Jump to content


  • Content Count

  • Joined

  • Last visited

About Mull7965

  • Rank
    InsanelyMac Protégé

Profile Information

  • Gender
  • Location
  1. Mull7965

    Problem with disable PCI-e Lane 2

    When you have successfully disabled the NVIDIA card what does the Graphics/Displays section in the System Information app look like? (You know, the one that opens when you click on the System Report button in the About this Mac panel?) I'm trying to do the same thing as you, very similar setup, but I don't know whether I have successfully disabled the card. For me it looks like this:
  2. I'm using a Gigabyte Z77X-UP5 TH with the latest BIOS (F12). Last weekend, I booted into the existing Windows 10 installation (current version, 64bit) and when I restarted the computer Windows did some updates. On the actual restart the computer booted directly into Windows 10, not the Clover boot menu as before. No problem, I thought, I'll just boot from a USB stick I always have handy for such cases, and fix the bootloader on the EFI partition. To my surprise, the BIOS suddenly didn't show the USB drives anymore. It only showed the harddrives. I'm sure the USB drive works and is bootable. I confirmed this by switching to an alternative BIOS (the mainboard has two BIOSes that can be selected with a physical switch). But with that latest version of the BIOS, after running Windows, it is no longer possible to choose a USB drive. Secure Boot is turned off by the way. Has anyone seen something similar? Can Windows 10 really modify the BIOS (any BIOS?) in such a way that valid boot options are not shown anymore?
  3. Apologies if this is going off-topic. I installed the 1.0.0d2 version of the driver and ran it for a while. Didn't notice anything odd. However, when I looked at the output from netstat I realised that this driver doesn't seem to do checksums in software. Output from "netstat -s" after using the system for a while with the hnak driver: ip: 238317 total packets received 0 bad header checksums 219647 headers (4393268 bytes) checksummed in software ... 212038 packets sent from this host ... 195009 headers (3900208 bytes) checksummed in software Output after using the system for a while with the mausi driver (similar network usage profile): ip: 411934 total packets received 0 bad header checksums 0 headers (0 bytes) checksummed in software ... 382249 packets sent from this host ... 75 headers (1508 bytes) checksummed in software Now, I do admit that I don't understand the details. My assumption is that headers that are not checksummed in software are checksummed in hardware in the NIC, right? Is it a good thing that the Mausi driver does almost no checksumming in software?
  4. I had a similar problem in the Mountain Lion days. My board has an Intel ethernet controller and I was using whatever version of hnak's driver was current at the time. In my quest to eliminate possible causes I went back from a slightly overclocked BCLK and I added an SSDT to describe my CPU overclock correctly. That stoped the crashes for me. Any chance you have overclocked your system? Another idea: if your board has Thunderbolt, you could buy a Thunderbolt-Ethernet dongle from Apple to try to figure out whether the problem is with the OS or the driver. Then again, for the price you can probably buy a supported ethernet PCI card...
  5. Mull7965

    Graphics - No GPU Power Management

    The actual bug seems to have been around since at least 10.8.5. It just gets triggered more easily in Mavericks. Most notably Preview and QuickLook can now trigger it.
  6. Mull7965

    Graphics - No GPU Power Management

    This is not how I understand the description of the patch. The author talks about three key function calls in the OpenCL framework: one to do basic initialisation, one to initialise CPU support, and a third call to initialise GPU support. Remember, OpenCL code can be run on either CPU or GPU. In some cases the CPU is faster, in other cases the GPU is. Either way, the patch simply stops the third call from happening. To me this implies that OS X will use the CPU for running all OpenCL code, because it doesn't think it has a GPU to run the code on.
  7. Mull7965

    Graphics - No GPU Power Management

    Can you please clarify the "it works" part? When you look at a picture in Preview or Quicklook, does your GPU clock down to below 1GHz afterwards? For most of us the GPU clocks down from the maximum boost frequency but it is stuck at just over 1GHz, rather than clocking down to the much lower regular idle frequency.
  8. Mull7965

