Jump to content
Welcome to InsanelyMac Forum

Register now to gain access to all of our features. Once registered and logged in, you will be able to contribute to this site by submitting your own content or replying to existing content. You'll be able to customize your profile, receive reputation points as a reward for submitting content, while also communicating with other members via your own private inbox, plus much more! This message will be removed once you have signed in.


  • Content count

  • Joined

  • Last visited

About ReddestDream

  • Rank
    InsanelyMac Protégé

Recent Profile Visitors

681 profile views
  1. nVidia GPUs and the infamous lag issue, Why is that happening?

    Awesome! Thanks for the info! Looks like 10.13.4 has improved things! Would be cool tho if someone could confirm this on UEFI and with an iMac SMBIOS (maybe 18,3 like mine) tho . . . The cards I'd hope to pair are my 1080ti in the top slot and RX 480 in the bottom in my Z270, 7700K build. But I'd rather that Nvidia just fix the drivers because it's unnecessary heat generated and power drawn. But it would be nice to at least have the option if the Nvidia drivers don't get fixed . . .
  2. nVidia GPUs and the infamous lag issue, Why is that happening?

    Cool. Thank you. And that's fine. I plan to update to 10.13.4 once it is released anyway, so if it's better than I expect, that's great!
  3. nVidia GPUs and the infamous lag issue, Why is that happening?

    AFAIK, Maxwell and later cards do show in Displays, at least on Sierra, even with no driver or monitor installed. They have incomplete info because of the missing driver, but they do appear to be detected by the VESA driver and they can operate in VESA mode w/o the Webdriver. Also, AFAIK, the bug is caused by having a discrete VESA adapter that is not attached to a monitor. And all the reports I've read of people on Reddit and elsewhere attempting AMD+Nvidia configs on High Sierra have ended in gioscreenlockstate bug tears. I believe that hiding the card is necessary, at least on 10.13-10.13.3. Not sure if 10.13.4 has improved anything. I believe it is in fact a bug new in High Sierra in the VESA detection process that leads to the gioscreenlockstate lockup. It did not happen prior to High Sierra AFAIK. To perform the hiding you must add the Nvidia properties here, turn on Inject Nvidia, and make the AMD card primary in the BIOS, and have no monitors attached to the Nvidia card, only the AMD one. You'll need WhateverGreen/RadeonDeInit or whatever you use because the Nvidia card won't be able to help you since it will disappear as soon macOS sets up the graphics. https://github.com/RehabMan/OS-X-Clover-Laptop-Config/commit/9b8610f32098ce224af044c4e482d2b3da2abae0 See the question is, I'm not sure if bogus properties + Inject Nvidia will make the VESA driver hang up on /all/ dGPUs or just that Nvidia one if it detects the Nvidia card first. I can only fit the Nvidia card in my first slot and the AMD one in my second . . . :/
  4. nVidia GPUs and the infamous lag issue, Why is that happening?

    Has anyone here tried running both an AMD and an Nvidia GPU at once, hiding the Nvidia GPU from macOS's VESA driver using Inject Nvidia and bogus properties to avoid the gioscreenlockstate bug in High Sierra when a headless dGPU is present? Such a method works on Optimus laptops, but I'm not sure if it would on a Desktop quite the same way. I have been able to use it to hide the Nvidia card in order to use only the Intel graphics on my Desktop. But I'm wondering if anyone has tried it in an Nvidia+AMD dual dGPU config . . . If I am forced by Nvidia's horrible drivers to use AMD in macOS and Linux, at least this way I can keep my Nvidia card for gaming in Windows . . . :/ https://github.com/RehabMan/OS-X-Clover-Laptop-Config/commit/9b8610f32098ce224af044c4e482d2b3da2abae0
  5. nVidia GPUs and the infamous lag issue, Why is that happening?

    NvidiaGraphicsFixup isn't going to fix it, no matter what version. At this point, if the Webdriver that comes out with 10.13.4 doesn't match or beat the 10x series of drivers in terms of UI performance, I'm going to come up with a way to plug in both my RX 480 and my 1080ti at the same time and not use the 1080ti in macOS (using Inject-Nvida to inject bogus data via the same method laptops with Optimus use if necessary) to get around the High Sierra bugs with multiple dGPUs. This way I can use the 1080ti in Windows for games, but it's a big PITA because I'll have to swap monitor connections around and 4K KVMs I've run across are too expensive to be feasible. On the bright side, my Linux install would also benefit from AMD graphics . . .
  6. nVidia GPUs and the infamous lag issue, Why is that happening?

