Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 1,1-5,1 GPU Compatibility List for cMP
- Thread starter Pentaxer
- WikiPost WikiPost
- Start date
- Sort by reaction score
- Status
If no Aux power required, then shouldn't be powering issue.
You can install RX560 in slot 1, and GT120 in slot 2 (if you have concern about "too close". Or it seems the GT120 block the RX560's cooling, you can install the GT120 into slot 3 or even slot 4. That doesn't really matter). Then connect the monitor to the GT120.
Then you should able to see something like this. Of course, in your case, the card in slot 1 will be RX560, but not 1080Ti.
If you have no idea that's patched or not, then most likely it's unpatched. We can set this aside first.
So, please do this again, and then when the screen show you this info, please unplug the monitor from the GT120, and connect that to the RX560 (You can disconnect and connect DisplayPort / HDMI cable when the power is on).
And please make sure you try ALL the ports on the RX560, just in case.
If possible, you should further update that to 144.0.0.0.0, but this shouldn't make any difference to your existing problem.
Oh wow!! That worked!
Unbelievable… thank you!!
Does that mean that I need to keep the GT120 there all the time?
Also, the graphics are a little bit on something. Some colours around the letters . Its quite nice but wondered if theres something there to do. Also trying to install Mojave but it says I need a compatible metal card and that one is metal supported so not sure!
Thanks a million for your help!
Unbelievable… thank you!!
Does that mean that I need to keep the GT120 there all the time?
Also, the graphics are a little bit on something. Some colours around the letters . Its quite nice but wondered if theres something there to do. Also trying to install Mojave but it says I need a compatible metal card and that one is metal supported so not sure!
Thanks a million for your help!
You need to remove the GT120 in order to install Mojave.
No, the GT120 should not be required. You may try to remove that now and see if the RX560 can work stand alone.
The colour around letters most likely related to the font AA. What's your monitor's resolution? And which rendering resolution you are using now?
Thank you that worked!
The RX560 is working on its own.
I am trying now to install mojave but its super frustrating . It keeps on saying the file that I already have on my computer is damaged. So I got rid of it, try to download a new file from the AppStore but it says open instead of download… grrr
The RX560 is working on its own.
I am trying now to install mojave but its super frustrating . It keeps on saying the file that I already have on my computer is damaged. So I got rid of it, try to download a new file from the AppStore but it says open instead of download… grrr
You have to remove the "damaged" file in order to download the new one. Otherwise, it seems the cMP can always locate the Mojave installer even you move it to another drive.
Anyway, before you try to remove -> re-download the installer again, please try disconnect your cMP from the internet, and set the date to 21 Jul 2021 (assuming you have the latest Mojave installer).
If you don't know which version of Mojave it is, or the date above still give you the "damaged" error, you may try to set the date to 24 Sep 2018.
Anyway, if the graphic card problem is fixed, you better move to a Mojave thread (e.g. this one) to discussion the Mojave installation issue.
It boots fine, but none of the Mozilla based browsers display anything. Although I can see the pointer changing from arrow to a hand (hovering over links). The only browser which works is Chrome.
The "easiest" with decent performance would be the Sapphire Radeon Pulse RX580. Decent performance, recommended by Apple, fits in Slot 1 without impinging on Slot 2, and doesn't need a Pixlas mod for power. The Vega 56 with an EVGA Powerlink can also work without power mods - others will have to tell you exactly how to do that but it's the most powerful GPU that can be used without power mods.
For any option - if you haven't looked at it - check out EnableGop to enable native boot menu with the "Option" key.
Regards,
sfalatko
For any option - if you haven't looked at it - check out EnableGop to enable native boot menu with the "Option" key.
Regards,
sfalatko
Thanks for the info , I definitely will buy evga power link. The power input are the boards 2mini 6 pins? I was thinking titan x for the 12 gb vram . Is it possible to install titan x without PIXLAS mod?
The thing with fe and vii is the mod I guess is required on the psu. What about 5700 vs vega 56 or 580 all the way ?
