Hackmodford Posted January 27, 2014 Share Posted January 27, 2014 It would be cool if there was a command we could use to restart the OpenCL framework... I'm also curious about the CPU usage. This is interesting... After opening preview my GPU clock gets locked at 1.08 GHZ I then ran Luxmark and got this during the benchmark. It actually ramped up to the Turbo state. This is the benchmark score I got. After the benchmark it went back down to 1.08 GHZ Here is my lowest clock state. I get the same luxmark score when starting from a "non-stuck" state as I do from a "stuck" state. Also the clock remains stuck at 1.08GHz after running luxmark. So I guess more confirmation it is OpenCL In any case I guess my GPU is not stuck at its highest state because it can go a little higher... Link to comment Share on other sites More sharing options...
Gringo Vermelho Posted January 27, 2014 Share Posted January 27, 2014 But if you killed it then you'd break Preview and whatever else depends on OpenCL. Even if there was a command that could "restart" OpenCL, I'm pretty sure that the GPU clocks would still be stuck. Link to comment Share on other sites More sharing options...
Hackmodford Posted January 27, 2014 Share Posted January 27, 2014 Yes I meant restart (changes post) Link to comment Share on other sites More sharing options...
Gringo Vermelho Posted January 27, 2014 Share Posted January 27, 2014 Mine (660 GTX) doesn't get stuck at its highest state either, it stays at the second highest I think (GPUSensors FakeSMC plugin says 992Mhz). Highest is 1.12GHz, so it does ramp down, just not all the way. The clocks ramping up are clearly a consequence of OpenCL doing calculations on the GPU, but I still don't believe that OpenCL itself is to blame for the clocks not dropping again afterwards. /EDIT yay 1 Link to comment Share on other sites More sharing options...
yokonunz Posted January 27, 2014 Share Posted January 27, 2014 what's your CPU load like when doing the same things that caused the GPU clocks to get stuck? My CPU (Intel G630) is correctly scaled from 1.58GHz to 2.67GHz since i patched and I've never rebooted by system. Link to comment Share on other sites More sharing options...
Gringo Vermelho Posted January 27, 2014 Share Posted January 27, 2014 Thanks, but you didn't really answer my question Link to comment Share on other sites More sharing options...
Hackmodford Posted January 27, 2014 Share Posted January 27, 2014 Has anyone filed a bug report with Apple about this? Link to comment Share on other sites More sharing options...
nomaj Posted January 27, 2014 Share Posted January 27, 2014 It would be cool if there was a command we could use to restart the OpenCL framework... I'm also curious about the CPU usage. This is interesting... After opening preview my GPU clock gets locked at 1.08 GHZ I then ran Luxmark and got this during the benchmark. Screen Shot 2014-01-26 at 7.36.32 PM.png It actually ramped up to the Turbo state. This is the benchmark score I got. Screen Shot 2014-01-26 at 7.38.17 PM.png After the benchmark it went back down to 1.08 GHZ Screen Shot 2014-01-26 at 7.38.23 PM.png Here is my lowest clock state. Screen Shot 2014-01-26 at 7.43.39 PM.png I get the same luxmark score when starting from a "non-stuck" state as I do from a "stuck" state. Also the clock remains stuck at 1.08GHz after running luxmark. So I guess more confirmation it is OpenCL In any case I guess my GPU is not stuck at its highest state because it can go a little higher... With this patch LuxMark wont start.. because patch kills OpenCL ( read the first page of patch ). What i find that for example Photoshop uses OpenCL - after patch PShop don't find OpenCL support at all - works with CPU. This is not perfect solution... what i see nice about it - is that your gfx card wont be stuck at high frequency, less heat and power consumption. For people who don't care about mentioned they just need to stop using quick-preview and instead Preview apk to use something like Xee or similar. Hope that Apple or Nvidia will fix this annoying thing. I'm not sure that this bug can be found on real Mac-s. Link to comment Share on other sites More sharing options...
digital_dreamer Posted January 29, 2014 Share Posted January 29, 2014 This is awesome. Patch works great on my GTX 680. Thanks, guys. MAJ Link to comment Share on other sites More sharing options...
Riley Freeman Posted January 29, 2014 Share Posted January 29, 2014 I just patched the framework in 10.8.5 and can confirm that neither LuxMark nor Oceanwave will run any benchmarks now that the GPU support is disabled. It's a little easier on ML as there's no need to codesign the binary. I just used two terminal commands (one to backup and one to patch) sudo cp /System/Library/Frameworks/OpenCL.framework/Versions/A/OpenCL /System/Library/Frameworks/OpenCL.framework/Versions/A/OpenCL.bak sudo perl -pi -e 's|\xE8\x52\xBE\x01\x00\xF6\xC3\x05\x74\x13|\xE8\x52\xBE\x01\x00\xF6\xC3\x00\x74\x13|' /System/Library/Frameworks/OpenCL.framework/Versions/A/OpenCL 1 Link to comment Share on other sites More sharing options...
maroder Posted January 29, 2014 Share Posted January 29, 2014 00F6C3057413 for 10.9.2 change to 00F6C3007413 Just tried it on 10.9.1 and it works great - no more stuck clock. Thank you! I do not understand your hint regarding 10.9.2 - it's identical? Link to comment Share on other sites More sharing options...
