank you. I sold my CF HD5870s back then, over what I thought was utter crap crossfire scaling (I never took the time to test each card individually), perhaps it wasn't bad CF scaling/stuttering, but rather the clocks never consistantly reaching full speeds, lol... Oh, well. I have a single GTX580, waiting on a HD7970 (or something) for my scheduled jump back and forth between GPU vendors every gen. :p
 
Last edited:
jeremyshaw said:
Ahh....I see, thank you. I sold my CF HD5870s back then, over what I thought was utter crap crossfire scaling (I never took the time to test each card individually), perhaps it wasn't bad CF scaling/stuttering, but rather the clocks never consistantly reaching full speeds, lol... Oh, well. I have a single GTX580, waiting on a HD7970 (or something) for my scheduled jump back and forth between GPU vendors every gen. :p
Click to expand...

It was, but they fixed it with Cayman.
 
Last edited:
burningrave101 said:
The biggest bug in AMD's driver package is that pile of crap Catalyst Install Manager. I've been having to switch between different drivers and SDK's to check performance on each and I've never seen something be so irritating than when the Install Manager gets hung halfway through the install or when I go to uninstall it does the same leaving me to have to reformat Windows just to get the right drivers installed and working again. They need to scrap that installer and start over.
Click to expand...



Sorry try again, the "Format" windows fix just doesn't work anymore as an excuse to make it sound worse than it is.

We are well beyond the win98se/WinME days of having to reformat..
 
H-street said:
Sorry try again, the "Format" windows fix just doesn't work anymore as an excuse to make it sound worse than it is.

We are well beyond the win98se/WinME days of having to reformat..
Click to expand...

I'm not trying to make it sound worse than it is, I'm running Windows 7 x64 and the only solution that fixed the Catalyst Install Manager refusing to install or uninstall the driver package is to reinstall the OS and start fresh. Using a driver cleaner in Safe Mode or deleting ATI directory files or whatever else I tried didn't fix it. And the issue is prevalent enough when switching between driver packages and different series of AMD/ATI cards that it became quite a bit of annoyance after happening several times the past few weeks while I was going between 5XXX series and 6XXX series single and dual GPU cards.
 
Last edited:
Unknown-One said:
Ok, I can complain, they sent me an HD 5870 with a BIOS from September 14th, 2009, which means this particular card has been kicking around AMD since the press event on the USS Hornet (September 10th, 2009) and predates the actual launch date of the cards (September 23rd, 2009) by a wide margin.

Guess what a BIOS this old brings along with it? More bugs from launch-day! My 3rd monitor, the one connected to DisplayPort via an active adapter, is blinking on-and-off at random right now. I saw this same behavior on my HD 5850 as well, but it was fully resolved by a BIOS update shortly after launch.
Click to expand...

this thread is not inspiring confidence in AMD's operations or products. I have a sinking feeling.
 
Reading some topics in this section and in other forums, it seems to me the UVD priority bug is not limited exclusively for HD5000 series. Or, at least, this is not the only reason.

IMHO drivers behavior can still improve a lot, beginning from DCOM issue (apparently involved with 2d/3d clockspeed).
 
weah said:
Reading some topics in this section and in other forums, it seems to me the UVD priority bug is not limited exclusively for HD5000 series. Or, at least, this is not the only reason.
Click to expand...

I only have 100% confirmation on UVD behavior on the HD 5000 series, but any card that doesn't say "UVD 3" next to it in this list may be subject to the same issues: UVD enabled GPUs

weah said:
IMHO drivers behavior can still improve a lot, beginning from DCOM issue (apparently involved with 2d/3d clockspeed).
Click to expand...
Agreed, but I can't seem to reproduce the DCOM error. I've got it listed, but I really need some more information on it... By the way, does it cause any problem besides adding an entry to the event log on shutdown?
 
Sent off another email to AMD this weekend. Hopefully this one FULLY clarifies the idle clockspeed issue, so they can pin it down and get a fix going. After I get confirmation back that they've understood what's wrong with AMD Overdrive, I'll deploy an email dealing with the DCOM errors. Working through one issue at a time seems to work better, here.

