Message boards : Questions and problems : problems with ATI 10.3
Message board moderation
Previous · 1 · 2 · 3 · Next
Author | Message |
---|---|
Send message Joined: 2 Jun 10 Posts: 3 ![]() |
Same with BOINC 6.10.58 with ATI 4850 Radeon on Win7 x64. Had BOINC running fine with CC 9.7 drivers. When switched to 10.4 or even 10.5, boinc.exe started throwing Access Violation Exceptions. Reverting back to 9.7 drivers didn't had any effect. So I dl all drivers from version 9.7 to 10.5 and installed each, every time on a fresh install of Win7 x64 and tested boinc. Exception started to appear when installed drivers with ATI's version of Visual C++ 2008 Redistributable - x64 (9.0.30729.4148) which is not the same version as the one you can dl from Microsoft's site. ATI started distributing this version with CC 9.12 drivers and later. Hope this info helps the devs to figure out what's wrong. Jimi |
![]() Send message Joined: 29 Aug 05 Posts: 15628 ![]() |
Same with BOINC 6.10.58 That's still a future version, at this time not released for testing even. It's still got to be build. I think you mean 6.10.56. ;-) |
Send message Joined: 2 Jun 10 Posts: 3 ![]() |
Same with BOINC 6.10.58 You are correct. It was a typo. 6.10.56 is the version I used. |
Send message Joined: 2 Jun 10 Posts: 3 ![]() |
I would like to state some further corrections to my posts. The version of Visual C++ 2008 is actually the SP1 version. I re-done the testings and till ATI CC 10.2 drivers boinc works fine. Exceptions start with version 10.3 and later. Sorry if I caused any inconvinience. Jimi |
Send message Joined: 15 Apr 10 Posts: 6 ![]() |
Has anybody tried the latest drivers? Or are we all still stuck on CCC 10.2? |
![]() Send message Joined: 29 Aug 05 Posts: 15628 ![]() |
I have no problems with 10.6, but haven't tried 10.7 yet. |
Send message Joined: 23 Apr 07 Posts: 1112 ![]() |
I've tried Cat 10.7, combined with Forceware 258.96, same as before, i can install both sets of drivers O.K, Boinc 6.10.58 at least starts now, but disables ATI GPU, offline runs of Hybrid ATI Astropulse Brook+ apps still crash, With SDK_2.2 installed, offline ATI OpenCL Astropulse apps crash too, Cuda apps always work no problem, reverted back to Cat 10.2 again, Brook+ apps work again, Claggy |
Send message Joined: 25 Nov 05 Posts: 1654 ![]() |
Problem reported here with 10.7, on cpdn graphics. |
![]() Send message Joined: 29 Aug 05 Posts: 15628 ![]() |
Oh, that effect. I thought that was normal. Seen it as well, and even reported it on the CPDN forums at the time. |
Send message Joined: 15 Apr 10 Posts: 6 ![]() |
Ageless - forum moderator, project administrator, You guys (associated with BOINC - I assume you are) seem quite reserved in your comments regarding the problems with the ATI drivers with versions >10.2. Claggy seems to saying that the ATI GPU is being disabled with versions 10.7 now but the BOIC subsystem seems to run otherwise. What's up? Is the team having problems determining the issue/fix and just disabled ATI GPU use? Maybe what I read between the lines is wrong - if so, please correct my understanding. If the team is having problems, bring the issue to the general community. You would likely find a system/driver programmer that would be willing to help out. |
![]() Send message Joined: 29 Aug 05 Posts: 15628 ![]() |
associated with BOINC - I assume you are I'm a volunteer just like you are. I may have close contacts with the developers, but I am not one of them. What's up? Is the team having problems determining the issue/fix and just disabled ATI GPU use? It's not a BOINC problem in that sense. BOINC's method of detecting the GPU hasn't changed between GPU driver versions, as is shown by reverting back to an earlier driver and things will work again. Which make it a driver problem, making it ATI's problem. I suspect very much that part of the problem is that ATI no longer invests any time, money or manpower into their CAL/Brook+/Stream computing software. They've turned all of their attention to developing OpenCL compatibility/compliancy instead (Catalysts 10.4 and up). The BOINC developers are in contact with the ATI driver development team, but from what I understand, getting information from a rock is easier. ATI isn't very much interested in supporting BOINC or even interested in telling people how to use their software/code. Perhaps that will change in some future. PS: The BOINC developers hardly read these forums. Yes, I have pointed out this thread, several times, but even then... If you want to get directly into their inbox and possibly get an answer back, it is better to post on the BOINC Development and Alpha email lists. |
