MP All Models AMD Polaris, Vega, & Navi GPU macOS Support

[atonaldenim]

Hi,
Did you get your MSI Air Boost ?
If so, do you have any fan issue with it, or is this problem solved with 10.14.5 ?
If anyone else has the answers, don't be shy ^^

Thanks

Yes, the fan speed is normal when I install the MSI Air Boost Vega 56 OC into a Mac Pro 2010 running MacOS Mojave 10.14.6. When the system is powered on, the fan spins up a bit during the boot screen phase (no picture), but when the OS loads and the screen turns on, the fan stops completely. Even running Cinebench R15 GPU test did not make the GPU fan start spinning (Activity Monitor showed the GPU load ~50%). Only when running intense apps like Valley, Luxmark, Furmark did the fan spin at all, and I would say it was spinning moderately fast, not jet engine loud.

While I'm at it, I thought I would share some power consumption and performance benchmark screenshots. I've powered the Vega 56 with a dual 6-pin > 8-pin cable into an EVGA Powerlink. I don't know why Hardware Monitor shows 0 load on PCIe aux connector 1, guessing that's just a bad sensor.

I didn't run any of these stress tests for an extended period of time, but a few short 1 minute Furmark sessions didn't start any fires or make the computer crash, phew!

Vega 56 MacOS 10.14.6
Peak power consumption: (reported by Hardware Monitor)
214W Furmark (91.5W aux power x2 + 30.7W slot power)
209W Valley (93.9W aux power x2 + 20.8W slot power)

Idle power consumption: 31W

 

[Djeross]

Man, thank you very much for this detailed answer, couldn't be more thorough
I ordered everything to do the Pixla's mod, to avoid any power issue with any GPU I may put in my MP, so on this end I should be fine.
Thanks again !

 

[Pval]

What is the status of custom Vega PowerPlay Tables in 10.14.5, with its improved native Vega drivers?

Still using the tables without issue. Native 10.14.6 fan control is still so so, my fans won't spin up beyond 1500rpm, with a custom table they will spin up.
10.14.6 native drivers; fans spin up to 1500rpm then remain there, GPU starts to throttle.

10.14.6 tweaked PPT; fans spin up to 2100rpm, GPU frequency remains stable.

 

[zedex]

Advice NEEDED!!
I'm using the RX480 at the moment.

I'm looking to buy a SAPPHIRE AMD RADEON RX VEGA 64 (Reference) but I'm not 100% clear on the current issues I might need to address if I proceed with the purchase..

Based on my existing (and preferred) OS build (macOS 10.13.6 High Sierra), I've learned the following:

(1) I will need to do the PIXLAS mod to ensure the card is properly powered
(2) I can expect that the card will be properly identified in 'About This Mac' and 'System Information'
(3) I can expect that the card will run the GPU fans at nearly maximum speeds (is there a hack to fix this on Vega 64)?
(4) Hardware acceleration can only be accessed/implemented using the hack described by @h9826790

Can someone correct/clarify the statements above if they're wrong..?

Also - extra notes on my planned upgrade are very welcome

 

[krakman]

Buy a recerenc

Advice NEEDED!!
I'm using the RX480 at the moment.

I'm looking to buy a SAPPHIRE AMD RADEON RX VEGA 64 (Reference) but I'm not 100% clear on the current issues I might need to address if I proceed with the purchase..

Based on my existing (and preferred) OS build (macOS 10.13.6 High Sierra), I've learned the following:

(1) I will need to do the PIXLAS mod to ensure the card is properly powered
(2) I can expect that the card will be properly identified in 'About This Mac' and 'System Information'
(3) I can expect that the card will run the GPU fans at nearly maximum speeds (is there a hack to fix this on Vega 64)?
(4) Hardware acceleration can only be accessed/implemented using the hack described by @h9826790

Can someone correct/clarify the statements above if they're wrong..?

Also - extra notes on my planned upgrade are very welcome

If you inted to edit video and photos you can buy a reference vega 56
put bios switch on lower power mode
Then use a EVGA powerlink and just use the 2 mini 6 pin power adaptors.

Search tis forum for enable amd h264 encode decode and follow the instructions therein.

