Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MisterAndrew

macrumors 68030
Original poster
Sep 15, 2015
2,881
2,363
Portland, Ore.
AMD announced their new MI200 series accelerators based on a new 6 nm CDNA 2 GPU today. This technology could eventually be found in workstation GPUs appropriate for the Mac Pro. The performance is out of this world with up to
383 TFLOPs of FP16 and 47.9 TFLOPs of FP32 & FP64 compute performance, with 3.2 TB/sec memory bandwidth. It's hard to imagine Apple coming up with something that fast. Maybe the Intel Mac Pro will stick around longer than we expected.

 
Last edited:

Stex

macrumors 6502
Jan 18, 2021
280
188
NYC
I just got a new Radeon RX 5700 that I want to install in my cMP 4,1>5,1 on PCIe slot 1, where I currently have a Sapphire Pulse RX580.

I'm hoping to get some help/confirmation from all the experts in this thread on the following:

1) The RX 5700 does not have a plate. Should I place a protective layer of something (vinyl?) on the top part of the CPU cage? Or there is nothing to worry about even if it makes contact with the CPU cage top?

2) I found info on this in previous posts here or maybe on another thread, but I'd like to double check that I have the correct/proper information: The RX 5700 (180W) has less power demands than my RX 580 (225W), so there is no need to power it via Pixlas or external PSU, correct?

3) Any other consideration/tip I should keep in mind for this upgrade?

Thanks in advance!

Edit: Corrected the power needs of the RX 5700.
 
Last edited:

avro707

macrumors 68000
Dec 13, 2010
1,754
974
I just got a new Radeon RX 5700 that I want to install in my cMP 4,1>5,1 on PCIe slot 1, where I currently have a Sapphire Pulse RX580.

I'm hoping to get some help/confirmation from all the experts in this thread on the following:

1) The RX 5700 does not have a plate. Should I place a protective layer of something (vinyl?) on the top part of the CPU cage? Or there is nothing to worry about even if it makes contact with the CPU cage top?

That cage should have a protective piece on top of it already, at least it does on my 5,1.

I'm interested how you go with the RX5700, mainly so I can get one or two of them as a contingency for my two 5,1 Mac Pros.
 
  • Like
Reactions: prefuse07

prefuse07

Suspended
Jan 27, 2020
895
1,066
San Francisco, CA
I just got a new Radeon RX 5700 that I want to install in my cMP 4,1>5,1 on PCIe slot 1, where I currently have a Sapphire Pulse RX580.

I'm hoping to get some help/confirmation from all the experts in this thread on the following:

1) The RX 5700 does not have a plate. Should I place a protective layer of something (vinyl?) on the top part of the CPU cage? Or there is nothing to worry about even if it makes contact with the CPU cage top?

2) I found info on this in previous posts here or maybe on another thread, but I'd like to double check that I have the correct/proper information: The RX 5700 (180W) has less power demands than my RX 580 (225W), so there is no need to power it via Pixlas or external PSU, correct?

3) Any other consideration/tip I should keep in mind for this upgrade?

Thanks in advance!

Edit: Corrected the power needs of the RX 5700.

2) Yes that's correct -- given the similarities in TDP, it should be a direct swap with the RX-580.

BTW, you may want to read this: https://www.tonymacx86.com/threads/list-of-amd-radeon-navi-cards-in-catalina-that-work.285432/

I am staying tuned to see if you installed it, and I'd like your feedback, as I too am thinking about getting either a 5600XT or a 5700 to replace my 580 (but I am starting to think about pixlas mod)

Thanks!
 
Last edited:

Stex

macrumors 6502
Jan 18, 2021
280
188
NYC
thanks @avro707 and @prefuse07 .

have not gotten to install and test the RX 5700 yet, should be able to do so in the next 24 hours. i will write a report but will likely post it on a different thread as, in retrospect, i think this was not the proper thread for my questions. once posted, i will add the link to my other post here to redirect you both as well as future readers.
 