About motherboard's mini-molex's power delivery (in 4,1->5,1),
Nobody has ever measured what it can deliver?
Apple kept their numbers in secret, which of course was not included in wikileaks...?
Usually there would be a lot of YT clips to proove the claims nowadays...
Nobody has ever measured what it can deliver?
Apple kept their numbers in secret, which of course was not included in wikileaks...?
Usually there would be a lot of YT clips to proove the claims nowadays...
It doesn't "required", I proved that dual mini 6pins can power a Radeon VII as long as you know how to connect the card, and how to set the PowerPlay table accordingly.
How to - downvolt a Radeon VII
I confirmed it's possible to run a radeon VII with the mini 6pin ONLY (no Pixlas mod, no extra PSU, no SATA power, but with power balancing) with my PowerPlay Table kext. I limit the card's power draw to 204W (170 x 120%). This is NOT a random number, but when I perform tests in Windows, I...
forums.macrumors.com
But Pixlas mod is the recommend way to go anyway.
If you worry about the GPU may draw too much, then you can stick with a graphic card that only has one 8pin, or two 6pins. Anything more than that (e.g. 6+8pin) may draw too much.
For RX580, the popular Sapphire PULSE RX580 is a single 8pin card. This is one of the reason why it is so popular.
About 120W, shutdown protection may kicks in once go beyond that.
It's no in any manual. But MR member contacted Apple support, and their official reply is 75W each. Same as normal 6pin.
So,
if I tried to find some old cheap but powerful enough for, lets say, 10.14, I could trust that this mp can drive a gpu with max TDP (110+110+75=) 295W?
I wonder why
if I tried to find some old cheap but powerful enough for, lets say, 10.14, I could trust that this mp can drive a gpu with max TDP (110+110+75=) 295W?
I wonder why
The Definitive Classic Mac Pro (2006-2012) Upgrade Guide
says that you need a psu mod if over 250W...?
The shutdown limits and the 'safe for your Mac' are different.
I would not run your Mac daily up to the shutdown limits, that's asking for trouble.
As recommended above, a pixlas mod is the solution.
I know it’s been discussed in numerous threads and in this thread too, but if you let me gather my thoughts about decision making in this problem, in a form of a list:
A few things to consider closing a GPU for your cMP:
a) is it supported aka is there a driver for my OS
b) does it fit in the case
c) how to feed enough power to it
A few things to consider after choosing a GPU that draws more power than is available according to Apple specs:
d) how much does it draw power from the PCIe slot
e) what is the maximum TDP in wattage for that card
f) how high of power peaks does there occur - because they do occur at some time
Normally almost no card draws 75W from PCIe slot. Probably never more than half of that, and more often even less than half of the spec. So you can’t add that number up to the max power draw for a card. Because of that, you would need the difference to be fed some other way to your card. That would be, from the PCIe power feed connectors of the card itself. There are 2x mini 6-pins. There are SATA power feeds. There is a power supply to tap into in not supported ways (Pixlas or straight inside the PSU, even a motherboard thief connector has been in the design I believe).
I have tried Martin’s aforementioned downvolting and it works just great. And there was no performance hit, might have been the way around in some tests even. With that kext in effect, you could run a Radeon VII without Pixlas mod: 2x mini 6-pin to 8-pin, 2x SATA-power to 6-pin (or 8-pin). These all could be onnected to an EVGA Powerlink, and with that to a VII. And it would work. Probably it would work great and not shutting down even at power peaks appearing at times either.
That would be called ”power balancing” between all the power sources the GPU draws juice from - except the PCIe power, it’s not in the combined balanced draw. PCIe would be the least significant of all of them, and the most moderate with almost any card there is on the market. But it helps with the math of course. A couple of old cards drew excess amounts from the PCIe slot, and it was a catastrophe back in days. I have read about that from MR posts, and from internet archives too, but luckily not experienced it myself. Those cards burned a lot of motherboards with too much power draw from PCIe slot.
ps.