You can search my user name as we have discussed this thoroughly.

If you intend to play games then you will need to do the pixlas mod because the power draw can spike and cause a shut down.

 

[DanSilov]

(1) I will need to do the PIXLAS mod to ensure the card is properly powered
(2) I can expect that the card will be properly identified in 'About This Mac' and 'System Information'
(3) I can expect that the card will run the GPU fans at nearly maximum speeds (is there a hack to fix this on Vega 64)?
(4) Hardware acceleration can only be accessed/implemented using the hack described by @h9826790

Can someone correct/clarify the statements above if they're wrong..?

1. For Vega 64 — yes. But with so many examples and photos, it is much easier than it might sound.
2. Yes.
3. Works OOB for me without any issues with Vega 64. Latest versions of Mojave had some fixes for fan issues, as far as I know.
4. Hardware acceleration for HEVC video decoding needs this hack. HEVC encoding is not possible at all. Regular 3D acceleration works OOB, of course.

 

[zedex]

Buy a recerenc

If you inted to edit video and photos you can buy a reference vega 56
put bios switch on lower power mode
Then use a EVGA powerlink and just use the 2 mini 6 pin power adaptors.

Thanks for this recommendation @krakman - I really get the impression the Vega 56 is a better choice for me (no render or gaming requirement). I'm really only leaning in to the RX Vega 64 LE because of the beautiful silver enclosure. My side panel is permanently removed and I love the look of well-engineered PCIe components.

Also - how do I get my hands on a EVGA Powerlink these days..?

The biggest unclear question I have is really about the fans and whether or not - based on the final High Sierra build - I'm going to have to deal with excessive GPU fan noise discussed thoroughly at the beginning of this thread. Most of the problems described related to early HS builds and BETAs.. Most of the solutions described relate to the improvements made in Mojave.

In very precise terms - I have not been able to ascertain if there were any "fan problem" improvements made by the time the final High Sierra build was released.

PS - Good clear answers from you @DanSilov - I really appreciate it.

 

[orph]

EVGA Powerlink
https://www.amazon.co.uk/EVGA-600-P...VGA+Powerlink&qid=1568212556&s=gateway&sr=8-1

vega 56 is relay a good option, i think it's only 5% slower or so and uses less power = less heat = less fans

im not sure what support is at the mo, think it's ok but i use a RX 580 and that works super well

there are power play table hacks you can use to in osx to 'edit' the power use

 

[zedex]

EVGA Powerlink (THUMBS UP)

..there are power play table hacks you can use to in osx to 'edit' the power use

YEP - The power performance payoff puts the Vega 56 way out in front IMO. I've read about the powerplay table hacks - seems like an excellent way to dial down the TPD (although I don't understand why this hack can be applied to V56 but not V64).

At this stage - I'm pretty certain that I want to put the V56 board inside the V64 LE shroud - i don't have the budget for a move like this but it's the perfect solution.

 

[AlexMaximus]

YEP - The power performance payoff puts the Vega 56 way out in front IMO. I've read about the powerplay table hacks - seems like an excellent way to dial down the TPD (although I don't understand why this hack can be applied to V56 but not V64).

At this stage - I'm pretty certain that I want to put the V56 board inside the V64 LE shroud - i don't have the budget for a move like this but it's the perfect solution.

The idea is good but might not work because the PCB from the Vega 56 is totally different than the 64 in some cases.
There are the reference models that I would not recommend because of the really bad duct cooling fan design.
I used to have the regular 64, then the LE 64 version that you want, and now the Vega 7.
The LE looks great, but both Vega 64 versions got extremely hot in my case. Both cards developed an electrical smell of some sorte, so I sold both cards again. With that past lessons learned, I would not get any reference designed card at all. What I could read about those cards, the very best option and value would be the Powercolor Vega 56 Red Dragon version. This one used to have a shorter PCB and uses lower TPW. I have read somewhere that you can do this card even without the pixlas mod, -using only a EVG Powerlink and the low power bios. I do however agree in terms of aesthetics, the LE looks the best.

 

[zedex]

Very helpful feedback - can you comment on the high-speed fan issue @AlexMaximus ..? Did you ever use either of these GPUs on the final build of High Sierra..?