tracer1

macrumors regular
Jan 6, 2013
240
38
SC
I have a Biostar Radeon RX560 4gb gaming ( dual fan ) that for some unknown reason shows the colors faded or washed out. Don't know if that's the correct terminology for a description of it. You can see the screen and everything works ok just very hard to see it. Anyone have any suggestions as to what would cause this. It had been fine until recently when I "tried" to install oclp, which failed. This was on my 5,1 12 core, 32gb ram. Is the gpu toast now and should I just replace it? Hope this is the right place to post this, if not, let me know.
Thanks
 

ikir

macrumors 68020
Sep 26, 2007
2,134
2,288
I have a Biostar Radeon RX560 4gb gaming ( dual fan ) that for some unknown reason shows the colors faded or washed out. Don't know if that's the correct terminology for a description of it. You can see the screen and everything works ok just very hard to see it. Anyone have any suggestions as to what would cause this. It had been fine until recently when I "tried" to install oclp, which failed. This was on my 5,1 12 core, 32gb ram. Is the gpu toast now and should I just replace it? Hope this is the right place to post this, if not, let me know.
Thanks
Maybe it is an issue with HDR and your display, try to turn off HDR
 

tracer1

macrumors regular
Jan 6, 2013
240
38
SC
My display is a 2012 Apple Cinema Display. I haven't seen anything on it to adjust or turn off HDR. Is there somewhere else this applies. The gpu has been playing fine up to a few days ago when I attempted to install OCLP. Did something wrong evidently, Had to erase my HD completely and since have had this problem. I have a 680, 7980 & a 120 and those play as they always have. Have no idea what the problem is.
Thanks
 

tracer1

macrumors regular
Jan 6, 2013
240
38
SC
Did a little troubleshooting on my RX 560 today. Found out the diplay port doesn't work at all. The DVI port has dim faded screen, so that's out also as far as I'm concerned. HDMI works fine. So not much left to it.
 

tracer1

macrumors regular
Jan 6, 2013
240
38
SC
One other thing on my RX560 is it now shows the card as a RX460 in the "about this Mac". It did show it as a RX560. Anyone know what would cause this or how to fix it? Running Catalina 15.7 currently.
Thanks
 

Ashok.Vardhan

macrumors member
Oct 4, 2017
87
33
Visakhapatnam, India
AMD Radeon Pro WX9100 (stock) - No Display in 10.14.6
---------------------------
Upgraded my OS to 10.14.6 with NVIDIA GeForce GTX TITAN-X using OpenCore 0.7.9 (which uses a generic driver) so there was display. After upgrading OS, shutdown, installed the new GPU: AMD Radeon Pro WX9100 with Mojave 10.14.6, there was no display from any of the mini Display ports of the WX-9100.

Tried all different ports, with different cables (all working cables) still no-go.

I can hear the fans/HDDs running in beat to a booted up OS so I was able to use keyboard shortcuts and blindly shut down the cMP without any issues. So I am guessing the system is booting into OS and thought it might be a driver issue?

So removed the AMD Radeon Pro WX9100, put the NVIDIA GeForce GTX TITAN-X back in and updated OS to latest version using software update. Shutdown, removed the TITAN and inserted the WX9100 back again, even with latest patches, still no-go.

I can hear the fans/HDDs running in the background when booted to OS so I was able to use keyboard shortcuts and blindly shut down the cMP without any issues.

What am I doing wrong? What's the issue?

The AMD Radeon Pro WX9100 is listed as supported by Apple?

Thanks for your insights.
 
  • Like
Reactions: dabotsonline

tsialex

Contributor
Jun 13, 2016
13,004
13,237
Not all AMD WX cards work are supported - Apple never officially supported one with MacPro5,1 like with Sapphire Pulse 580 or MSI Gaming RX 560, just mentioned that "might also be compatible". Some AMD WX video cards have special personalities that clash with the Apple defined frame buffers or are based on GPUs that don't even have drivers on macOS - not the case with other VEGA 10 based video cards, like the VEGA 56 and 64, these two have drivers and compatible frame buffers/personality and reference models are known to work very good with MacPro5,1.

First thing that you can do is to enable screen sharing and access your Mac Pro from another Mac to see if the card is even recognized on System Information Display + PCIe. If not, the card could be DOA.

Other thing, it's relatively common AMD GPUs going crazy when installed on Mac Pros that previously had TITAN GPUs that used NVIDIA web drivers. Some NVIDIA web drivers save some data inside the Mac Pro NVRAM that make AMD GPUs go crazy.
 
Last edited:

Ashok.Vardhan

macrumors member
Oct 4, 2017
87
33
Visakhapatnam, India
First thing that you can do is to enable screen sharing and access your Mac Pro from another Mac to see if the card is even recognized on System Information Display + PCIe. If not, the card could be DOA.

