Recent Posts

Saw this thread and can't respond.

https://forums.macrumors.com/threads/gtx-750ti-high-sierra-10-13-4-mac-pro-2008-graphics-problems.2113393/

Lots of people talking, pretty few of them know...the answer. Actually none.

The dosdude tool does very useful stuff to get antiques running newer OS's. The "recommended" patches for 3,1 Mac Pro replace a few pieces of AMD drivers and a GPU framework, IIRC. What is not clearly mentioned is that running the patch breaks Nvidia drivers. As in wrecks them. They run like EFI drivers at that point.And just reinstalling them a gazillion times won't fix it.

I tried to undo this patch on my X-Serve and found myself vexed. The patcher itself only gives a partial list of what gets modded. In point of fact, newer AMD drivers are apparently;y written for Nehalem or newer CPUs only. So to get 3,1 creakily through doorway he discovered that using pieces of older drivers worked.

But only way I undid it was to compare kexts and frameworks with real install with no patches on newer machine. Would be nice if there was either an uninstaller or a definitive list of the kexts and frameworks replaced.

Anyhow, I'll go toss a 750Ti in my 3,1 and see what I get. Pretty sure it will be fine.

When you install this patch on 3,1 this set of patches comes up as recommended. So it is easy for folks to install by accident.

What was that about Dave?

Anyone thinking of altering and flashing their logic boards boot rom should use extreme care, all the normal warnings apply, it could brick your Mac, burn your house down, kick your dog, or knock up your teenaged daughter.

Anyway, two completely different targets, obviously  if you are dealing with an add on PCI card that has a ROM chip on it, it's best to just flash that rom chip with a Mac compatible firmware if the device needs to take part in the boot process. However what users have been doing is adding NVME booting to their Boot ROM, because we have no source of M.2 boards with a ROM chip, and I'm not aware any even exists.

There is the hope of the "Holy Grail" of getting cards with GOP roms working with cMPs, but it's just idle talk, no one is really actively working on it, and other than a few reports no one really knows how to do it, or if it will even work.

it seems currently at least existing windows installs dont like the new microcode patched BootROMs for some reason...

if you do a fresh install of windows from what I have read it will work then.

as for the X5687 im wondering if its a multiplier issue the same one that plagues the Blackbird SR2 board when running X5698s, I mentioned it in this thread here https://forums.macrumors.com/threads/mac-pro-x5687-had-to-try-it.2115849/#post-26006902

a card or a mac is in bad hw state.

try to reseat card in pci-e slot, clean somehow pci-e contacts on card, etc, signals not going well somewhere

This is a sequel to this post, so same specs.
Updated since then to 10.13.4 w/ driver 387.10.10.10.30.107.

I can't get past the OS X login screen. After the MVC EFI boot text is done, Monitor 1 shows garbled dots, and monitor 2 shows gray. After a force shutdown, the next attempt shows black at the same point. And almost always, after another force-shutdown, everything boots and logs in fine.  Sometimes it takes more fidgeting, like replugging monitors or rebooting w/ OS driver and back to nVidia driver, and then everything runs fine for a few days.
At first this happened once every 1-2 weeks, so there wasn't much to troubleshoot. It has gotten progressively worse, to the point that it now happens every day. Also, it started happening not just at login. The whole system hangs, about once a week. Keyboard still works, but the UI freezes.

Console shows many gpuRestart kernel panic reports, almost all of them of this type:

Code:

NVDA(DMA): Channel exception! Exception type = 0x1f Access Violation Error (MMU Error 2)
Channel Info: [7, 0x8, 0x3, 0xe8]
Version Info: [com.nvidia.web.GeForceWeb, 10.3.1, 0x4e554e, 1234, 387.10.10.10.30.107, 1]
MMU Error: FAULT_INFO_TYPE_UNSUPPORTED_KIN at 0x5001ef000

And a few

Code:

NVDA(Graphics): Channel exception! Exception type = 0x2c Access Violation Error (MMU Error 3)
Channel Info: [68, 0x43, 0x3a, 0xa03b5]
MMU Error: FAULT_PDE at 0x0

I also still experience this glitch 100% of the time (when dragging windows across screens)
https://imgur.com/a/QP6SAhJ

No Avid card, just a Newertech USB3 card.

I upgraded to 10.13.4 and sleep - at least initiated sleep - does work. (But there's an unrelated bug which Apple declined to fix whereas sleep kills power to PCIe cards so drives get force-ejected upon sleep. So I rarely use that feature anymore.)

But my problems only got worse. Will create a separate post since they have nothing to do with sleep anymore..

Courtesy of Dosdude and lightbulb fun and H982687 we now have easy way to give a little "new" back to these venerable machines.

I was able to add microcode to have latest from Intel.

I blew $40 and got a 5687, no love. Just sits there dumb and deaf. No response to anything but power button.

This was predicted but sure would have been nice if we'd gained a couple CPUs.

Adding NVME boot was also trivial with provided tools and guides. What you see in profiler doesn't really change, but all of a sudden you can boot into NVME drive with ease. And I was happy to see my first gen Samsung reach the 1500 MBs speeds. Beats the 200 or so I got with SATA SSD.

Now for the down side...I have so far confirmed that Windows is just BORKED by this. I have 2 identical 4,1/5,1s. Same CPU (X5690) and I can easily swap drives. My Windows 10 flashing environment SSD boots right to desktop in one, but now flashed 4,1 it crashes at first spinner. Over and over...Windows even told me it was hosed and I needed to repair via outside media.

I put it back in unflashed 4,1 and it got right to desktop. (after "installing new devices")

I decided to be even more scientific, I flashed the 4,1 back to stock boot rom, slid Windows drive in and it booted right up again.

Flashed it back with improvements, Windows a train wreck again.

Next I will try to do a fresh Windows install.

I recall seeing someone else say they had Windows problems, can't recall if they figured out.

But modding for these 2 things is easy now. The way I see it, next for 4,15,1 is to find out why the top CPUs are still unsupported using Intel's latest Microcode.

Got to be a "kosher CPUs" list somewhere else in that rom.

So, I shook off the cobwebs and dug out a 3,1 last night.

The NVME fix is now so easy a young toddler could knock it out, and never drop binky from other hand.

Updated microcode for these?

The big question now is APFS.

Has anyone added this to 3,1 or X-Serve yet? Found an APFS driver, used the DXEInject thing but no magic.

But it didn't put there driver were it goes in a 4,1, in a PE32 image.

Also, I'm having an issue with Windows, will try new install but currently my 4,1 will NOT boot into Windows with modified boot rom. But that will be another thread.

Yeah, the NVME is pretty easy now that others figured it out.

Nice of Apple to leave that supported machines boot plist the same so you can just use a 10.10 one in 10.14 and everything is easy breezy.

Tackled the 3,1 last night, but that will be another thread.

Next is APFS.

Anyone get this done yet?

Sorry Rominator,
I read the fist post, but I couldn't imagine that my Mac is
one of the unsupported Mac's..
Is is like a blind spot to recognise how old my MacPro is..

Hope for the 10.14 beta2 or beta3 and for a new and modern MacPro mid 2019

Regards Zodi

P.S. I'm a little bit disappointed about the strategies of the new Apple leadership