Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Status
The first post of this thread is a WikiPost and can be edited by anyone with the appropiate permissions. Your edits will be public.
Maybe difference in M4000M is because you did not boost clock for testing.

Have you tried boosting P3000M clock to 1215 MHz ?? According to GPU-Z info you posted standard clock is 1088 MHz and boost is 1215 MHz.
Pascal p3000 core frequency is not limited to accelerate without overclocking, testing time gpu-z view frequency high when close to 1500ghz, Pascal impact score is the power consumption wall and temperature wall, it is clear that the temperature is no problem that is the power consumption wall received a limit to affect the frequency has been accelerated, testing p3000 power consumption has been in the 50w-70w between changes. m4000m can always work in 95w-100w

I'm pretty sure that both m4000m are are refreshed my own modified 1127ghz run scores do have a slight difference.


I bought the p4000 on my side for three times the price of the m4000m. p3000 is twice the price of the m4000m.


Translated with www.DeepL.com/Translator (free version)
 
Pascal p3000 core frequency is not limited to accelerate without overclocking, testing time gpu-z view frequency high when close to 1500ghz, Pascal impact score is the power consumption wall and temperature wall, it is clear that the temperature is no problem that is the power consumption wall received a limit to affect the frequency has been accelerated, testing p3000 power consumption has been in the 50w-70w between changes. m4000m can always work in 95w-100w

I'm pretty sure that both m4000m are are refreshed my own modified 1127ghz run scores do have a slight difference.


I bought the p4000 on my side for three times the price of the m4000m. p3000 is twice the price of the m4000m.


Translated with www.DeepL.com/Translator (free version)
If limit on Pascal is TDP, there are tools to change it (don't know if this works on Px000M cards) like Mobile Pascal TDP Tweaker

Great price on the P4000M, here in Europe I'm getting asked around $550 for a single P4000M :(

Can you please post vbios that came with your P3000M and P4000M ?? I'd like to take a look.
 
Last edited:
If limit on Pascal is TDP, there are tools to change it (don't know if this works on Px000M cards) like Mobile Pascal TDP Tweaker

Great price on the P4000M, here in Europe I'm getting asked around $550 for a single P4000M :(

Can you please post vbios that came with your P3000M and P4000M ?? I'd like to take a look.
I tried the Mobile Pascal TDP Tweaker tool and couldn't modify anything very limited. I can only edit the hex code by myself to work.
 
iMac Quadro P3000M & P4000M roms:

- Enables the internal display
- Enables GOP boot-screen
- Brightness control with OpenCore
- Full Boost clocks in MacOS & Windows
- Power Limit increased (with AfterBurner)

Sadly there is still no certificate hacked nvflash for Pascal,
so patched roms must be flashed using an EEPROM Clip
if you are using a CH341A you MUST use the 1.8V adapter!

1.8v.png



You should be able to install un-modified Windows drivers,
or the correct Web Drivers for your exact MacOS build:

http://www.macvidcards.com/drivers.html

https://dortania.github.io/GPU-Buyers-Guide/modern-gpus/nvidia-gpu.html

Booting from OpenCore will now enable the display even without drivers,
so you can actually boot newer MacOS versions with no web driver support,
(obviously with no acceleration but basic 2D works, even Youtube videos etc)

Brightness control in MacOS requires OpenCore to do these four things:
- Load the web drivers using nvda_drv_vrl=1 (add to OpenCore's boot args)
- Load a backlight kext to configure the brightness range on the iMac display
- Spoof the "shiki ID" or Board ID to iMac 14,2 (Mac-27ADBB7B4CEE8E61)
- Inject some additional Device Properties, just add them to your config.plist:

