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.
Dual Gpu in Mac Pro 5,1
- Thread starter AlbertoQuemada
- Start date
- Sort by reaction score
Nah i won’t go for 2 cards in the same cMP if i sell the 2 RX580 for a Frontier or RX5700 XT.
But to answer your question, hypothetically if i would go for 2 GPU in the same cMP it would be much easier with these 2 reference models because they are smaller and have no DVI port.
So i would just need to bend a little piece of aluminium on the back of the cMP to make the card fits into slot 4.
Exactly like this:
But to answer your question, hypothetically if i would go for 2 GPU in the same cMP it would be much easier with these 2 reference models because they are smaller and have no DVI port.
So i would just need to bend a little piece of aluminium on the back of the cMP to make the card fits into slot 4.
Exactly like this:
Attachments
Alright so i bought a new RX 5700 XT.
Now since i have 2, i’ll still want to test one of my cMP with both in it, just for the kick of it and some benchmarks.
I’ll have to make a double Pixlas Mod since the other cMP i got has a Dual 8 pins Pixlas Mod which will not be enough to power 2 (6+8 pins) cards.
Is a double Pixlas Mod is kindn of the same as a single, i just double the T-Taps on the PSU?
What should i be aware of?
Thanks guys.
Now since i have 2, i’ll still want to test one of my cMP with both in it, just for the kick of it and some benchmarks.
I’ll have to make a double Pixlas Mod since the other cMP i got has a Dual 8 pins Pixlas Mod which will not be enough to power 2 (6+8 pins) cards.
Is a double Pixlas Mod is kindn of the same as a single, i just double the T-Taps on the PSU?
What should i be aware of?
Thanks guys.
It's the other way around : All GPU manufacturers are mistaken .... (lazy saturday afternoon).
I'd double the T-splitters so: 2 cables going down each with 2X 6+2 pins.
Note that you may get problems passing 2 Pixlas cables + 1 SATA cable through the hole in the bottom of the optical drive bay...
I assume you will have to leave off the aluminium cover in the bay due to the bulky clutter of cables and perhaps the optical drive sled may not slide all the way in ...
Last edited:
How did you get the bootscreen to workI don’t have the magic of two Radeon VII but the configuration setup I have suites meAlso have the “bootscreen” option!
Also have the “bootscreen” option!
That's the Sapphire HD7950 Mac Edition card, has boot screen support for cMP from factory.
All 7xxx Mac EFI ROM out there are using the Mac EFI that extracted from this card.
Why make things so complicated?
Why not just install OpenCore to get the boot screen for RX580.
And keep the 5770 in a box safely as an emergency card.
It's not about PSU or slots count.
But why keep the 5770 at there to degrade the system stability (possibly, due to mixing GPU), and makes your cMP unable to install / run Mojave and newer macOS properly (due to 5770 isn't supported).
If just for boot screen, OpenCore or simply use rEFInd+ can achieve that already.
A fair question. It's system stability that I prize above all else.
I believe (but do not have proof) that when macOS boots up it calls the AMD Graphics package (kexts) if you have AMD GPU(s) installed OR it calls the NVDA Graphics package (kexts) if you have NVIDIA GPU(s) installed.
As long as the NVIDIA GPUs are pre-Maxwell (Kepler and earlier) and the AMD GPUs have featured in other Mac hardware releases, then I can expect Apple-engineered stability.
Mixing Nvidia and AMD GPUs is bad as we all know. This could be because macOS is calling 2 different Graphics packages (kexts) concurrently which causes kernel panics, IO conflicts and generally unreliable/unstable graphics performance.
Opencore, flashing graphics cards, installing third-party kexts all affect stability in ways that I can't easily diagnose.
I've got a pretty white Mac Edition HD 7950 in another cMP which I only bought because I forgot that the last time I owned one, it intermittently caused screen glitches 3 times an hour. This GPU is "Apple-blessed" (aftermarket) and it is truly a sub-standard GPU. The problem here is that if you flash this 'official' ROM to a cheaper (PC) release of the HD 7950 - you've still got rubbish.
So in summary - endless trial and error have led me to the most stable GPU solutions for native support with no sacrifices and the formula is simple: Pair an old and gutless Apple OEM GPU with a more modern powerhouse GPU of the same team.