Here's the bulk of the email, visual aid and all. I've offered to walk them through it over the phone if necessary:
Alright, I've gone ahead and put together a visual aid for the idle clockspeed issue experienced when AMD Overdrive is enabled. I've attached the image to this email, and if that doesn't go through, I've also uploaded the image here: http://i.imgur.com/wEiP5.png

The first three boxes (figure 1, 2, and 3) are representative of single-display behavior, these are all working correctly, I've only included them for reference and for the sake of completeness. The last three boxes (figure 4, 5, and 6) represent the behavior AMD Overdrive exhibits when using multiple displays in an extended desktop configuration (not an eyefinity display group).

Notice that normal multi-monitor operation (no overclock) in Figure 4 and 5 has the card idle at 400MHz core, 1000 MHz memory. The multi-monitor PowerPlay table on an HD 5850 is as follows:
Idle: 400 / 1000
2D: 550 / 1000
3D: 725 / 1000

However, as soon as an actual overclock is applied (see Figure 6), the driver reverts to the single-monitor PowerPlay table. The immediate effect of this glitch is that the idle speed drops from 400 / 1000 to 157 / 300. Looking closer, the entire PowerPlay table has reverted to the one designed for single-display configurations, which is as follows.
Idle: 157 / 300
2D: 550 / 900
3D: 725 / 1000

Notice that the memory clockspeed is constant on the multi-display table, but variable on the single-display table. Memory clockspeed is locked down for multiple monitors because shifting memory clockspeed causes secondary display heads to experience a "horizontal line flash" type of momentary flickering.

What this basically means is that users of multiple monitors cannot overclock their cards with AMD overdrive, because doing so unlocks memory clockspeed, which causes constant "horisontal line flashes" as memclock goes up and down based on load. The driver needs to be fixed so that it continues to use proper multi-display clockspeeds when AMD Overdrive is in use.


This appears to be fixable at the driver level, and in fact, I've already developed a temporary fix that involves nothing more than editing an AMD Vision Engine Control Center preset. I see no reason why the driver couldn't fix this itself, automatically. The manual fix procedure is as follows:
1. Open AMD Vision Engine Control Center
2. Click "Performance" and then "AMD Overdrive"
3. Set an overclock. For this example, we'll try something that should be safe on any HD 5850: 740MHz core, 1000MHz Memory.
4. Apply the overclock, your secondary displays will probably start flickering, ignore it for now.
5. Click "Presets" and then "Add Preset"
6. Name the new preset "Overdrive Fix" and save it.
7. Navigate to "C:\Users\%username%\AppData\Local\ATI\ACE\Profiles"
8. Open "Overdrive Fix.xml" in Notepad
9. Change this block:
Code: 
        <Feature name="CoreClockTarget_0">
          <Property name="Want_0" value="15700" />
          <Property name="Want_1" value="55000" />
          <Property name="Want_2" value="74000" />
        </Feature>
        <Feature name="PowerControl_0">
          <Property name="Want" value="0" />
        </Feature>
        <Feature name="MemoryClockTarget_0">
          <Property name="Want_0" value="30000" />
          <Property name="Want_1" value="90000" />
          <Property name="Want_2" value="102500" />
        </Feature>
        <Feature name="CoreVoltageTarget_0">
          <Property name="Want_0" value="950" />
          <Property name="Want_1" value="1038" />
          <Property name="Want_2" value="1088" />
        </Feature>
So that it looks like this:
Code: 
        <Feature name="CoreClockTarget_0">
          <Property name="Want_0" value="40000" />
          <Property name="Want_1" value="55000" />
          <Property name="Want_2" value="74000" />
        </Feature>
        <Feature name="PowerControl_0">
          <Property name="Want" value="0" />
        </Feature>
        <Feature name="MemoryClockTarget_0">
          <Property name="Want_0" value="100000" />
          <Property name="Want_1" value="100000" />
          <Property name="Want_2" value="100000" />
        </Feature>
        <Feature name="CoreVoltageTarget_0">
          <Property name="Want_0" value="1000" />
          <Property name="Want_1" value="1038" />
          <Property name="Want_2" value="1088" />
        </Feature>
10. Save the file over itself.
11. Go back to AMD Vision Engine Control Center, click the "Overdrive Fix" preset.
12. The card will begin to idle at the correct speed 400 / 1000. It will also scale properly to 550 / 1000 and up to the requested overclock of 740 / 1000 in 3D mode. Any display flickering should stop immediately.
Click to expand...

