MP 1,1-5,1 Minimal setup for flashing GPU

[barneyntd]

I have a MP 3,1 with an unflashed GTX680 4MB, which I usually use with Catalina (installed with macOS Catalina Patcher) and it works fine. Occasionally, however, I need to run something 32bit, so I keep a copy of 10.11 (the last supported osx for this mac) on a hard disk. Restarting in 10.11 is easy: just use startup disk. Switching back to Catalina is a real pain, because 10.11 can't read the APFS disk it's installed on.

What is the simplest, cheapest, easiest version of windows (or other) I could install to flash the GTX680 and enable the boot options screen? I have no other use for windows, and I don't have money to throw at this, so dodgy hacks preferred over expensive supported methods!

Barney.

 

[sailmac]

Here's a method I used to flash a GPU directly on my 5,1 cMP. I expect it should work on your 3,1. You'll need to tweak some steps to match your specific GPU and to find the GTX680 mac ROM file if you don't already have it, but the main takeaway will be the recipe for creating a bootable linux thumb drive and using the included nVIDIA/AMD tools to flash a GPU. Free, and no windows or dodgy hacks required!

 

[LangdonS]

Windows has like no good things about it.
Use Linux all the way.
Rescatux is always a handy thing to have in a CD somewhere.
Rescatux.org

 

[LangdonS]

Can't you just use HFS+?
I get that APFS is better but HFS+ is still good (Still much better than NTFS)

 

[Macschrauber]

I use a grml Linux thumb drive with nvflash on it.

 

[sailmac]

I use a grml Linux thumb drive with nvflash on it.

Indeed, a grml Linux thumb drive with nvflash (and amdvbflash) is created by the steps I linked to.

Not arguing, just clarifying what's included in the steps! 😀

 

[Bigwaff]

Can't you just use HFS+?

Catalina requires APFS for install.

 

[Macschrauber]

Indeed, a grml Linux thumb drive with nvflash (and amdvbflash) is created by the steps I linked to.

Not arguing, just clarifying what's included in the steps! 😀

Tbh did not read it

 

[LangdonS]

Rescatux is like lithium grease or an OLFA knife or dish soap.
It just has so many uses.
On a computer that old, I would just reccommend Pop!_OS
macOS is simply not worth the trouble on a machine that old.

### WARNING! ###:
DO NOT attempt to erase all partitions and install Pop!_OS.
The EFI system partition is necessary for the Mac to boot.
I once installed Pop!_OS via the Erase all button and had to reformat my drive with a macOS bootable USB drive.
The Pop!_OS installer may say an error occurred during installlation.
If it boots, don't worry about it.

 

[LangdonS]

I use a grml Linux thumb drive with nvflash on it.

This looks interesting.
Unfortunately, Rescatux appears to be somewhat abandoned at this point.
GRML looks like it's still maintained.

 

[Macschrauber]

Its just needs the basic command line, Flashrom runs, too, for dumping the Mac firmware as plan B when MacOs is not booting no more.

 

[barneyntd]

Here's a method I used to flash a GPU directly on my 5,1 cMP.

Thanks sailmac. I tried this, but had problems. Booting into grml worked fine, but xwindows wouldn't start up: my monitor kept cycling between awake with blank screen and asleep with no signal, repeating about every ten seconds. After that I tried nvflash without xwindows (the text is absolutely tiny, but readable if you peer closely) but had problems saving the backup flash: I saved four versions, and they're all different, by 40 ± 5 bytes out of 183296. Something is wrong.

Anyone have any idea what's going on? If I have a hardware error, why does Catalina work? What diagnostics should I run?

Thanks, Barney.

 

[sailmac]

Barney, sorry, but scrutinizing the innards of VROM files is out of my depth! However, I did notice this post in another thread which suggest slight variations between backup dumps is normal.

Did you proceed with flashing a new ROM to the GPU? If yes and the system behaves correctly I wouldn't be overly concerned about what's going on with backups of the old ROM. If necessary, presumably you could find an equivalent original on TechPowerUp and use it instead of one of your "suspicious" backups of the original. But if the new ROM flashed okay and your GPU is running, I'm guessing there is probably very little chance you will ever need to flash it back to original in the future.

 

[barneyntd]

Barney, sorry, but scrutinizing the innards of VROM files is out of my depth! However, I did notice this post in another thread which suggest slight variations between backup dumps is normal.

Did you proceed with flashing a new ROM to the GPU? If yes and the system behaves correctly I wouldn't be overly concerned about what's going on with backups of the old ROM. If necessary, presumably you could find an equivalent original on TechPowerUp and use it instead of one of your "suspicious" backups of the original. But if the new ROM flashed okay and your GPU is running, I'm guessing there is probably very little chance you will ever need to flash it back to original in the future.

That's reassuring! The readme says backup twice and run diff as a check, which fails.

I did, in fact, try the flash but it did some kind of check which said a byte was wrong and reverted the flash process. The nvflash man page lists more than four large screens of almost unreadably tiny text of options, so I haven't actually worked out what nvflash -6 (as recommended by the readme) does. I'll report back after further investigation, but I'm going to be cautious, as I'm a bit worried about bricking the thing.

I wonder if it's been overclocked? I didn't get it new.

 

[sailmac]

I haven't actually worked out what nvflash -6 (as recommended by the readme) does. I'll report back after further investigation, but I'm going to be cautious, as I'm a bit worried about bricking the thing.

I wonder if it's been overclocked? I didn't get it new.

I did a quick dig on the internet to see if I could find something that helps. After piecing together a few sources it looks like

-6 allows firmware and adapter PCI subsystem ID mismatch

There used to be (and maybe still are?) options

-4 allows firmware and sub-vendor or hardware ID mismatch (for example when the VBIOS you want to flash is from a different vendor than the one who originally branded the GPU)

-5 allows firmware and adapter PCI device ID mismatch

To use all of these together one would enter

nvflash --index=1 -4 -5 -6 your.ROM

In case of a mismatch confirm the flash with 'y'

I personally haven’t used the nvflash tool, but if you are sifting through the utility’s help pages that’s probably the gist of what your eyes are squinting to see. BUT NOTE: the info I collected might be out of date! I am not certain all of those options exist in the latest version of nvflash.

If the utility is complaining about a single byte being wrong, perhaps what it really means is it encountered one of the mismatches described above? Seems like it would be safe to try flashing with -4 -6 enabled now that you know it will abort if the ROM itself fails to checkout. And if -4 is no longer available the command probably won’t even run.

If you add a -L to the command it should produce a log that may give you more explicit information about what is not working.

BTW, if you have a second computer on the network and the cMP is connected via ethernet you could trying SSH into it and that way avoid staring at tiny screen font on the cMP — you’d be staring at Terminal in the second machine which should be much easier to interact with.

Also, even if your used GPU was overclocked, that shouldn’t prevent flashing it with a different ROM.