MY FAVOURITE COMBINATIONS..
NVIDIA: GT-120 [SLOT4] paired with a GTX Titan [SLOT1] (or Quadro M4000 [SLOT3] if you are determined to keep both of the x16 slots free)
AMD: HD5770 [SLOT3] paired with a RX 580 / VEGA 56 [SLOT1] (WITHOUT a PCIe expansion chassis - CUBIX or NETSTOR - you WILL lose a 16 lane slot to an AMD GPU).
It's worth a thought
I believe (but do not have proof) that when macOS boots up it calls the AMD Graphics package (kexts) if you have AMD GPU(s) installed OR it calls the NVDA Graphics package (kexts) if you have NVIDIA GPU(s) installed.
As long as the NVIDIA GPUs are pre-Maxwell (Kepler and earlier) and the AMD GPUs have featured in other Mac hardware releases, then I can expect Apple-engineered stability.
Mixing Nvidia and AMD GPUs is bad as we all know. This could be because macOS is calling 2 different Graphics packages (kexts) concurrently which causes kernel panics, IO conflicts and generally unreliable/unstable graphics performance.
Opencore, flashing graphics cards, installing third-party kexts all affect stability in ways that I can't easily diagnose.
I've got a pretty white Mac Edition HD 7950 in another cMP which I only bought because I forgot that the last time I owned one, it intermittently caused screen glitches 3 times an hour. This GPU is "Apple-blessed" (aftermarket) and it is truly a sub-standard GPU. The problem here is that if you flash this 'official' ROM to a cheaper (PC) release of the HD 7950 - you've still got rubbish.
So in summary - endless trial and error have led me to the most stable GPU solutions for native support with no sacrifices and the formula is simple: Pair an old and gutless Apple OEM GPU with a more modern powerhouse GPU of the same team.
MY FAVOURITE COMBINATIONS..
NVIDIA: GT-120 [SLOT4] paired with a GTX Titan [SLOT1] (or Quadro M4000 [SLOT3] if you are determined to keep both of the x16 slots free)
AMD: HD5770 [SLOT3] paired with a RX 580 / VEGA 56 [SLOT1] (WITHOUT a PCIe expansion chassis - CUBIX or NETSTOR - you WILL lose a 16 lane slot to an AMD GPU).
It's worth a thought
You can argue OpenCore may destabilise the system, but so far, hundreds (if not thousands) cMP installed OpenCore, no such evidence.A fair question. It's system stability that I prize above all else.
I believe (but do not have proof) that when macOS boots up it calls the AMD Graphics package (kexts) if you have AMD GPU(s) installed OR it calls the NVDA Graphics package (kexts) if you have NVIDIA GPU(s) installed.
As long as the NVIDIA GPUs are pre-Maxwell (Kepler and earlier) and the AMD GPUs have featured in other Mac hardware releases, then I can expect Apple-engineered stability.
Mixing Nvidia and AMD GPUs is bad as we all know. This could be because macOS is calling 2 different Graphics packages (kexts) concurrently which causes kernel panics, IO conflicts and generally unreliable/unstable graphics performance.
Opencore, flashing graphics cards, installing third-party kexts all affect stability in ways that I can't easily diagnose.
I've got a pretty white Mac Edition HD 7950 in another cMP which I only bought because I forgot that the last time I owned one, it intermittently caused screen glitches 3 times an hour. This GPU is "Apple-blessed" (aftermarket) and it is truly a sub-standard GPU. The problem here is that if you flash this 'official' ROM to a cheaper (PC) release of the HD 7950 - you've still got rubbish.
So in summary - endless trial and error have led me to the most stable GPU solutions for native support with no sacrifices and the formula is simple: Pair an old and gutless Apple OEM GPU with a more modern powerhouse GPU of the same team.
MY FAVOURITE COMBINATIONS..
NVIDIA: GT-120 [SLOT4] paired with a GTX Titan [SLOT1] (or Quadro M4000 [SLOT3] if you are determined to keep both of the x16 slots free)
AMD: HD5770 [SLOT3] paired with a RX 580 / VEGA 56 [SLOT1] (WITHOUT a PCIe expansion chassis - CUBIX or NETSTOR - you WILL lose a 16 lane slot to an AMD GPU).