Code:
<key>DeviceProperties</key>
<dict>
<key>Add</key>
<dict>
<key>PciRoot(0x0)/Pci(0x1,0x0)/Pci(0x0,0x0)</key>
<dict>
<key>@0,backlight-control</key>
<data>AQAAAA==</data>
<key>@0,built-in</key>
<data>AQAAAA==</data>
<key>@0,display-cfg</key>
<data>/////w==</data>
<key>@0,display-connect-flags</key>
<data>BAAAAA==</data>
<key>@0,pwm-info</key>
<data>AhgAZHhpAAAQJwAAjAoAAAAEAAAAAAAA</data>
<key>APPL,backlight-control</key>
<data>AQAAAA==</data>
<key>agdpmod</key>
<string>vit9696</string>
<key>applbkl</key>
<data>AQAAAA==</data>
<key>shikigva</key>
<integer>40</integer>
</dict>


These roms have a standard un-patched Nvidia GOP that will only load on UEFI 2.0
So to enable boot-screens with OpenCore / rEFInd+ you must check your config file to
make sure "reload_gop" or "ReloadOptionROMs" + "ForgeUefiSupport" are enabled.
OpenCore Config.plist is attached, it might also fix the sleep issue? (up to High Sierra only)

External displays are enabled but UNTESTED so might not work with these vBIOS.
PWR versions have their power limits patched with Mobile Pascal TDP Tweaker,
so it can be increased in AfterBurner, but I'm not sure these changes are effective..??


Going off-topic (again), those Turing RTX 3000 / 4000 MXM cards could also be good options.
There are no MacOS web drivers for any Turing cards but they might have one last surprise:
It is possible that these cards could support the NATIVE Apple boot-screen / boot-selector!
It is known that desktop Turing cards can show the boot-selector on MacPro (without OpenCore)
If you can find a card at a good price it would be an interesting experiment, to complete the set...


PS - Updated iMac Quadro M4000M & M5000M roms:

I updated the M4000M rom to clock higher
and have tried to enable external displays.
I have also included a rom for the M5000M.
 

Attachments

  • P3000M+P4000M.zip
    602.6 KB · Views: 324
  • M4000M+M5000M.zip
    283.4 KB · Views: 287
  • Config.zip
    7.7 KB · Views: 293
Last edited:
Thanks again @Santa's Little Helper! I ran a few benchmarks this morning with the new higher clock (1050mhz) on M4000M.

@1050mhz the M4000M uses approximately 82w and sits at 70C in Vulkan. Bumping the core clock to 1185mhz increases the GPU wattage to 94w and temp to 80C but completes the test. I also ran Valley at 1185mhz and scored 4091 with no black screen issues or driver crashes.

M4000M @ 1050mhz Vulkan
Vulkan1050.png


M4000M @ 1185 (1050 + 135 from MSI Afterburner)
Vulkan1185HWM.png



M4000M Valley @ 1185mhz (1050 + 135 from MSI Afterburner)
Valley1185.png
 
Awesome work @Santa's Little Helper !!! Thanks a lot for this!

I've been testing the M4000M new vbios, and can confirm that the external display works on port closest to the USB ports. I only get 1080p but I think this is normal as I'm just using a passive cable, maybe with an active Mini DisplayPort to HDMI Adapter it is possible to go beyond 1080p.

At 1185 Mhz I get similar watts and temps as @G3llings, but driver/OS crash on Valley test and 3DMark Fire Strike (they both make use of DX11). They crash early on the test, and the card does not even have time to heat beyond 60ºC. Geekbench 5 Vulkan and 3DMark Time Spy pass the tests however (Vulkan and DX12 are more "parallel" processing optimized).

Just by lowering overclock to 1150 Mhz it seems I can pass all the tests without problem.


I feel tempted to buy one of those Pascal cards for testing...
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Awesome work @Santa's Little Helper !!! Thanks a lot for this!

I've been testing the M4000M new vbios, and can confirm that the external display works on port closest to the USB ports. I only get 1080p but I think this is normal as I'm just using a passive cable, maybe with an active Mini DisplayPort to HDMI Adapter it is possible to go beyond 1080p.

At 1185 Mhz I get similar watts and temps as @G3llings, but driver/OS crash on Valley test and 3DMark Fire Strike (they both make use of DX11). They crash early on the test, and the card does not even have time to heat beyond 60ºC. Geekbench 5 Vulkan and 3DMark Time Spy pass the tests however (Vulkan and DX12 are more "parallel" processing optimized).