    Graphics - No GPU Power Management

    After some tests I think I can confirm that it's not a problem with HWSensors. I removed all FakeSMC plugins and measured power usage at the wall. After booting, my system idles at about 59W, which is the same as it was with ML. After watching a video in VLC the system idles at 94W. This makes it pretty clear, I think, that the GPU isn't throttled back. Now, to me it's not surprising that editing Info.plist in the AGPM kext doesn't make a difference. With ML my system changed GPU frequency up and down even without AGPM edits, i.e. even when I got "[AGPM Controller] unknownPlatform" in the log. This seems to indicate that the change is somewhere deeper in OS X. But where?
  9. I have a Kepler card myself and I was hoping, too, that sooner or later Apple would have a Mac with a desktop Kepler card. Well, fat chance now... So, I started to experiement a bit and if I'm not mistaken I have good news about power management for Kepler. First, I tried to get power management working, patching the Info.plist in AppleGraphicsPowerManagement.kext. I got to the point where OS X recognised my card, and loaded the section I wanted, i.e. system.log had an entry like this on boot: kernel[0]: [AGPM Controller] build GPUDict by Vendor10deDevice1189 For the actual content I tried variations based on the GTX 680M and similar cards Apple has details on. Unfortunately, I never saw any entries in the logfile about OS X trying to switch power states. So, clearly something wasn't working. At the same time though, my system is very quiet and I'm pretty sure that if the GPU was running at full clock I would be able to hear the GPU fan over the rest. I didn't. Next, I thought that this is about power management and therefore changes would also be observable with a power meter on the wall. I plugged that in and measured a couple of things: idle power consumption, power consumption while playing 1080p in Youtube (Firefox), running the OpenGL test in Cinebench. Readings are averaged. I repeated this with AGPM not building a GPUDict, i.e. AGPM unknownPlatform, and, as a control, the same thing on Windows 7, which presumably has fully working power management. Results: idle youtube cinebench AGPM with GPUDict 59W ~64W ~130W AGPM unknown platform 59W ~64W ~130W Windows 7 59W ~64W ~130W I also checked whether the system returned to the measured idle power consumption after running Cinebench. It did. If I read this right then we don't have to do anything at all to get power management working for Desktop Kepler cards, right? Unfortunately, there is a fly in the ointment, though. The Cinebench result under OS X (in both configurations) is 48fps, while on Windows it's 62fps. Does that mean the card doesn't ramp up properly under OS X? But why the increased power consumption? Which is identical to the Windows one?
  10. Thanks for the change hnak. One question: did you commit the code? I'm looking at SourceForge (http://sourceforge.net/p/osx86drivers/code/45/log/) and the last commit is from late June. Am i missing something?
  11. Thanks Stickpin, this is a great starting point. A couple of comments, but please bear in mind that I am no expert in any of this. I got confused that the jmicron kext was still loading despite the fact that you have removed IDE from the DSDT. Maybe this is obvious to you but it took me a while to realise that you also have to disable the "Onboard IDE contoller" in the BIOS settings. Sleep only works when I check the "Start up automatically after power failiure" option in the energy saver preferences. (If I don't check that option my machine will turn off rather than sleeping.) Is this your experience, too? Or could this be a side effect of the CST settings I'm using? In your DSDT you define 3 C-States. Is this for C0, C3, and C5? Do you have C2/C2E and C4/C4E disabled in your BIOS settings? I hope this is not veering into generic speed stepping territory... Repoman, I like how you avoided duplication for _PSS and _PSD in your DSDT. However, this makes it even more obvious that you're using different C-States for the different cores. What is the reason for this? Also, in this block: Method (_CST, 0, NotSerialized) { Return (Package (0x04) { 0x03, Package (0x04) { ResourceTemplate () { Register (FFixedHW, 0x01, 0x02, 0x0000000000000000,,) Shouldn't the last line read: 0x0000000000000000,0x01,) And in the following blocks, you keep using FFixedHW when most other CST use SystemIO. I have no idea what that means but I noticed the difference. Is this intentional?
  12. Mull7965

    10.6.2 Released!

    Updated from Software Update without any problems. Vanilla install on Gigabyte EP45-UD3LR with NVIDIA 9800GT, configured through DSDT. More details in my installation report: http://www.mulle-kybernetik.com/artikel/hackintosh/
  13. Dual booting of Windows 7 and Snow Leopard from the same hard-drive is relatively easy to achieve. It took me a few hours of trial and error, though, to find the easy way. First, use Chameleon and a USB stick to boot the Snow Leopard installer. This is described in detail in the Lifehacker[1] guide for example. In the Snow Leopard installer partition your hard disk. Use a GUID partition table and create (at least) two partitions; the first one a FAT32 partition that will hold Windows 7 later, and the second one a Journaled HFS+ partition for Snow Leopard. After installing Snow Leopard make sure that you can boot it from the hard-drive without the USB stick. The procedure to achieve this is described in the Lifehacker guide mentioned before. On my system, for example, I used the following commands. Your disk and volume names may vary. The three boot files, boot0, boot1h, and boot came from the Chameleon distribution. sudo fdisk -f boot0 -u -y /dev/rdisk0 sudo dd if=boot1h of=/dev/rdisk0s3 sudo cp boot "/Volumes/Hackintosh HD" Next, make sure the USB stick isn't still plugged in and boot your system from the Windows 7 installer DVD. The installer should see the FAT32 partition created earlier and, after a conversion to NTFS, should allow you to install Windows 7 onto this partition. After the Windows install, when your system reboots it will boot into the Windows bootloader, which will only show the just installed Windows 7. Don't panic. Back in Windows 7 download EasyBCD[2] and add the Snow Leopard partition to the Windows boot loader. You must use a tool like EasyBCD because beginning with Vista the Windows boot loader no longer uses an easy to access text file. This should be it. More detail and background in my Hackintosh guide: http://www.mulle-kybernetik.com/artikel/hackintosh/ Hope this helps. [1] http://lifehacker.com/5351485/how-to-build...start-to-finish [2] http://neosmart.net/dl.php?id=1
  14. Just to confirm, it can be done this way and I didn't even encounter any errors along the way. Partition, install Snow Leopard, install Windows 7, setup boot loader with Easy BCD. Done. Details here: http://www.mulle-kybernetik.com/artikel/ha...sh/#dualbooting