Gringo Vermelho Posted January 29, 2014 Share Posted January 29, 2014 No it isn't, look again! Link to comment Share on other sites More sharing options...
maroder Posted January 29, 2014 Share Posted January 29, 2014 No it isn't, look again! If I look at both threads this is what I get with "Copy & Paste": 10.9.1: 00F6C3057413 to 00F6C3007413 10.9.2: 00F6C3057413 to 00F6C3007413 It's identical - but maybe I'm just blind ... Link to comment Share on other sites More sharing options...
Hackmodford Posted January 29, 2014 Share Posted January 29, 2014 Isn't this the difference they're referring to? 00F6C3057413 00F6C3007413 Update: Re read stuff... now I'm confused Link to comment Share on other sites More sharing options...
maroder Posted January 29, 2014 Share Posted January 29, 2014 Isn't this the difference they're referring to? 00F6C3057413 00F6C3007413 Yes sure but where is the difference in patching 10.9.2 in regard to 10.9.1? That was my question Link to comment Share on other sites More sharing options...
Hackmodford Posted January 29, 2014 Share Posted January 29, 2014 Re read stuff... now I'm confused 1 Link to comment Share on other sites More sharing options...
Gringo Vermelho Posted January 29, 2014 Share Posted January 29, 2014 I didn't even read it and now I'm confused too. Link to comment Share on other sites More sharing options...
Hackmodford Posted January 29, 2014 Share Posted January 29, 2014 I have a bad feeling about this... 1 Link to comment Share on other sites More sharing options...
Riley Freeman Posted January 29, 2014 Share Posted January 29, 2014 The original search string for 10.9 and 10.9.1 was E83DE10100F6C3057413. I couldn't find this in 10.9.2 so nomaj said to search for 00F6C3057413 which is only half of the original string. The second half of the string is the same for 10.8.5 and the various 10.9 builds. Only the first half varies. 1 Link to comment Share on other sites More sharing options...
maroder Posted January 29, 2014 Share Posted January 29, 2014 The original search string for 10.9 and 10.9.1 was E83DE10100F6C3057413 I couldn't find this in 10.9.2 so nomaj said to search for 00F6C3057413 which is only half of the original string. LOL Now I get it. Thank you for clearing things up Link to comment Share on other sites More sharing options...
david_osx Posted January 29, 2014 Share Posted January 29, 2014 What is the exact string for 10.9.2 I cant get this to work at all :/ Ok maybe I search for 00F6C3057413 and replace it with 00F6C3007413 then save the file or sign it first then save it what is the exact steps for this patch.. Im having hard time to make it work.. either from not being able to repair permissions or booting back with missing-bluetooth-controller problem with my mavericks cleaned install… btw i used myhack to make the usb but according to people i have to restart as soon my hack promote me to choose what extra folder kext i want to install or the bug my fix have changes IRQs of the installation etc.. maybe thats why I'm getting missing-bluetooth-controller problem.. maybe i need to complete the install.. make the usb install with another method.. anyway what is the exact steps for making this patch work with 10.9.2 thanks guys.. Link to comment Share on other sites More sharing options...
Gringo Vermelho Posted January 29, 2014 Share Posted January 29, 2014 the missing bluetooth messages are harmless, I've had those for a long time. I don't have any bluetooth hardware, I think it's triggered by the model identifier. Link to comment Share on other sites More sharing options...
david_osx Posted January 29, 2014 Share Posted January 29, 2014 sure its harmless but the problem is i cant boot to the desktop and the boot process hangs at that message.. that problem happens when i save the edited opencl file and boot back so the permission is change for opencl file and i have to repair the permission using the install usb/ Disk utility.. anyway what is the exact hex string for 10.9.2 ?? 2. after i enter the string and replace it with what exactly and then save the file or not ?? 3. should i signe the file and restart right away or repair permission before restarting..?? Dont know what exactly im doing wrong and everyone has already made the patch and done by now maybe i first of all need a new clean install of mavericks with another method then myhack.. I cant get this patch to work what ever step I do or change.. Link to comment Share on other sites More sharing options...
Hackmodford Posted January 29, 2014 Share Posted January 29, 2014 This whole disabling opencl thing just sits wrong with me. I can't see this being a good idea or a worthy compromise. 2 Link to comment Share on other sites More sharing options...
Gringo Vermelho Posted January 29, 2014 Share Posted January 29, 2014 Totally agree. I thought OpenCL would just run on the CPU instead of the GPU and everything would be fine but if it breaks things then I'm not patching. So far we have only one person saying that though. sure its harmless but the problem is i cant boot to the desktop and the boot process hangs at that message.. What I meant to say was, it might be the last message you see, but it's harmless and has nothing to do with why your hack hangs at that point. Link to comment Share on other sites More sharing options...
Recommended Posts