Just by lowering overclock to 1150 Mhz it seems I can pass all the tests without problem.


I feel tempted to buy one of those Pascal cards for testing...
I am going to go back and run valley again later this evening. Based on my score I'm thinking I may have run at the 1050mhz since it was lower than your previous score. I will see if my card can pass at 1185mhz or not.

Edit: I confirmed by re-running at 1185mhz and scored very similarly. No crashes. We must have different settings, causing the lower score. I did play a couple different games as well with no crashes.
 
Last edited:
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
For reference, here is the OpenCore EFI I used for High Sierra testing on the M4000M. It is just a quick hack I did for testing with Santa's indications. Brightness control works (max brightness is too dim). It's in no way optimized and I think it loads some kext not intended for High Sierra, and may even create some conflicts, but it worked for me. I may try to make a better one once I reinstall High Sierra as I just accidentally wiped the High Sierra partition while installing windows 10 :rolleyes:

I compressed it with 7zip as it was too big to upload as a zip file, so if extraction does not work just change extension from .zip to .7z


Testing this version of OpenCore EFI boot does not match p3000 I don't know if @Santa's Little Helper modifications to vbios are incompatible or a configuration issue. There is no simulated boot screen, Windows 10 enters the system with the driver loaded will have an internal screen display. Add by registry modification also does not support brightness adjustment. There are two monitors inside the settings. Maybe it's an OpenCore configuration problem!

Translated with www.DeepL.com/Translator (free version)
 
I am going to go back and run valley again later this evening. Based on my score I'm thinking I may have run at the 1050mhz since it was lower than your previous score. I will see if my card can pass at 1185mhz or not.

Edit: I confirmed by re-running at 1185mhz and scored very similarly. No crashes. We must have different settings, causing the lower score. I did play a couple different games as well with no crashes.
I guess not all cards are made equal, and we may be running very close to the limits. Also Valley is quite an old benchmark and CPU plays a role (I think you had a Xeon and I have an i7). Anyway, great it works for you! :)
 
For some reason I can't find
Testing this version of OpenCore EFI boot does not match p3000 I don't know if @Santa's Little Helper modifications to vbios are incompatible or a configuration issue. There is no simulated boot screen, Windows 10 enters the system with the driver loaded will have an internal screen display. Add by registry modification also does not support brightness adjustment. There are two monitors inside the settings. Maybe it's an OpenCore configuration problem!

Translated with www.DeepL.com/Translator (free version)
I think there's a a new OpenCore parameter needed for reloading non patched EFI 2.0 on Pascal cards. For some reason @Santa's Little Helper post with vbios and instructions for Pascal cards have disappeared from thread :eek: so I'll have to look for it.

Edit: post is back :) Seems when you edit a post and have not reached a certain number of messages it gets hold back for a while...
 
Last edited:
Update p3000 The latest test results are not bad than the previous no internal screen display remote desktop test results are quite high, close to the performance of the desktop graphics card gtx1060.

Afterburne overclocking can not do anything, I think to overclock only adjust vbios to work.

With simulated boot options, macOS system brightness adjustment normal brightness can be maximum brightness. windows 10 through the registry to add code to achieve brightness adjustment.

I received a p4000 is bad card has been returned to the seller, I will look to buy a new p4000 to test.

I use a multi-speed voltage adjustment programmer to refresh vbios.
 

