Jump to content

SwithDrawn

Members
  • Content Count

    35
  • Joined

  • Last visited


Reputation Activity

  1. Like
    SwithDrawn got a reaction from Fergarth in iMacPro SMBIOS and new horizens   
    I switched over from MacPro6,1 to iMacPro1,1 on my 10.13.4 build and it is working great so far. 
     
    I got the black screen issue but removing NvidiaGraphicsFixup/Lilu solved it -- it's nice these are no longer needed! I didn't have to futz with PlatformSupport.plist either, it worked OOTB (possibly due to the beta build). Just had to remote in to get around the black screen and remove those kexts. 
     
    Also lost my USB2 ports. Updating USBInjectAll to the latest fixed that. 
     
    I'm not noticing any video lag (156 drivers) it actually SEEMS snappier as far as UI animation. Geekbench results were all a little lower than before, though. 
     
    Most importantly... unlock with Apple Watch works now!   Not sure if due to the new smbios itself, or something to do with my wifi/bt card which seems to work more consistently on this smbios. 
     

  2. Like
    SwithDrawn reacted to bbmatias3 in iMacPro SMBIOS and new horizens   
    Yes I’ll give it another go

    thanks!! 157 worked! 
  3. Like
    SwithDrawn got a reaction from cheneyveron in new possibilities for X79 AppleACPIPlatform panic   
    Thanks to everyone here for the shares -- finally got my old X9DR3-F upgraded from Sierra. I'm using x.di's config minus the included ssdt (got a kp related to P-States).
     
    I also had to use SunKi's fix for ICH10 in order to get passed the "still waiting for root device" with garbled screen & stop sign. I was able to boot High Sierra from a usb drive but got this error with SATA. 
     
    Everything working great so far.
     
  4. Like
    SwithDrawn reacted to SavageAUS in [Pre-Release] MacOS High Sierra 10.13.2   
    Beta 6 installed on by my hacks without issue.
  5. Like
    SwithDrawn got a reaction from chris1111 in [Pre-Release] MacOS High Sierra 10.13.2   
    Build 17C85a (beta 6) working well. I started with a working 13.2.1 and had to boot with -liluoff to avoid kernel panic once the install process began, but once it installed simply updating Lilu to 1.2.1 fixed the KP.  
     
    Used chris1111's Nvidia Payload Packager on WebDriver-378.10.10.10.20.109, have not run benchmarks but the 1080 Ti seems to be just fine.,, 
  6. Like
    SwithDrawn reacted to x.di in new possibilities for X79 AppleACPIPlatform panic   
    @Arkanis Give it a try Do not replace AppleAHCI or IntelCpu . With this EFI you will be able to Install and boot vanilla HighSierra and everything will be patched for you
    EFI.zip
  7. Like
    SwithDrawn got a reaction from Max.1974 in AMD Polaris IDs on Sierra / High Sierra   
    Good find, but what was your Luxmark score before the change? 14k is about consistent with one of my 480 8GB cards on Luxball. I added AMDBaffinGraphicsAccelerator to AppleMGPUPowerControl but didn't see any difference. Actually went down by 100, lol. 
  8. Like
    SwithDrawn got a reaction from jmacie in AMD Polaris IDs on Sierra / High Sierra   
    A huge thanks to everyone working on this -- was able to get my two Asus Dual-Fan RX480 8G's working on 10.12.3! 
     
    1) Started with a single 480 and Nvidia GT710 as a helper card (I have a Matrox IGPU so it's no good). For some reason the clover injection methods weren't working for me, I ended up editing info.plist in AMDRadeonX4100.kext, changing 0x67FF1002 to 0x67DF1002. That gave me acceleration on the 480, and I applied the CU patch to unlock all 36 compute units. I still needed the GT710 to boot. I tried putting in the second 480 but wasn't able to get a boot even with the GT710 helper card -- the setup only worked with a single 480 + helper card. 
     
    2) I modified the 480's vbios so I wouldn't need the GT710 (thanks to Rhys216's method). This worked fine, I hoped this would allow me to insert the second 480 without needing to mod its bios but this wasn't the case. I couldn't get OS X to boot up with the unmodded 480 alongside the modded 480 even with the GT710 helper card still in the case. 
     