It's worth a thought
However, rEFInd+ won't inject anything. it's a pure boot loader to provide boot screen, not way to destabilise anything.
And macOS isn't work that way. HS still OK, but really, just OK to mix GPU. It's not about AMD or Nvidia, but mixing dGPU is can cause strange issues. AMD cannot mix with Nvidia is a Windows thing, not macOS. macOS load driver base on the device ID, even both 5770 and RX580 are AMD card, they use different drivers. And can cause conflict.
For older OSX, we can mix almost any GPU without adverse effect. But for recent macOS, mixing GPU have high chance to hit some strange bug (not necessary whole system crash, but can be as simple as preview shows transparent photos).
I am not 100% sure which kind of glitched you talk about for the 7950, if you mean a monetary shuttering on the screen, that's a known driver issue. Apple never fix that since this 7950 release.
The work around is simple, use a HDMI dummy or connect multiple monitors to the 7950, for the VRAM stay at 1250MHz, then there will be no more shuttering. This glitch only happen during VRAM speed up / slow down.
Anyway, if your current setup works flawlessly, there is no point to change it.
But if you want to go Mojave or higher, then IMO, it's time to ditch the "boot screen card".
Last edited:
Thanks for taking the time to reply.
I'm very encouraged by the high acceptance of OpenCore but the arrival of Mojave signalled the end of my interest in new macOS releases.
I recently bought a Macbook Air with Catalina pre-loaded and it's affirmed my reservations about the future direction of macOS. I'll probably get on board again in the 2nd Gen of 'apple silicon'.
Great tip regarding the HD 7950. I'll give it a try and see if it works on my setup. Cheers - David
I'm very encouraged by the high acceptance of OpenCore but the arrival of Mojave signalled the end of my interest in new macOS releases.
I recently bought a Macbook Air with Catalina pre-loaded and it's affirmed my reservations about the future direction of macOS. I'll probably get on board again in the 2nd Gen of 'apple silicon'.
Great tip regarding the HD 7950. I'll give it a try and see if it works on my setup. Cheers - David
Hello, I do video editing. I have a similar combo (GTX980 slot 1/GT120 slot 2). How do I know which card is being used for what? In davinci, Premier and AE I can assign GPU but not in FCP7. In the software where I cant select the GPU, how do I know which GPU is being used on that software? I can do the opencore and all that but I like the GT120 because of the minidisplay port I can use directly to my Apple cinema display without frankeinsteining adaptors. Another weird thing is that in ABOUT THIS MAC, I can see the 120 but not the 980. However if I further click on “system report” I can see both cards in the list. I surely dont want to be using the 120 in FCP or Premier. This is my first time using dual cards. What card is doing what? Is PCI 1 the main? Or is PCI 1 exactly the same as 2? I know 3&4 are x4 so those are different. Or are they? Why are some of you installing a GPU on a x4 slot? Does that mean I could install the 120 in slot 3 for a bit of airflow to 980? The way I have it is not covering the fan though.A fair question. It's system stability that I prize above all else.
I believe (but do not have proof) that when macOS boots up it calls the AMD Graphics package (kexts) if you have AMD GPU(s) installed OR it calls the NVDA Graphics package (kexts) if you have NVIDIA GPU(s) installed.
As long as the NVIDIA GPUs are pre-Maxwell (Kepler and earlier) and the AMD GPUs have featured in other Mac hardware releases, then I can expect Apple-engineered stability.
Mixing Nvidia and AMD GPUs is bad as we all know. This could be because macOS is calling 2 different Graphics packages (kexts) concurrently which causes kernel panics, IO conflicts and generally unreliable/unstable graphics performance.
Opencore, flashing graphics cards, installing third-party kexts all affect stability in ways that I can't easily diagnose.
I've got a pretty white Mac Edition HD 7950 in another cMP which I only bought because I forgot that the last time I owned one, it intermittently caused screen glitches 3 times an hour. This GPU is "Apple-blessed" (aftermarket) and it is truly a sub-standard GPU. The problem here is that if you flash this 'official' ROM to a cheaper (PC) release of the HD 7950 - you've still got rubbish.
So in summary - endless trial and error have led me to the most stable GPU solutions for native support with no sacrifices and the formula is simple: Pair an old and gutless Apple OEM GPU with a more modern powerhouse GPU of the same team.