Attachments

  • IMG_0808.jpg
    IMG_0808.jpg
    808 KB · Views: 263
  • IMG_0806.jpg
    IMG_0806.jpg
    950.2 KB · Views: 272
  • IMG_0809.jpg
    IMG_0809.jpg
    777.7 KB · Views: 242
  • IMG_0812.jpg
    IMG_0812.jpg
    867.9 KB · Views: 216
  • IMG_0815.jpg
    IMG_0815.jpg
    422.3 KB · Views: 228
  • IMG_0814.jpg
    IMG_0814.jpg
    737.7 KB · Views: 215
  • IMG_0816.jpg
    IMG_0816.jpg
    836.1 KB · Views: 215
  • QQ20210924-051431@2x.jpg
    QQ20210924-051431@2x.jpg
    206.3 KB · Views: 205
  • IMG_0782.jpg
    IMG_0782.jpg
    931.8 KB · Views: 216
  • IMG_0781.jpg
    IMG_0781.jpg
    428.6 KB · Views: 228
  • IMG_0777.jpg
    IMG_0777.jpg
    428.2 KB · Views: 228
Update p3000 The latest test results are not bad than the previous no internal screen display remote desktop test results are quite high, close to the performance of the desktop graphics card gtx1060.

Afterburne overclocking can not do anything, I think to overclock only adjust vbios to work.

With simulated boot options, macOS system brightness adjustment normal brightness can be maximum brightness. windows 10 through the registry to add code to achieve brightness adjustment.

I received a p4000 is bad card has been returned to the seller, I will look to buy a new p4000 to test.

I use a multi-speed voltage adjustment programmer to refresh vbios.
Nice testing, these benchmark values look great! From 3DMark database the P4000M should score close to 4000 on Time Spy test.

There is a very long thread about Pascal TDP tweaker here. I have not read all of it, but it seems for Pascal it is more efficient for overperformance to make a hardware mod (shunt mod) on the cards than patching vbios. The hardware mod (reduce shunt resistor value) makes the GPU think it is consuming less power and avoids throttling based on power limit. Changing a R005 to R004 shunt resistor will make GPU think it is consuming 20% less power.

I don't know if this mod is possible for MXM cards. Shunt resistors look like this:
1632785071382.png


@jay508 can you please post a high res pictures of your P3000 card ? From online pictures it is very hard to find something on the card pcb.
 
Last edited:
Nice testing, these benchmark values look great! From 3DMark database the P4000M should score close to 4000 on Time Spy test.

There is a very long thread about Pascal TDP tweaker here. I have not read all of it, but it seems for Pascal it is more efficient for overperformance to make a hardware mod (shunt mod) on the cards than patching vbios. The hardware mod (reduce shunt resistor value) makes the GPU think it is consuming less power and avoids throttling based on power limit. Changing a R005 to R004 shunt resistor will make GPU think it is consuming 20% less power.

I don't know if this mod is possible for MXM cards. Shunt resistors look like this:
View attachment 1850418

@jay508 can you please post a high res pictures of your P3000 card ? From online pictures it is very hard to find something on the card pcb.
Yes Afterburnner has no effect on Pascal overclocking can be adjusted can not be saved to take effect. I wonder if the power wall limit can be adjusted by modifying vbios.

I do not have a professional camera, shot with iPhone 12 pro Max do not know if it is clear enough.
 

Attachments

  • IMG_0836.jpg
    IMG_0836.jpg
    663.7 KB · Views: 199
  • IMG_0833.jpg
    IMG_0833.jpg
    766.6 KB · Views: 203
  • Like
Reactions: m0bil
Yes Afterburnner has no effect on Pascal overclocking can be adjusted can not be saved to take effect. I wonder if the power wall limit can be adjusted by modifying vbios.

I do not have a professional camera, shot with iPhone 12 pro Max do not know if it is clear enough.
Yes, pictures are clear enough, thank you.

Good news is that there are shunt resistors on the card:

1632815887000.png

1632816026889.png


So, in theory card can be shunt modded to increase TDP. I will search the net to see if someone has done this on a MXM card before.
 
Cool! I did this on my desktop 1080, but that was water-cooled.
Maybe it's time to drill a big hole in the iMac to mount an AIO cooler!
Don't even think about using liquid metal on them, that will end badly...


Shunt mods could actually be quite dangerous on a mobile card's VRM.
They only help if the card is actually power-limited (green bars in GPUz)
and still won't increase the boost clock limit, just help hold that max clock.
The boost clock seems to be locked and protected on Quadro mobile cards...

Before you think about shunt mods I suggest you try the PWR versions first,
they should let you increase the power limit to about 120% in AfterBurner.
You could even try cross-flashing the P4000 roms onto the P3000 card?!?