 

[AlexMaximus]

Very helpful feedback - can you comment on the high-speed fan issue @AlexMaximus ..? Did you ever use either of these GPUs on the final build of High Sierra..?

Sorry, I did rest on El Capitan a very long time and then jumped directly to Mojave. I did skip Sierre & High Sierra. However the very first version of Mojave was kind of similar on the fans I guess. I did not have super high pitch fan noise, more of a constant middle speed fan on the GPU. The more disturbing thing was the higher speed running pci boost fan for me, and that one somehow came from the higher power draw. The Apple fan speed system works not through temp sensors but in power monitoring. They somehow changed the PCI power draw settings in the firmware update, meaning in Mojave the PCI slot is allowed to get less power without impact on fans, and get more power through the dual 6 pin socket on the mainboard.
You need to speek with someone that has the Powercolor Red Dragon card on the HighSierra OSX to hear about his experience.
But as far as I could read about this, it seems this card belongs to the better bunch of V56, because it is still a 2 slot design together with a three fan design, meaning a better cooling overall. Read the recommendation on the early blog entry, there should be folks out there with this setup. By the way, best greetings to Australia, your country is truly awesome!

 

[MisterAndrew]

I wouldn't buy a Vega card to use with High Sierra. Get a Sapphire Pulse RX 580 and mod it to have the correct Apple dev kit part number so it works with the Orinoco frame buffer. I would recommend Vega for 10.14.5+ only.

 

[Pval]

Get a Sapphire Pulse RX 580

He’s already got a rx480 which is a rx580 running 100MHz slower... which he can easily flash or fix via kext...

 

[zedex]

He’s already got a rx480 which is a rx580 running 100MHz slower... which he can easily flash or fix via kext...

Thanks for the input from both @MisterAndrew and @Pval

The reality is that I will probably upgrade to a Sapphire Pulse RX 580 because my RX 480 Reference Card only has a 6-pin PCI-e power connector and I'm getting system resets (power cuts out without any warning) whenever I screen record on a 4K screen. It's annoying and (unfortunately) a flash or a kext won't allow me to feed the card more power.

MisterAndrew has outlined a strong case in favour of the Sapphire Pulse RX 580 (and maximum compatibility is rightly prized amongst cMP users). My biggest problems with the Sapphire Pulse RX 580 are as follows:
(1) it's a new kind of ugly - a heretofore unseen brand of ugly that is rivalled in ugliness only by previous generations of graphics cards released by an assembly of random 'brand partners' that NVIDIA and AMD have inexplicably blessed to handle product distribution.
(2) I don't know anything about flashing or GPU BIOS editing. Can I even make these sort of changes on a cMP..? or do I need to get my hands on a cheap Windows tower first before I even consider trying to 'flash' a suitable GPU..?

I've always been unclear about the initial system requirements on graphics card flashing. Some clarity on this would be appreciated.

 

[Ludacrisvp]

Thanks for the input from both @MisterAndrew and @Pval

The reality is that I will probably upgrade to a Sapphire Pulse RX 580 because my RX 480 Reference Card only has a 6-pin PCI-e power connector and I'm getting system resets (power cuts out without any warning) whenever I screen record on a 4K screen. It's annoying and (unfortunately) a flash or a kext won't allow me to feed the card more power.

MisterAndrew has outlined a strong case in favour of the Sapphire Pulse RX 580 (and maximum compatibility is rightly prized amongst cMP users). My biggest problems with the Sapphire Pulse RX 580 are as follows:
(1) it's a new kind of ugly - a heretofore unseen brand of ugly that is rivalled in ugliness only by previous generations of graphics cards released by an assembly of random 'brand partners' that NVIDIA and AMD have inexplicably blessed to handle product distribution.
(2) I don't know anything about flashing or GPU BIOS editing. Can I even make these sort of changes on a cMP..? or do I need to get my hands on a cheap Windows tower first before I even consider trying to 'flash' a suitable GPU..?

I've always been unclear about the initial system requirements on graphics card flashing. Some clarity on this would be appreciated.

You could get some more cables to power your card from both 6 pins on the motherboard.