Edit: Aug, AMD's email system dropped some of the message, I just had to resend it...
 
Last edited:
Well...they don't want to fix AMD Overdrive either...
AMD said:
This is expected behavior of our driver, or you also can say it&#8217;s a limitation with this board.
Click to expand...
The thing is, I know it's not a limitation of the hardware. I sent them a fix that involves nothing more than editing a profile generated by Catalyst Control Center.

AMD really seems to have something against those using HD 5000 series cards. I really don't like how they're treating the entire group of people who were first in line to adopt Eyefinity... I really think they need to recall the entire HD 5000 series, if the cards themselves have so many of these "limitations" (read: bugs) that they can't resolve.

Edit: Sent off another email, with suggestions for alternate behavior that would be preferable to the current broken behavior. Lets see what they say...
 
Last edited:
I wonder if a petition needs to be started by all 5000 series adopters to gather signatures for AMD to fix the bugs or they will boycott their products forever. Imagine a CEO's or high level Corporate managers trying to talk to the driver creators about the premise of losing tens of thousands of loyal customers because they have such a god complex. It's a pain in the ass to fix it, we get it!! Just fucking do it!! If they refuse or whatever fire them!

I mean, it's as if the driver creators can do whatever the hell they want and the managers have no call. Maybe the driver teams boss needs to be fired and the executive above him so that some true heavy handed prioritizing can occur at AMD.


Anyway, /End Rant, I don't even have a 5000 series card and their attitude annoys me greatly.
 
If you think their attitude is bad, just wait until you see how that tool mod/admin "spyre" treats AMD paying customers on AMD's official forums.
 
Raudulfr said:
If you think their attitude is bad, just wait until you see how that tool mod/admin "spyre" treats AMD paying customers on AMD's official forums.
Click to expand...

Wow really! I have never posted in their forum. A shame AMD lets people like this work for them when there is so much unemployment going around and a huge talent pool of good people to choose from. They make good hardware though, shame they can't put the final touches on the software to really make the customers happy and to do justice to the amazing engineering involved.
 
Had another conference call with AMD a couple hours ago... you guys aren't going to like this.

I have now been told that no resources are being devoted to fixing major issues with HD 5000 series graphics cards, and that I should not expect that situation to change. In their own words, going back and fixing issues with the Evergreen family of GPU's would "throw a wrench into the works" of their current development cycle. They also made it clear that the feedback that we are giving them isn't going to waste, and will be considered during the development of upcoming graphics hardware (which probably means the HD 7000 series).

I wonder if they're aware of the message this stance sends to their customers... It basically amounts to: "thanks for beta-testing our hardware, now plunk down some more cash to get the fixed version." Not cool...

Edit: In other news, they are interested in fixing the DCOM error. I'm compiling an email with all the relevant information now.
 
Last edited:
Unknown-One, thanks for all of your hard work here.

I haven't read the entire thread, but I can assure you I have experienced some of the same problems you are describing, and my solution was finally to just modify the BIOS. I went through XFX and received a response similar to what AMD is providing you "That's how it is supposed to work."

Yeah... Shit graphics does not a $750 video card make.

Needless to say, I will be weighing the next generation very heavily against their direct competition.
 
Unknown-One said:
wonder if they're aware of the message this stance sends to their customers... It basically amounts to: "thanks for beta-testing our hardware, now plunk down some more cash to get the fixed version." Not cool...

Edit: In other news, they are interested in fixing the DCOM error. I'm compiling an email with all the relevant information now.
Click to expand...

Makes sense, they are expected to deliver new updated products with limited development resources. Every tech company I worked for does the same thing. I'm not defending, but it is the reality.

Back in my main frame and network days, we always felt like Beta testers for IBM and Cisco. Nothing has changed.....
 
@Unknown-one: First of all, I would like to thank you for reporting Dcom issue. Unfortunately there are other issues, and I compiled a schematic list (including some suggestions for CIM/CCC usability).