Make sure you max the fans out, or you'll be temperature limited anyway,
on Pascal the boost clock drops like a stone as the "hot spot" temp increases.

Good luck - don't blame me if the card explodes! ;)
 
Last edited:
Meanwhile, I'm awaiting my WX4170m to be delivered from China.
Hi Nguyen,
Have you got your order delivered?
If yes could you kindly share photo with me, please? (You can send in DM if for some reason don't want to share it on public)
I'm asking cause I was waiting for the same parcel from China (may be even from the same shop as you), but as result got a very strange piece of equipment. And when I try to install it, my Mac doesn't even fully start. It tries to spin coolers for few second and stop, than retries in few seconds and even 2nd LED never comes on not talking about all the rest.
Regards Sergiy
 
There would be no "fully supported" GPU for iMac 2009~2011 and OS newer than High Sierra.
Apple original GPUs => High Sierra.
Quadro K*100m => OS supported but H264 encoding is broken (not a big deal to many, just some extra workload dumped to the CPU)
AMD WX41*0 => No native boot screen, internal LCD need kick-starting (for High Sierra). Work nicely on Big Sur (not Mojave, not Catalina, just Big Sur).
There are other options as well: GTX 680m, 780m, 880m. They are just pricey and aged.

Meanwhile, I'm awaiting my WX4170m to be delivered from China. Not very sure to install it right away or wait until my HD4850m start dying.
Still struggling with the question whether to buy a M4000m and switch off Mac OS forever on my iMac 2009.
Your description is incomplete:

All AMD WX41X0, the WX7100, and the RX480 run with all software features enabled using OCLP with

- macOS Mojave
- macOS Catalina
- macOS Big Sur
- macOS Monterey Beta 1-7

This is covered in detail on the very first post of the iMac GPU thread. You can use OCLP to run with High Sierra, too.

Only on the Mid 2011 models it would be necessary to manually enable the HD3000 iGPU in the DeviceProperties section again, otherwise you loose Airplay …one could strip down the OLCP config then and disable spoofing and all extensions but WhateverGreen and Lilu to enable the display on boot.
 
Hi Nguyen,
Have you got your order delivered?
If yes could you kindly share photo with me, please? (You can send in DM if for some reason don't want to share it on public)
I'm asking cause I was waiting for the same parcel from China (may be even from the same shop as you), but as result got a very strange piece of equipment. And when I try to install it, my Mac doesn't even fully start. It tries to spin coolers for few second and stop, than retries in few seconds and even 2nd LED never comes on not talking about all the rest.
Regards Sergiy

Delivery to Vietnam is always slower than to Rusia because Aliexpress shipping service cannot go directly. The package must go to Singapore first, then to Vietnam. 45 days from Sep 8th is my expectation. If I could receive the card before Oct 15, it would be a magic.
 
  • Like
Reactions: Ausdauersportler
Cool! I did this on my desktop 1080, but that was water-cooled.
Maybe it's time to drill a big hole in the iMac to mount an AIO cooler!
Don't even think about using liquid metal on them, that will end badly...


Shunt mods could actually be quite dangerous on a mobile card's VRM.
They only help if the card is actually power-limited (green bars in GPUz)
and still won't increase the boost clock limit, just help hold that max clock.
The boost clock seems to be locked and protected on Quadro mobile cards...

Before you think about shunt mods I suggest you try the PWR versions first,
they should let you increase the power limit to about 120% in AfterBurner.
You could even try cross-flashing the P4000 roms onto the P3000 card?!?

Make sure you max the fans out, or you'll be temperature limited anyway,
on Pascal the boost clock drops like a stone as the "hot spot" temp increases.

Good luck - don't blame me if the card explodes! ;)
Well, watching this Gallium video is scary :) It seems gallium ends up dissolving solder tin under the shunt resistors, and being conductive it can spill to nearby components causing shorts, absolutely not recommended to use it at all.