    3) I modded the second 480's vbios and presto - both cards working in OS X and I don't need the GT710. Edit: for those interested in using the modded bios 480's in Windows, I had to use Crimson 16.9.1 as recent drivers check the bios signature and I got Error 43 in the Device Manager. 
     
    4) With just the two 480's and no post/bios/bootloader showing up, I have two config.plist files - one specifies the MacOS disk as default and the other selects the Windows disk. I switch them out depending on which OS I want to boot up. I used this guide to mount the clover EFI partition in Windows. 
     
    My Luxmark scores about doubled from the single 480. That being said, even on just one 480 Adobe Premiere is very unstable with OpenCL but I can use Metal API for rendering and it seems solid. My test scene with Metal API trounced the same scene on Windows with OpenCL, and they were both way better than my previous setup (two R9 280s). Edit: I've found that removing the extra CU patch (so CUs go from 36 to 16) improves my OpenCL stability in Premiere. 
     

  9. Like
    SwithDrawn got a reaction from jmacie in AMD Polaris IDs on Sierra / High Sierra   
    I have the Asus Rx 480 Dual-Fan 8GBs (I believe they are non reference). With regards to flashing, here's the original post (though it looks like you've already seen it lol). His instructions are a little vague, I used ATIFlash to back up the original bios, modified it like in his video and reflashed. That prevented the card from posting. Obviously do this at your own risk. It wasn't clear to me why he was using a downloaded BIOS instead of the original extracted one, which worked fine for me -- probably less of a big deal if you have a reference card. 
     
    I wasn't able to boot into Windows without a helper card until I installed the Crimson drivers, which cleared Error 43 in the Device Manager. After that, it booted up on the 480s just like OSX (no Windows loading/logo but video turns on when the login screen appears). 
  10. Like
    SwithDrawn got a reaction from jmacie in AMD Polaris IDs on Sierra / High Sierra   
    A huge thanks to everyone working on this -- was able to get my two Asus Dual-Fan RX480 8G's working on 10.12.3! 
     
    1) Started with a single 480 and Nvidia GT710 as a helper card (I have a Matrox IGPU so it's no good). For some reason the clover injection methods weren't working for me, I ended up editing info.plist in AMDRadeonX4100.kext, changing 0x67FF1002 to 0x67DF1002. That gave me acceleration on the 480, and I applied the CU patch to unlock all 36 compute units. I still needed the GT710 to boot. I tried putting in the second 480 but wasn't able to get a boot even with the GT710 helper card -- the setup only worked with a single 480 + helper card. 
     
    2) I modified the 480's vbios so I wouldn't need the GT710 (thanks to Rhys216's method). This worked fine, I hoped this would allow me to insert the second 480 without needing to mod its bios but this wasn't the case. I couldn't get OS X to boot up with the unmodded 480 alongside the modded 480 even with the GT710 helper card still in the case. 
     
    3) I modded the second 480's vbios and presto - both cards working in OS X and I don't need the GT710. Edit: for those interested in using the modded bios 480's in Windows, I had to use Crimson 16.9.1 as recent drivers check the bios signature and I got Error 43 in the Device Manager. 
     
    4) With just the two 480's and no post/bios/bootloader showing up, I have two config.plist files - one specifies the MacOS disk as default and the other selects the Windows disk. I switch them out depending on which OS I want to boot up. I used this guide to mount the clover EFI partition in Windows. 
     
    My Luxmark scores about doubled from the single 480. That being said, even on just one 480 Adobe Premiere is very unstable with OpenCL but I can use Metal API for rendering and it seems solid. My test scene with Metal API trounced the same scene on Windows with OpenCL, and they were both way better than my previous setup (two R9 280s). Edit: I've found that removing the extra CU patch (so CUs go from 36 to 16) improves my OpenCL stability in Premiere. 
     

  11. Like
    SwithDrawn got a reaction from jmacie in AMD Polaris IDs on Sierra / High Sierra   
    A huge thanks to everyone working on this -- was able to get my two Asus Dual-Fan RX480 8G's working on 10.12.3! 
     
    1) Started with a single 480 and Nvidia GT710 as a helper card (I have a Matrox IGPU so it's no good). For some reason the clover injection methods weren't working for me, I ended up editing info.plist in AMDRadeonX4100.kext, changing 0x67FF1002 to 0x67DF1002. That gave me acceleration on the 480, and I applied the CU patch to unlock all 36 compute units. I still needed the GT710 to boot. I tried putting in the second 480 but wasn't able to get a boot even with the GT710 helper card -- the setup only worked with a single 480 + helper card. 
     