Other thing, it's relatively common AMD GPUs going crazy when installed on Mac Pros that previously had TITAN GPUs that used NVIDIA web drivers. Some NVIDIA web drivers save some data inside the Mac Pro NVRAM that make AMD GPUs go crazy.
Thanks for your suggestions tsialex. Learnt new tools for trouble shooting today: screen-sharing.

I tried these steps:

1. Reset NVRAM (Long-Press power button, Power LED blinks-long beep-chimes and booted. CMD+Option-P+R method (cycled through 6 times) still no go.
2. Used screen-sharing and checked System Report under PCI and looks the the card is not detected. Which is really strange.

Before I installed this card internally in this cMP I had bought it would an eGPU enclosure and it was detected just fine externally in a eGPU enclosure (Razer Core X Chroma Thunderbolt 3) on a macMini M1, so am not sure why it is not getting detected. I sold the eGPU enclosure thinking everything should be fine, so do not have access to the enclosure anymore.

3. I upgraded 10.14.6 with the latest security update Mojave 2021-005

Any other suggestions?
 

tsialex

Contributor
Jun 13, 2016
13,004
13,237
Thanks for your suggestions tsialex. Learnt new tools for trouble shooting today: screen-sharing.

I tried these steps:

1. Reset NVRAM (Long-Press power button, Power LED blinks-long beep-chimes and booted. CMD+Option-P+R method (cycled through 6 times) still no go.
2. Used screen-sharing and checked System Report under PCI and looks the the card is not detected. Which is really strange.

Before I installed this card internally in this cMP I had bought it would an eGPU enclosure and it was detected just fine externally in a eGPU enclosure (Razer Core X Chroma Thunderbolt 3) on a macMini M1, so am not sure why it is not getting detected. I sold the eGPU enclosure thinking everything should be fine, so do not have access to the enclosure anymore.

3. I upgraded 10.14.6 with the latest security update Mojave 2021-005

Any other suggestions?
The card is correctly powered, it's a 1x 6-pin + 1x 8-pin that is rated for 230W and is know to draw much more? I wouldn't eve try this one without Pixla's mod or at least an eVGA PowerLink (not safe for the backplane even with the PowerLiunk). The PCIe power cables are confirmed working?

Test the card with a PC, btw.

Before I installed this card internally in this cMP I had bought it would an eGPU enclosure and it was detected just fine externally in a eGPU enclosure (Razer Core X Chroma Thunderbolt 3) on a macMini M1, so am not sure why it is not getting detected. I sold the eGPU enclosure thinking everything should be fine, so do not have access to the enclosure anymore.

This is weird, maybe I'm missing something here, but AFAIK Apple Silicon Macs don't yet support eGPUs and Apple don't even released AMD drivers on Apple Silicon:

 
Last edited:
  • Like
Reactions: Ashok.Vardhan

Ashok.Vardhan

macrumors member
Oct 4, 2017
87
33
Visakhapatnam, India
The card is correctly powered, it's a 1x 6-pin + 1x 8-pin that is rated for 230W and is know to draw much more? I wouldn't eve try this one without Pixla's mod or at least an eVGA PowerLink (not safe for the backplane even with the PowerLiunk). The PCIe power cables are confirmed working?

This is weird, maybe I'm missing something here, but AFAIK Apple Silicon Macs don't yet support eGPUs and Apple don't even released AMD drivers on Apple Silicon:
1. Powering is not an issue. The input power cables needed to power the WX9100 are similar to what I was using for the TITAN-X Maxwell inputs (1 x 6-pin and 1 x 8-pin). Currently running it off the motherboard, but will eventually use the Pixlas mod.

2. In regards to the macMini, I cannot be sure, my friend said it was the latest one (perhaps he did not know the difference between the M1 and earlier versions?) However, even if it were an older version macMini with a thunderbolt input, the card was detected in that machine, which is what is baffling for me when inserted into this cMP (5.1 mid-2012) that I am using as my main workstation. Unfortunately, I cannot check in an eGPU with another mac, as am in a different, very distant, city with very low penetrations of macs here.

Last try will be to check inside a PC and verify.
 

kmax3991

macrumors newbie
Aug 9, 2021
16
5
I would put the pixlas mod on first, I have the Vega FE AC 16GB running in there, which has basically the same power consumption and eventually it will fry your board. Trust me, I know ?