It seems not clear that hard modding has any benefit for MXM cards, but I think the right way would be to replace shunt resistors with lower value, and that is difficult: as they conduct high current - low voltage the solder surface on the PCB is large, and special SMD equipment and/or heat gun is required.

Looking at P3000M pictures I've seen the power measurement chip using the shunt resistors is the well known INA3221:
1632869566204.png


So, there's probably an easier to perform and better mod soldering low ohm (10 to 100 ohm) resistors on top of theses input capacitors. Some additional info in chinese.

But, I fully agree that vbios modding is the way to go, I'm just leaving the info here as a reference if ever needed (and before I forget the details :))

Also, hard modding and maybe breaking an expensive and good working card for a bit of performance is quite a risk.
 
Last edited:
Cool! I did this on my desktop 1080, but that was water-cooled.
Maybe it's time to drill a big hole in the iMac to mount an AIO cooler!
Don't even think about using liquid metal on them, that will end badly...


Shunt mods could actually be quite dangerous on a mobile card's VRM.
They only help if the card is actually power-limited (green bars in GPUz)
and still won't increase the boost clock limit, just help hold that max clock.
The boost clock seems to be locked and protected on Quadro mobile cards...

Before you think about shunt mods I suggest you try the PWR versions first,
they should let you increase the power limit to about 120% in AfterBurner.
You could even try cross-flashing the P4000 roms onto the P3000 card?!?

Make sure you max the fans out, or you'll be temperature limited anyway,
on Pascal the boost clock drops like a stone as the "hot spot" temp increases.

Good luck - don't blame me if the card explodes! ;)
I have tested the p4000 vbios refresh to p3000 is not available can not be used properly can detect p3000 gpu-z view vbios information blank related information are errors can only be detected to the core code gp104

Several versions of vbios including PWR do not make any difference are limited by the power wall.
 
I have tested the p4000 vbios refresh to p3000 is not available can not be used properly can detect p3000 gpu-z view vbios information blank related information are errors can only be detected to the core code gp104

Several versions of vbios including PWR do not make any difference are limited by the power wall.
What is the power limit you are seeing on the P3000 (how much watts) ?
Also, which is the maximum temperature you get ? Do you get PerfCap reason other than "None" ?

Edit: for reference, stressing my power-hungry 780M I get around 100 watts from PCIe (MXM) slot and 75ºC with ODD fan at 3800 rpm and 3 pipe heatsink.

1632958406778.png
 
Last edited:
Well after about an hour of gaming I had my first grey screen freeze at 1185mhz on m4000m. Turning the display off and back on got me to crash to the desktop and was able to drop down to 1175mhz and go back into the game. We must be right on that line of maximum stable clock like @m0bil said before.
 
  • Like
Reactions: m0bil
What is the power limit you are seeing on the P3000 (how much watts) ?
Also, which is the maximum temperature you get ? Do you get PerfCap reason other than "None" ?

Edit: for reference, stressing my power-hungry 780M I get around 100 watts from PCIe (MXM) slot and 75ºC with ODD fan at 3800 rpm and 3 pipe heatsink.

View attachment 1852457
Look at the chart above I posted a test chart it, the highest load can go to 75w but not always in 75w stable in 45w 65w 75w between changes in the core frequency is always stable. Heavy stress test p3000 temperature stability within 70 degrees. The same conditions test m4000m at about 85 degrees.
 

Attachments

  • QQ20210930-084158@2x.jpg
    QQ20210930-084158@2x.jpg
    733.4 KB · Views: 164
  • Like
Reactions: m0bil
Look at the chart above I posted a test chart it, the highest load can go to 75w but not always in 75w stable in 45w 65w 75w between changes in the core frequency is always stable. Heavy stress test p3000 temperature stability within 70 degrees. The same conditions test m4000m at about 85 degrees.
Well, it seems clear the card is being limited by 75W TDP. Maybe this card can benefit from a hard mod to increase TDP.

The P4000M has TDP of 100W, so little improvement possible here.

Also, I saw there are more modern versions of this Pascal cards, P3200M and P4200M but seem hard to find (some are available on Aliexpress).
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.