 

[zedex]

Yeah - I've already done this. 2 mini6pins to one full size 6pin. It doesn't cut out until pushed now but it still cuts out (and I don't render or game). It was distressing at first - thought I might have a failing PSU on my hands but then I read a post somewhere that said the 8-pin trigger gives the PSU "permission" to draw 150W - the flip side of the argument is that it doesn't matter how much power I provide (i.e. 2 mini6pins) if the card physically constrains you to the use of one 6pin - it will only give the PSU "permission" to draw 75W.

I found this argument interesting and it seems to be true because the 'total available power' from 2 mini6pins has definitely not doubled (or PSU/GPU power cutout problem would be gone completely).

 

[tsialex]

Yeah - I've already done this. 2 mini6pins to one full size 6pin. It doesn't cut out until pushed now but it still cuts out (and I don't render or game). It was distressing at first - thought I might have a failing PSU on my hands but then I read a post somewhere that said the 8-pin trigger gives the PSU "permission" to draw 150W - the flip side of the argument is that it doesn't matter how much power I provide (i.e. 2 mini6pins) if the card physically constrains you to the use of one 6pin - it will only give the PSU "permission" to draw 75W.

I found this argument interesting and it seems to be true because the 'total available power' from 2 mini6pins has definitely not doubled (or PSU/GPU power cutout problem would be gone completely).

Some RX-480 use more than 75W from the PCIe slot even with AMD driver corrections.

 

[Pval]

(1) it's a new kind of ugly - a heretofore unseen brand of ugly that is rivalled in ugliness only by previous generations of graphics cards released by an assembly of random 'brand partners' that NVIDIA and AMD have inexplicably blessed to handle product distribution.

Well, some configs look nice, but I'm biased; IOCrest & Vega64 Liquid look the part

 

[zedex]

Well, some configs look nice, but I'm biased; IOCrest & Vega64 Liquid look the part

Dude - that is super clean and super impressive. I was having a shot at the littany of disgracefully designed gaming GPUs - not the beautiful reference cards. That's the first time I've seen a liquid cooled installation in a cMP. It looks AMAZING!!

Does it perform well..? I wasn't sure the cMP would even turn on if I removed the PCIe fans from the front middle section - have there been any temperature control issues...?

 

[MisterAndrew]

I don't know anything about flashing or GPU BIOS editing. Can I even make these sort of changes on a cMP..? or do I need to get my hands on a cheap Windows tower first before I even consider trying to 'flash' a suitable GPU..?

https://forums.macrumors.com/thread...x580-pulse-into-the-mac-edition-card.2101909/

 

[AlexMaximus]

Well, some configs look nice, but I'm biased; IOCrest & Vega64 Liquid look the part

Besides the obvious, nice cooling mod on the DT120's ... most people forget that aspect. I love this setup, really very clean & great pic.
Kudos to the Holland Geek Squad!

 

[cube]

AMD added Radeon Image Sharpening to RX470, RX480, RX570, RX580, and RX590 in DX12 and Vulkan.

 

[vadergr]

Hello All,

I am thinking of buying a RX560 card for my Mac Pro 5.1.
Probably the MSI RX 560 AERO ITX or a Sapphire one.

The procedure is just plug n play, just installing the new card and wait for the auto installation of new drivers from MacOS, or is there any other procedure i should pay attention to?

Thanks in advance.

 

[zedex]

I am thinking about buying a Radeon Pro WX 5100.. I'm puzzled by something though..

My RX 480 Reference Card only has a 6-pin PCI-e power connector and I'm getting system resets (power cuts out without any warning) whenever I screen record on a 4K screen.

Am I begging for trouble if I move to a Radeon Pro WX 5100 with no external power connector (and TDP of 75W)..?

The Radeon Pro WX 5100 specifications sheet suggests that it can handle my workflow requirements in its sleep (and @MisterAndrew suggests that a WX 4100 might even be enough for a user like myself).

BUT..... I simply DO NOT understand how a GPU with less power access could perform at or above the level of the RX480 6-PIN powered GPU I have installed at the moment..?

Also - @Pval if I wish to install this card in SLOT 3, will it get the full 75W power draw that it needs..? or will it be much lower because SLOT 3 is only x4 electrically active..?