    2) I modified the 480's vbios so I wouldn't need the GT710 (thanks to Rhys216's method). This worked fine, I hoped this would allow me to insert the second 480 without needing to mod its bios but this wasn't the case. I couldn't get OS X to boot up with the unmodded 480 alongside the modded 480 even with the GT710 helper card still in the case. 
     
    3) I modded the second 480's vbios and presto - both cards working in OS X and I don't need the GT710. Edit: for those interested in using the modded bios 480's in Windows, I had to use Crimson 16.9.1 as recent drivers check the bios signature and I got Error 43 in the Device Manager. 
     
    4) With just the two 480's and no post/bios/bootloader showing up, I have two config.plist files - one specifies the MacOS disk as default and the other selects the Windows disk. I switch them out depending on which OS I want to boot up. I used this guide to mount the clover EFI partition in Windows. 
     
    My Luxmark scores about doubled from the single 480. That being said, even on just one 480 Adobe Premiere is very unstable with OpenCL but I can use Metal API for rendering and it seems solid. My test scene with Metal API trounced the same scene on Windows with OpenCL, and they were both way better than my previous setup (two R9 280s). Edit: I've found that removing the extra CU patch (so CUs go from 36 to 16) improves my OpenCL stability in Premiere. 
     

  12. Like
    SwithDrawn reacted to nawcom in OS X Mavericks 10.9.3 to dev   
    The native kext NVDAStartup is actually an injector kext among other things from Apple. (I say other things because I notice that if I prevent it from loading on my laptop with the GT525M Fermi card which also requires GraphicsEnabler to boot up right, I get good ol 1024x768 VESA, so it's depended on by other kexts like the one for Fermi architecture cards.) If you open the kext's binary up in a disassembler app like IDA Pro you can see where it probes your video card for its architecture (Kepler, Maxwell, Tesla, Fermi) and in other functions, it injects its determined NVCAP, NVDA,Display-*, etc. So I don't use GraphicsEnabler or any other form, mainly because
     
    a.) It only supports one card, and I have 2 cards I want enabled. I of course can't use SLI in OS X but nonetheless I want both of my cards working separately and enabled.
    b.) As of now GraphicsEnabler for Chameleon (I haven't checked on Clover) only supports single cards with the old limit of 2 displays enabled, NVDA,Display-A and NVDA,Display-B. Kepler cards (GK1xx) extend to NVDA,Display-C and NVDA,Display-D, which is how you're able to get more than 2 displays going. 
     
    I could inject via DSDT and include NVDA,Display-C and D but why should I if NVDAStartup injects it already? Same for NVCAP values - NVDAStartup.kext generates it. Pretty much if you have a Kepler architecture NVIDIA card you shouldn't have to use any injector for everything to work normally (for the most part, I'm sure there are situations where it isn't perfect, though I haven't run into any myself). And like I said before, the kext is still generating the same NVCAP from 10.9.0 in 10.9.3, it's just deciding to ignore one of my DVI ports.
     
    That's because NVDAStartup.kext is overriding those injected entries with its own. Pretty much what would happen if you had one NVCAP value injected via DSDT, but also used an injector kext like good ol' Natit.kext at the same time with a different NVCAP - Natit.kext will override since it loads after DSDT/EFI strings are injected and recognized. In the case for my 2 GTX 770s, if I try and boot with DSDT overrides (which I originally did before I understood NVDAStartup) I end up with a black screen, so using both at the same time can cause issues in the end.
     
    Now for my laptop with GT525M that uses GraphicsEnabler - the NVCAP is still the same as before and it worked fine with it, detecting the display and HDMI. Now with the 10.9.3 kexts its screen goes blank, as if the NVCAP is wrong. I roll back the kexts and the screen is back. There's something new in these updated kexts that's doing it. 
     
    GTX770M uses some newer core than desktop GTX770, mine is GK104 while yours is GK106 so there's some differences, though I wouldn't expect you to be having issues because of it.
     
    There's still a lot to figure out as of now. I'll be sure and post here about any updates while I {censored} around with this.
×