When it comes to the OC, I had some issues with my graphics, too, especially with the DP and the boot screen. After I updated my config.plist with the infos of my card (see post #1 in the OpenCore thread), I got rid of all issues concerning my Vega.
Maybe this could help ??‍♂️

Greetings
kmax
 
  • Like
Reactions: Ashok.Vardhan

Ashok.Vardhan

macrumors member
Oct 4, 2017
87
33
Visakhapatnam, India
I would put the pixlas mod on first, I have the Vega FE AC 16GB running in there, which has basically the same power consumption and eventually it will fry your board. Trust me, I know ?
I tried the Pixlas mod, still not getting detected in High Sierra or even in Mojave. Power is getting supplied fine (the mod was functioning with the Titan X).

Do not have access to another PC currently to check the card, but will eventually do that.

Could you please point out the specific instructions that you did to modify your config.plist file? I did check the first post but did not find anything helpful there.

What could be the issue? It was detected fine in an external enclosure as an eGPU in a mac mini.

Really puzzling.

Regards.
 

kmax3991

macrumors newbie
Aug 9, 2021
16
5
Could you please point out the specific instructions that you did to modify your config.plist file? I did check the first post but did not find anything helpful there.
Check "complete your setup" in the mentioned post concerning kexts. I did exactly, what is mentioned there and now I have the full potential of my Vega FE.
To enable hardware acceleration and other missing features such as Night Shift, Airplay to Mac and Universal Control, you'll need three extra components:

Component ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀Description ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀What to do ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀
LiluPatcherDownload the latest RELEASE version
WhateverGreenLilu plugin for GPU patchesDownload the latest RELEASE version
FeatureUnlockLilu plugin for missing featuresDownload the latest RELEASE version

➊ Copy the files "Lilu.kext", "WhateverGreen.kext" and "FeatureUnlock.kext" to your ESP as indicated below:

ESP
└── EFI
⠀⠀ ⠀├── BOOT
⠀⠀ ⠀└── OC
⠀⠀⠀⠀ ⠀ ⠀├── ACPI
⠀⠀⠀⠀ ⠀ ⠀├── config.plist
⠀⠀⠀⠀ ⠀ ⠀├── Drivers
⠀⠀⠀⠀ ⠀ ⠀├── Kexts
⠀⠀⠀⠀ ⠀ ⠀│⠀ ⠀├── Lilu.kext
⠀⠀⠀⠀ ⠀ ⠀│⠀ ⠀├── WhateverGreen.kext
⠀⠀⠀⠀ ⠀ ⠀│⠀ ⠀└── FeatureUnlock.kext
⠀⠀⠀⠀ ⠀ ⠀├── OpenCore.efi
⠀⠀⠀⠀ ⠀ ⠀├── Resources
⠀⠀⠀⠀ ⠀ ⠀└── Tools

➋ Enable the kexts:

Find ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀Replace ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀
<key>Kernel</key>
<dict>
<key>Add</key>
<array/>
<key>Block</key>






















































<key>Kernel</key>
<dict>
<key>Add</key>
<array>
<dict>
<key>Arch</key>
<string>x86_64</string>
<key>BundlePath</key>
<string>Lilu.kext</string>
<key>Comment</key>
<string>Patcher</string>
<key>Enabled</key>
<true/>
<key>ExecutablePath</key>
<string>Contents/MacOS/Lilu</string>
<key>MaxKernel</key>
<string></string>
<key>MinKernel</key>
<string></string>
<key>PlistPath</key>
<string>Contents/Info.plist</string>
</dict>
<dict>
<key>Arch</key>
<string>x86_64</string>
<key>BundlePath</key>
<string>WhateverGreen.kext</string>
<key>Comment</key>
<string>Lilu plugin for GPU patches</string>
<key>Enabled</key>
<true/>
<key>ExecutablePath</key>
<string>Contents/MacOS/WhateverGreen</string>
<key>MaxKernel</key>
<string></string>
<key>MinKernel</key>
<string></string>
<key>PlistPath</key>
<string>Contents/Info.plist</string>
</dict>
<dict>
<key>Arch</key>
<string>x86_64</string>
<key>BundlePath</key>
<string>FeatureUnlock.kext</string>
<key>Comment</key>
<string>Lilu plugin for missing features</string>
<key>Enabled</key>
<true/>
<key>ExecutablePath</key>
<string>Contents/MacOS/FeatureUnlock</string>
<key>MaxKernel</key>
<string></string>
<key>MinKernel</key>
<string></string>
<key>PlistPath</key>
<string>Contents/Info.plist</string>
</dict>
</array>
<key>Block</key>