- Bug reports to address


1) I discovered a new bug report concerning CCC Profiles, started from 11.6 CCC.
Source: http://www.rage3d.com/board/showthread.php?t=33979025

Praetor2 said:
I reported this issue to AMD as follows:

PROBLEM DESCRIPTION: Every time the ATI commandline tool "CLI.EXE" is used to apply a Catalyst Control Centre PRESET (profile), it leaves a resident copy of "MMLoadDrv.exe" (visible in Windows Task Manager).

So every time I use CLI.EXE to apply a PRESET (via a VBS script) before a game is started, it leaves ANOTHER copy of "MMLoadDrv.exe" resident. At the end of the day I have MANY MANY of these "MMLoadDrv.exe" files resident!

I've been using CLI.EXE for years, and this problem only started with Catalyst v11.6 and now also in v11.7.

Example of commandline tool usage:
C:\Program Files (x86)\ATI Technologies\ATI.ACE\Core-Static\CLI.exe Start Load profilename=MyProfile

Please rectify this issue as I still have to use v11.5 in the meantime.
Click to expand...

2) "CCC Profiles are Not Properly Applied" still not fixed in 11.8:
http://hardforum.com/showpost.php?p=1037652555&postcount=31
In this post you can find another workaround to add in the OP of this thread.

3) There is a bug in 11.8 regarding OpenGL32.dll module throws an exception when wglCreateContext is called
This bug was discovered from AIDA64 developers. No other drivers seem to be affected, except 11.8 (that's odd)
Full details:
http://forums.aida64.com/index.php?/topic/497-open-gl-not-listed/
New update from AIDA64 developers:
We've switched from dynamic DLL linking (of OPENGL32.DLL) to static linking. Apparently Catalyst 11.8 has a bug that only comes up when OPENGL32.DLL is dynamically loaded.

- Suggestions to address


1) It would be great an option during uninstall process, with the possibility to remove all custom settings/profiles.
Source: http://www.rage3d.com/board/showthread.php?t=33980557
Uninstall process needs to be improved in this sense, because it doesn't perform a clean uninstall.
A dedicated option is to be intended as suggestion for AMD.

2) Actually the infamous ULPS feature can be disabled or enabled only with a registry setting/third party tools:

Go to registry and disable Ulps:
[HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Control\Class\{4D36E968-E325-11CE-BFC1-08002BE10318}]
and change "EnableUlps" to 0 under the 0000 and 0001 or 0003 keys.
Click to expand...

It would be more useful a dedicated option in CCC, instead to go expressly in registry to enable/disable it.
It's amazing that a catalyst feature is not available in CCC as option.
 
Last edited:
I sent the following email off to AMD on Sunday, still waiting on a response back:
-- Issue #1: DistributedCOM error during shutdown --
A DistributedCOM error is logged in the "System" event log (Located under "Windows Logs" in the Event Viewer) every time the computer is shut down or restarted. The Event ID is 10010, and the error logged is as follows:

"The server {E10F6C3A-F1AE-4ADC-AA9D-2FE65525666E} did not register with DCOM within the required timeout."

Catalyst 11.4 Preview, March 7th [8.84.3-110226a-114256E], was the last released driver that DID NOT experience this problem.
Catalyst 11.4 Preview, March 23rd [8.84.2-110321a-115519E], was the first released driver that began experiencing this problem. Every single released driver after this is also effected.

A couple of work-arounds have been discovered that seem to fix the issue. The simplest one is to set the "AMD External Events Utility" service to "disabled," which seems to prevent the error from being logged, but also disables profile hot keys.

The second work-around is more involved.
1. Extract atiesrxx.exe and atieclxx.exe from Catalyst 11.4 Preview, March 7th [8.84.3-110226a-114256E]
2. Stop the "AMD External Events Utility" Service
3. Paste the extracted executables into "C:\Windows\System32\"
4. Allow Windows to overwrite the exiting files.
5. Start the "AMD External Events Utility" service again.
6. Reboot, you will no-longer see a DCOM error logged, and no driver functionality is disabled.