MY FAVOURITE COMBINATIONS..
NVIDIA: GT-120 [SLOT4] paired with a GTX Titan [SLOT1] (or Quadro M4000 [SLOT3] if you are determined to keep both of the x16 slots free)
AMD: HD5770 [SLOT3] paired with a RX 580 / VEGA 56 [SLOT1] (WITHOUT a PCIe expansion chassis - CUBIX or NETSTOR - you WILL lose a 16 lane slot to an AMD GPU).
It's worth a thought
Attachments
Last edited:
Is there a benefit in running both styles / vintage cards? Typically like cards offer better performance in applications that utilize such hardware, though my understanding of how the 5,1 handles Multi GPU generations is quite limited.
I only use the 120 for that minidisplay port that works great with my Apple Cinema Display. So that would be one benefit. Also, When I restart the computer having only the 980 installed, I don’t get the boot screen it looks like the Nvidia web driver resets to macOS driver. So with the 120 installed on top of 980, I always get that boot. I just want to know which card is doing what but for some reason, nobody seems to know this answer. Ive tried all kinds of facebook groups and forum searches. Its a complete mystery. I hope someone on here has an answer.
1. Which card is doing what if the software doesnt allow GPU selection?
2. why is the 980 not showing in “about this mac” but it does shows in “system report” list.
3. Are PCI SLOT 1, exact same as PCI SLOT 2? or does mac default to either as a priority?
4. Can I install the 120 in slot 3 being a x4 PCI?
Last edited:
1) Depends on the software. Therefore, no one can answer your question apart from the developer.
But in general. If the software can utilise multiple GPU, then all GPU are in use. Otherwise, the GPU that connected to the monitor which displaying the software is in use.
2) Because you installed both GPU, and GT120 is higher priority
3) Same. In your case, no.
4) Can
Not necessarily, some apps will happily use different cards but FCP only likes perfectly matched pairs of GPUs.
When I had a W5700X and 5700 XT in my 7,1 it was one or the other.
FCP can use multiple uneven GPU, I tested that myself.
I suspect that you were using H264 / HEVC hardware acceleration. In that case, no matter how many GPU installed, still only one hardware video engine can be used.
Im using FCP7 which I prefer over the x version. I dont see anywhere an option of selecting a GPU or an option where it shows which GPU is in use.
I was looking for some infos about dual CPUs in the cMP4,1/5,1 and it seems I found it.
But still, I have a question:
I am running my cMP 4,1 (flashed to 5,1) 12-Core on a Vega Frontier Edition. Pixlas Mod, OpenCore (0.7.9 currently), NVMe-SSD and many other exciting things. I have MacOSX Monterey for video and foto editing and Win10 for gaming.
In MacOS everything is fine, my editing does not concerne the Vega FE to much, but in Windows for the gaming it is getting quite noisy. I was trying a RX6900XT but, of course, hit a wall with Navi2 (still waiting for a wonder to happen, so it will still work on cMPs). So I got back to the FE and was thinking about putting another one in, to have more power for the gaming.
I do know, that dualGPU and gaming always has some flaws and you don´t get double the power by doubling the GPUs, but still I´d like to know, if someone has some experience with it? Does it work and improve the fps?
All the time I´m still waiting for the FE with watercooling, but there are close to none on the market.
Greets from switzerland 🏔️🗻
But still, I have a question:
I am running my cMP 4,1 (flashed to 5,1) 12-Core on a Vega Frontier Edition. Pixlas Mod, OpenCore (0.7.9 currently), NVMe-SSD and many other exciting things. I have MacOSX Monterey for video and foto editing and Win10 for gaming.
In MacOS everything is fine, my editing does not concerne the Vega FE to much, but in Windows for the gaming it is getting quite noisy. I was trying a RX6900XT but, of course, hit a wall with Navi2 (still waiting for a wonder to happen, so it will still work on cMPs). So I got back to the FE and was thinking about putting another one in, to have more power for the gaming.
I do know, that dualGPU and gaming always has some flaws and you don´t get double the power by doubling the GPUs, but still I´d like to know, if someone has some experience with it? Does it work and improve the fps?
All the time I´m still waiting for the FE with watercooling, but there are close to none on the market.
Greets from switzerland 🏔️🗻