To further configure WhateverGreen for hardware acceleration, you'll need a special utility:

Utility ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀⠀Description ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀What to do ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀
gfxutilFor getting device pathsDownload the latest RELEASE version

➌ Copy the executable "gfxutil" to your home folder.
➍ In Terminal, enter the following command:

xattr -c gfxutil

➎ Get the path of your graphics card by entering the following command:

./gfxutil -f display | sed 's|.*= ||'

If your graphics card is in slot 1, the path should be PciRoot(0x0)/Pci(0x3,0x0)/Pci(0x0,0x0).

➏ Add the device properties, replacing "PciRoot(0x0)/Pci(0x3,0x0)/Pci(0x0,0x0)" with the actual path:

Find ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀Replace ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀ ⠀
<key>DeviceProperties</key>
<dict>
<key>Add</key>
<dict/>
<key>Delete</key>







<key>DeviceProperties</key>
<dict>
<key>Add</key>
<dict>
<key>PciRoot(0x0)/Pci(0x3,0x0)/Pci(0x0,0x0)</key>
<dict>
<key>rebuild-device-tree</key>
<integer>0</integer>
<key>unfairgva</key>
<integer>1</integer>
</dict>
</dict>
<key>Delete</key>

You might want to hang on to gfxutil because it will be used in a later step; afterwards, you can delete it.

I think you already know, but always check the config.plist status via:
Verify your config

➊ Save your changes.
➋ In Terminal, enter the following command:

plutil -convert xml1 config.plist && plutil config.plist

This command will take care of fixing the formatting of your config (so don't worry about the indentation). You should see "config.plist: OK". If not, recheck your edits. Once the check passes, you can proceed to load your config.

Still, if there is any problem with the compatibility of the hardware I don't see, what else can be done.

Concerning the eGPU:
There is indeed a difference between a GPU being the primary or the secondary source for the system. A eGPU will always be peripheral, which makes it easier to implement into a system.

One last thought:
Through recent experience I know once more, how fragile a system with OC can be. I had one SSD with HighSierra as backup with modified EFI (not installed), one SSD with unmodified EFI, also on HighSierra and a third one was configured with Monterey, the EFI also modified for OC. After a single change in the PCIe hardware configuration I had no OC boot screen anymore. I removed the Monterey SSD, threw in the HighSierra with mods and.... Voila. OC boot with no problems. I reinstalled the Monterey SSD and it worked just fine again. Don't ask me, how. Sometimes my system is stuck in the loading screen. I do a hardreset and it works again. Sometimes it is a little troublesome o_O

Let's see, if it helped some, otherwise I can also recommend the thread about GPU Compatibility, there are also some interesting tipps for not officially supported GPUs for the cMP.

Greets
kmax3991
 
  • Like
Reactions: Ashok.Vardhan

Stex

macrumors 6502
Jan 18, 2021
280
188
NYC
Prices are almost back to OEM levels:
W6600
W6800
RX 6900 XT

... and @Syncretic (thanks!) has done his magic once again extending cMP longevity!

 

Ashok.Vardhan

macrumors member
Oct 4, 2017
87
33
Visakhapatnam, India
... and @Syncretic (thanks!) has done his magic once again extending cMP longevity!

Thanks for the update. Would not have known about this new development if it were not for it.

This part of Syncretic's message is what confuses me though:

"And now, back to the AVX integration problem that's been driving me up the wall"

My understanding is that the CPU that the cMPs have are not compatible because of AVX instructions? How does that change with this new patch?
 
Last edited:
  • Like
Reactions: prefuse07 and Stex

Stex

macrumors 6502
Jan 18, 2021
280
188
NYC
Thanks for the update. Would not have known about this new development if it were not for it.

The past of Syncretic's message is what confuses me though:

"And now, back to the AVX integration problem that's been driving me up the wall"

My understanding is that the CPU that the cMPs have are not compatible because of AVX instructions? How does that change with this new patch?

I don't know the answer, but my guess is that there might have been uncertainty on the AVX instruction problem being the actual cause, or maybe the patch is providing an alternative method and by-passing that problem...? It would be helpful to know more, you could ask for clarification on the AVX thread pasted below, also by @Syncretic :

 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.