-- Issue #2: Multiple instances of MMLoadDrv.exe --
Every time CLI.exe is called to apply a preset, it spawns another instance of MMLoadDrv.exe, and never closes it. As you can imagine, after applying a few AMD VISION Engine Control Center presets (via the start menu shortcuts created by the Control Center), this leaves quite a few extra processes running in the background that have to be killed manually from the Task Manager.

Catalyst 11.6 WHQL was the first released driver to experience this problem.
Catalyst 11.7 WHQL and Catalyst 11.8 WHQL continue to experience this problem.



-- Issue #3: ACE Event Log Errors --
A near-constant stream of exceptions is written to "ACEEventLog" (Located under "Applications and Services Logs" in the Event Viewer). These errors range from "Access Denied" due to insufficient privileges, to "File Not Found" because the driver is looking in the wrong place.

I have attached an excerpt of my log-file to this email as "ACEEventLog 8-22-2011.txt" for your review.

This issue goes back at least as far as Catalyst 11.5 (I haven't checked for this issue an any driver older than that), and persists through 11.6, 11.7, and now 11.8.



-- Issue #4: wglCreateContext Exception --
An exception is thrown when an wglCreateContext is called in OpenGL32.dll

Catalyst 11.8 WHQL is the first released driver to experience this problem.



-- Suggestion #1: AMD Video Driver "Clean Install" --
AMD's video driver uninstaller leaves behind files and settings, I assume this is so that settings are carried over through driver revisions. Carrying settings over from one Catalyst version to another can sometimes cause issues that cannot be resolved without uninstalling the driver, running a 3rd party utility such as Driver Sweeper, and finally installing the latest video driver.

I suggest a "clean install" option be added to the driver installer. When checked, this option should clear all remnants of previous drivers, then install the driver contained within the installation package.

Nvidia has recently implemented this functionality in their driver installer, and I can almost guarantee you it has reduced the number of reports they receive on driver-upgrade issues. It also offers a quick troubleshooting option that immediately creates a clean slate and common ground. Very valuable.



-- Suggestion #2: User-accessible toggle for ULPS (Ultra Low Power State) --
ULPS allows secondary cards in a crossfire configuration to slip into an "ultra-low" power state, drawing less power than a normal idle state when their performance is not needed. This technology can cause issues with temperature-monitoring and overclocking software when they attempt to access the secondary card while it is in ULPS mode.

As an example, GPU-Z and MSI Afterburner both crash if they are launched on a system that has Crossfire and ULPS enabled.

Currently, the only way to disable ULPS is to navigate to the following 4 keys (all that exist out of the 4), And change every instance of the DWORD Value "EnableUlps" from "1" to "0":
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E968-E325-11CE-BFC1-08002BE10318}\0000]
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E968-E325-11CE-BFC1-08002BE10318}\0001]
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E968-E325-11CE-BFC1-08002BE10318}\0002]
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E968-E325-11CE-BFC1-08002BE10318}\0003]

A setting that needs to be changed in order to use simple monitoring software like GPU-Z should not be buried in the registry. My suggestion is that, When crossfire is enabled, a check-box should appear in the "AMD Overdrive" section of AMD VISION Engine Control Center that allows the user to toggle ULPS on or off.
Click to expand...

@Weah: I think we have some confusion here...
weah said:
2) "CCC Profiles are Not Properly Applied" still not fixed in 11.8:
http://hardforum.com/showpost.php?p=...5&postcount=31
In this post you can find another workaround to add in the OP of this thread.
Click to expand...
The bug you're referring to can still be provoked into appearing, but the post you're linking to is about another issue entirely.

The bug in question deals with not all aspects of a previously saved profile being applied on the first try.
That post you've linked is complaining that, when saving a profile, only the contents of the currently visible panel are saved in the profile (rather than ALL settings in CCC).

If you want to make a profile that saves all settings in CCC rather than just those from a specific section, switch to basic mode and make the profile there. I'll suggest making the profile system more flexible, but this isn't really a bug.

Thanks for your help with the research, by the way. You're helping keep AMD and I very busy. I'll add these new bugs to the original post in a minute.
 
Last edited:
You must log in or register to reply here.
Back
Top