    I don't think the Nvidia lag is anything like the AMD black screen issue, which was an issue with the init state the macOS AMD drivers expect the card to be in. Fixes like WhateverGreen and RadeonDeInit work by putting the card in the state the driver expects. In fact, Linux users are also having lag issues with their latest Nvidia driver, so I think Nvidia just dropped the ball somewhere internally, maybe with the Spectre/Meltdown mitigations, and didn't test properly or didn't have time on non-Windows OSes. Like the macOS drivers, the Linux drivers are also tied to particular kernel versions and people are modding the older drivers to work on newer kernels in the meantime: https://bbs.archlinux.org/viewtopic.php?id=234241 In fact this is all so bad, I've been tempted to try to put my old RX 480 in alongside my 1080ti and only use the 1080ti in Windows . . .
  7. nVidia GPUs and the infamous lag issue, Why is that happening?

    Feel free to check and confirm. They were released so close together, I can't imagine it's otherwise. Nvidia starting doing iMac Pro builds because someone filed a request for it . . .
  8. nVidia GPUs and the infamous lag issue, Why is that happening?

    AFAIK, 104, 105, and 106 are identical save for the changed build number for iMac Pro. Also, 158 and 159 are only different by build number too.
  9. nVidia GPUs and the infamous lag issue, Why is that happening?

    So, I was reading and some MBP owners have also had Nvidia lag on High Sierra: https://forums.macrumors.com/threads/any-news-about-the-macbook-pro-nvidia-gpu-lag-yet.2088300/page-6 Some users there are also reporting that the 104 Web driver is the best driver they've had too . . .
  10. nVidia GPUs and the infamous lag issue, Why is that happening?

    The issue appears to be CPU Generation-dependent (Mostly affecting Skylake and newer). What CPU do you have in this hack that you say is not lagging?
  11. nVidia GPUs and the infamous lag issue, Why is that happening?

    So, seems like 158 fixes the lag on everything but Skylake and newer, regardless of card or card generation. And it also seems like maybe the lag is somewhat improved on Skylake and newer, but it's still present. I've found that when 158 is not lagging, it actually seems to perform better on my 7700K/1080ti hack than 104 did . . .
  12. Real macs had the lag issue on 156/157 with eGPUs, so . . . And 156 also ran worse than the stock Nvidia driver on my 2012 rMBP . . .
  13. nVidia GPUs and the infamous lag issue, Why is that happening?

    Considering how small the user base is and how easily Apple can break things, it probably does make sense to just wait. ;p AFAIK, 10.13.4 is in second beta. And there's some pressure to release sooner rather than later since it fixes the Indian text bug on the Mac side. https://www.macrumors.com/2018/02/15/telugu-crashing-bug-fixed-ios-11-3-macos-10-13-4/
  14. nVidia GPUs and the infamous lag issue, Why is that happening?

    May also have to do with 10.13.4's improved eGPU support and Nvidia's attempt to improve things on their end in preparation. But the fact that real macs are also having issues with this driver suggests to me that this driver is just flawed and this is a "bad release," but Nvidia doesn't really see a point in fixing it until 10.13.4 actually drops . . .
  15. The best answer I've found after attempting to analyze the changes between 104-106 and 156/157 with Hopper and a hex editor is that this is Nvidia's screw up and the driver is just bad. :/ There doesn't seem to be any magic new check or anything. This driver has some significant structural changes, and something they did, likely in an attempt to improve eGPU support, just resulted in inferior performance. Hopefully, the next release will be better. Even tho it doesn't have the characteristic lag, I've also noticed just outright inferior performance on my 2012 rMBP with the 156 webdriver vs. the 10.13.3 stock Nvidia driver, including issues with display hotplug and severe lag in VMWare virtual machines. Normally, the webdrivers provide the same or slightly better better performance over stock on my 2012 rMBP. Other real mac users with eGPU have also reported the same lag we are seeing on hacks.