![]() Send message Joined: 8 Jan 06 Posts: 448 ![]() |
Ageless' Project administrator title is more honorary than actual. He's a volunteer like us and isn't directly involved in Boinc development. It is to allow him to deal with forum issues which the moderator level title wouldn't and the admins don't have time for. He does have a more direct line to the developers to bring issues to their attention. Edit: I see Ageless answered for himself. Boinc V 7.4.36 Win7 i5 3.33G 4GB NVidia 470 |
Send message Joined: 15 Apr 10 Posts: 6 ![]() |
Thanks for the fast response, guys. The background you have provided paints in another piece of the big picture - something we (BOINC supporters, ATI loyal customers) cannot see without your help. Sounds like we're caught in a transitional period while ATI and the GPU industry moves to a set of common standards (OpenCL). If so, so be it. We'll come out better in the long run. I've always been disappointed that there were not more BOINC apps running on the ATI GPUs. NVidia had them out of the gate on an API and promoting GPU use. But these two companies have been leap frogging each other for years. That's the way it should be working. ;-) I, for one, will stay with CC 10.2 as I have no issues with the drivers at this time. And, I can run what few ATI GPU apps are available. |
Send message Joined: 23 Apr 07 Posts: 1112 ![]() |
I have a fix, Rename aticfx32.dll to aticfx32_temp.dll and aticfx64.dll to aticfx64_temp.dll in System32 and SysWOW64 folders, See: Regression: CAL samples not working with a Nvidia GPU on Windows 7 I've now successfully run Astropulse Hybrid ATI Brook+ apps, 2 different Astropulse OpenCL apps, HelloCL and CLinfo apps, all with my Nvidia GPU still fitted and latest drivers installed, and Boinc now detects the HD5770 with Cat 10.7 without disabling it, GPU-Z also no longer Crashes when SDK installed, here's my host at Collatz (not done any tasks with Cat 10.7 yet): hostid=20542 Claggy |
![]() Send message Joined: 5 Sep 10 Posts: 2 ![]() |
Thought I'd share my experiences with getting my 9800GT and HD4850 cards to 'play nice' together with latest drivers without resorting to any dll-renaming tricks! After sweating blood for a few hours trying various slot/monitor combinations - along with many un- & re-installs - I came to the conclusion that the A(T-Rex)i (or its drivers) were behaving in typical Dino-fashion - wanting to devour everything in sight! - so I adopted the following 'stealth' strategy.. 1. Remove NVidia card, do complete uninstall of NVidia drivers, and remove all traces of anything smelling of NVidia from the drive. 2. Install A(T-Rex)i card in primary 'Master!' slot & attach to primary 'Master!' monitor - Install Cat v10.7 drivers and get all working with Boinc. 3. Turn off and quietly slip NVidia card into secondary slot while the monster sleeps.. 4. Turn on and wait until A(T-Rex)i's attention is fully diverted while crunching on a few tasty DnetC tasks before (quietly!) installing NVidia (GeForce 258.96) drivers in the background... 5. A further reboot sees the two cards happily crunching side-by-side as if they'd been bosom-buddies all their lives!... Now I DO understand that all this anthropomorphism of these purely silicon-based (non-)life forms was merely the result of a mild delusion brought on by a severe lack-of-sleep while trying to get them to work together - but work together they did - so who am I to argue with that!.. :-) |
![]() Send message Joined: 22 May 10 Posts: 9 ![]() |
|
Send message Joined: 23 Apr 07 Posts: 1112 ![]() |
is there a way to notify developers of ati driver or BOINC developers to solve this problem? Already done, see my thread on the ATI Stream Forum, Claggy |
![]() Send message Joined: 22 May 10 Posts: 9 ![]() |
Already done, see my thread on the ATI Stream Forum, Yes, I have seen. The problem will be fixed in next release? Or do you know when? Because I Should play mafia 2 and I am locked to 10.2 |
Send message Joined: 23 Apr 07 Posts: 1112 ![]() |
Already done, see my thread on the ATI Stream Forum, Unlikely to be next release, AMD/ATI has a release cycle, they'll have to a fix it first, test it, test it some more, maybe tweak it, test it, etc, so it'll be a couple of releases before it appears, Claggy |
![]() Send message Joined: 22 May 10 Posts: 9 ![]() |
something new, or are we still waiting? |
Copyright © 2025 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.