With Martins downvolt kext and a dual Pixlas mod + the trick above, and one a little bent PCIe shield, I have run 3x VIIs in a cMP successfully for quite a some time (can’t remember how long) without any problems. Well, any problems I could regard to VIIs or their power draw at least. I still have that cMP, but I have jumped to other configurations for work already.
A few things to consider closing a GPU for your cMP:
a) is it supported aka is there a driver for my OS
b) does it fit in the case
c) how to feed enough power to it
A few things to consider after choosing a GPU that draws more power than is available according to Apple specs:
d) how much does it draw power from the PCIe slot
e) what is the maximum TDP in wattage for that card
f) how high of power peaks does there occur - because they do occur at some time
Normally almost no card draws 75W from PCIe slot. Probably never more than half of that, and more often even less than half of the spec. So you can’t add that number up to the max power draw for a card. Because of that, you would need the difference to be fed some other way to your card. That would be, from the PCIe power feed connectors of the card itself. There are 2x mini 6-pins. There are SATA power feeds. There is a power supply to tap into in not supported ways (Pixlas or straight inside the PSU, even a motherboard thief connector has been in the design I believe).
I have tried Martin’s aforementioned downvolting and it works just great. And there was no performance hit, might have been the way around in some tests even. With that kext in effect, you could run a Radeon VII without Pixlas mod: 2x mini 6-pin to 8-pin, 2x SATA-power to 6-pin (or 8-pin). These all could be onnected to an EVGA Powerlink, and with that to a VII. And it would work. Probably it would work great and not shutting down even at power peaks appearing at times either.
That would be called ”power balancing” between all the power sources the GPU draws juice from - except the PCIe power, it’s not in the combined balanced draw. PCIe would be the least significant of all of them, and the most moderate with almost any card there is on the market. But it helps with the math of course. A couple of old cards drew excess amounts from the PCIe slot, and it was a catastrophe back in days. I have read about that from MR posts, and from internet archives too, but luckily not experienced it myself. Those cards burned a lot of motherboards with too much power draw from PCIe slot.
ps.
With Martins downvolt kext and a dual Pixlas mod + the trick above, and one a little bent PCIe shield, I have run 3x VIIs in a cMP successfully for quite a some time (can’t remember how long) without any problems. Well, any problems I could regard to VIIs or their power draw at least. I still have that cMP, but I have jumped to other configurations for work already.
As mikas said, the problem isn't the total power draw, but how it draws the power.
e.g. a 250W GPU, draw 125W from mini 6pin A, another 125W from mini 6pin B, and 0W from the PCIe slot.
Then the shutdown protection will occur.
But another 300W GPU, draw 115W mini 6pin A, another 115W from mini 6pin B, and the remaining 70W from PCIe slot.
Then this card can run with dual mini 6pins only.
That "need PSU mod over 250W" is just a rough reference number. In reality, there is no GPU under 250W which draw zero power from PCIe slot, but completely rely on the aux power input. Therefore, as long as the card's TDP is 250W or below, then most likely can be powered by mini 6pins without any issue.
Straightly speaking, the PCIe card must has 225W TDP or below, and the aux pins must be less than single 8pin (or dual 6pin). But this is too limiting to get a good card for the cMP. Therefore, people tend to extend this limit a bit to certain range which is still safe in real world.
Sorry about slightly off topic, but just tested 2x VII, the other one downvolted and OC spoofed as a Vega II. Got 1000 points more in cinebench than without downvolting and spoofing. Don't know which one was the key to get it up that much. Maybe both.
It seems there's a bug in cinebench 2024. It only shows Radeon Pro Vega II once, but I certainly have selected them both to work on the bench. Maybe it's the OC spoofing of the other one of those GPUs, don't know.
Next I need to figure out how to assign the downvolt+spoof to PCIe slots 2 and 4. Trying to figure it out, but help needed desperately, please. Link: Today at 14:49
ps. by the way, this is a little over 600 pts (7%) more than Apple M2 Ultra top of the line 76 CU GPU.
It seems there's a bug in cinebench 2024. It only shows Radeon Pro Vega II once, but I certainly have selected them both to work on the bench. Maybe it's the OC spoofing of the other one of those GPUs, don't know.
Next I need to figure out how to assign the downvolt+spoof to PCIe slots 2 and 4. Trying to figure it out, but help needed desperately, please. Link: Today at 14:49
ps. by the way, this is a little over 600 pts (7%) more than Apple M2 Ultra top of the line 76 CU GPU.
Last edited:
My cMP 2009 although upgraded with 2 X5670s and 48GB of RAM has remained in prehistoric times when it comes to MacOS. Still running El Capitain 😬
I would like to give this beauty some love and update the OS. I don't really use it all the time, nor do I push it to extremes or need anything fancy (no gaming or major video processing needs or the like). I mostly work on my MBP these days. This said, I think it would be a good idea to use some fresher version of MacOS on this grandma.
The cMP currently has its stock GPU. I couldn't even go past High Sierra if I wanted to unless I now upgrade GPU. So here is my question and request for guidance. What GPU would fit these requirements:
- least amount of tweaking (no PSU mod) if possible;
- most compatibility with cMP;
- enable me to update MacOS as far a possible (Catalina / Big Sure / Monterey / Ventura ... am I asking too much?)
- if possible improved performance compared to the stock GPU (ATI Radeon HD 4870) but not really that important;
- has at least 1 DVI and or HDMI output;
- has boot screen or is flash-able (I would hate not to see that sweet apple logo). I see in the OP that many of these GPUs say "can be flashed only by MVC.". Does this mean I can't do it myself?;
- doesn't brake the bank
Am I asking too much from a 14 year old computer? 😙
Thank you.
I would like to give this beauty some love and update the OS. I don't really use it all the time, nor do I push it to extremes or need anything fancy (no gaming or major video processing needs or the like). I mostly work on my MBP these days. This said, I think it would be a good idea to use some fresher version of MacOS on this grandma.
The cMP currently has its stock GPU. I couldn't even go past High Sierra if I wanted to unless I now upgrade GPU. So here is my question and request for guidance. What GPU would fit these requirements:
- least amount of tweaking (no PSU mod) if possible;
- most compatibility with cMP;
- enable me to update MacOS as far a possible (Catalina / Big Sure / Monterey / Ventura ... am I asking too much?)
- if possible improved performance compared to the stock GPU (ATI Radeon HD 4870) but not really that important;
- has at least 1 DVI and or HDMI output;
- has boot screen or is flash-able (I would hate not to see that sweet apple logo). I see in the OP that many of these GPUs say "can be flashed only by MVC.". Does this mean I can't do it myself?;
- doesn't brake the bank
Am I asking too much from a 14 year old computer? 😙
Thank you.
Last edited:
Sapphire PULSE RX580 fits quite well.
It actually on the Apple's GPU list, you don't need to mod anything. And due to Apple use this particular GPU in their eGPU kit, which makes this GPU has native support by default.
It allow you to upgrade to Catalina and beyond (OCLP also make it can run on cMP in Ventura)
It's about 5x faster than HD4870.
It has both DVI port and HDMI ports.
It can be flashed by MVC. However, I personally recommend you flash the cMP, but not the card now. As long as you flash your cMP with EnableGop. A factory RX580 should able to show you the boot screen on cMP.
I don't know your budget, but RX580 isn't considered expensive for today's GPU price.
The biggest concern is that you cannot run this GPU in El Capitan. So, you have to plan your upgrade path properly. e.g. Upgrade to latest HS with you 4870, then swap in the RX580, and further upgrade to latest Mojave (to get the 144.0.0.0.0 firmware), then apply EnableGop to that 144.0.0.0.0 BootROM, then further upgrade to Monterey (latest macOS that can be loaded via OpenCore without any system file patching) or Ventura (need system files patching via OCLP)