It's a 180W card. There's another model for external use that's 150W. Neither one is suitable for using in the 5,1 without modifying the PSU. The peak current in Vega GPUs requires much beefier power supplies.
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 All Models AMD Polaris, Vega, & Navi GPU macOS Support
- Thread starter MisterAndrew
- Start date
- Sort by reaction score
It's a 180W card. There's another model for external use that's 150W. Neither one is suitable for using in the 5,1 without modifying the PSU. The peak current in Vega GPUs requires much beefier power supplies.
Hello everybody.
I installed the AMD Sapphire Nitro + RX580 8GB Special Edition (the blue one) in my Mac Pro (HS 10.13.4).
By booting the LuxMark bench the fans of the Mac Pro that increase in revolutions are those of Expansion Slots and Power Supply .... is it normal or should they stay at the default speed?
The LuxMark bench is 12913.
I noticed that with the old ati 5770 the fan that increased rpm with luxmark bench was only that of the Expansion Slots and not that of the Power Supply.
I installed the AMD Sapphire Nitro + RX580 8GB Special Edition (the blue one) in my Mac Pro (HS 10.13.4).
By booting the LuxMark bench the fans of the Mac Pro that increase in revolutions are those of Expansion Slots and Power Supply .... is it normal or should they stay at the default speed?
The LuxMark bench is 12913.
I noticed that with the old ati 5770 the fan that increased rpm with luxmark bench was only that of the Expansion Slots and not that of the Power Supply.
Last edited:
Asus RX570 OC 4GB: Power Supply Fan just about 900 RPM then card was loaded by extreme graphics, and that's after 10-15 minutes maybe with Unigine engine to make sure there is a good load.
In LuxMark the card didn't draw quite that much power than in Unigine. Watts stood just around 70W all the time, and fans stayed at around 800RPM:
In Luxmark PS RPM gets up to 818 (after two minutes) if I test both RX580+X5680.
In Luxmark PS RPM gets up to 807 (after two minutes) if I test only RX580.
You can read the watts for CPU and RX570 from the tables.
I suppose it should depend whether you are testing multiple openCL devices, or just one. The more devices, the more watts consumed and the higher the RPMs. As we can read from my measurements, the difference is quite minimal in LuxMark - at least in my case it is.
By the way, I used only one mini 6-pin to 8-pin with RX570, and it worked just fine. You can see there was almost 90W drawn from a single mini 6-pin motherboard connector (The spec says 75 W). I ran out of cables because I bought a new card (Sapphire RX580 Pulse 8GB), and just wanted to test this scenario by myself too. I don't know if this was to have any effect to my measurements and their usefulness.
FYI: I'm gonna use 2x6-pin to 8-pin from now on, just received the new cable.
In LuxMark the card didn't draw quite that much power than in Unigine. Watts stood just around 70W all the time, and fans stayed at around 800RPM:
In Luxmark PS RPM gets up to 818 (after two minutes) if I test both RX580+X5680.
In Luxmark PS RPM gets up to 807 (after two minutes) if I test only RX580.
You can read the watts for CPU and RX570 from the tables.
I suppose it should depend whether you are testing multiple openCL devices, or just one. The more devices, the more watts consumed and the higher the RPMs. As we can read from my measurements, the difference is quite minimal in LuxMark - at least in my case it is.
By the way, I used only one mini 6-pin to 8-pin with RX570, and it worked just fine. You can see there was almost 90W drawn from a single mini 6-pin motherboard connector (The spec says 75 W). I ran out of cables because I bought a new card (Sapphire RX580 Pulse 8GB), and just wanted to test this scenario by myself too. I don't know if this was to have any effect to my measurements and their usefulness.
FYI: I'm gonna use 2x6-pin to 8-pin from now on, just received the new cable.
Attachments
I bet pretty much everything I have it's never going to be Pcie..
Would not make sense for them
I updated my original post concerning my experience with the Reference Sapphire RX Vega 56 flashed with Sapphire Nitro+ RX Vega 56 BIOS. https://forums.macrumors.com/thread...-sierra-support.2083168/page-22#post-26060072
Most important new discovery: I mentioned a kind of full-screen glitch previously. It’s gone with another monitor! So it was not the cards fault.
Power draw is still a problem, though. I will try to get additional power from all four HDD SATA ports next. If that‘s no good I‘ll have to do the Pixla mod.
Most important new discovery: I mentioned a kind of full-screen glitch previously. It’s gone with another monitor! So it was not the cards fault.
Power draw is still a problem, though. I will try to get additional power from all four HDD SATA ports next. If that‘s no good I‘ll have to do the Pixla mod.
If you look at the card while in the Mac, in which “BIOS switch position” have you flashed the Nitro bios?
And does it matter?
I flashed the BIOS to the position facing the display connectors. I think that was the primary position. I used the secondary BIOS from the Nitro+. So that way I kept the original secondary BIOS in the position facing the front of the computer.
I have no idea if it makes a difference. I bet it doesn‘t, but it’s maybe worth a try to do it the other way around.
I have no idea if it makes a difference. I bet it doesn‘t, but it’s maybe worth a try to do it the other way around.
I don’t think pulling power from the SATA ports is a good idea.
I feared that you or h9 would say something like that
My current plan is to make two double-Y contraptions.
Imagine taking two sata ports, combine them to one 6pin, use that one and combine it with one of the onboard 6pins to form one 8pin. same for the second 8pin.
I thought that way each of the 8pins gets a little more headroom to handle the demand.
Sure, I have some fear that this might damage something and the load flows „the wrong way“. But shouldn‘t the load just distribute itself between all the wires? Sorry if I sound foolish.
Anyone else having overheating issues with VEGA FE? When playing games at high settings or running fur mark the card simply overheats and shuts down. The GPU heat sensor doesn't talk to the cMP and spin up the fans. Any way to active this? I have had to manually turn the PCIE fan on to 4500 RPM, and the GPU is still so hot it could cook an egg.
[doublepost=1527404937][/doublepost]
[doublepost=1527404937][/doublepost]
I don't think that using SATA power is a problem. Man cMP users do this, but just realize you might have to give up a lot of HDD slots. For example Vega FE draws up to 300, and 400 spikes. The PCIE cables on the cMP apparently provide up to 120 each, for a total of 240, so you would have to link at least two HDDs SATA ports to provide enough power. Better to use an external PSU or do PIXLAS mod.I feared that you or h9 would say something like that
My current plan is to make two double-Y contraptions.
Imagine taking two sata ports, combine them to one 6pin, use that one and combine it with one of the onboard 6pins to form one 8pin. same for the second 8pin.
I thought that way each of the 8pins gets a little more headroom to handle the demand.
Sure, I have some fear that this might damage something and the load flows „the wrong way“. But shouldn‘t the load just distribute itself between all the wires? Sorry if I sound foolish.
This “power available” only true if you know how to draw it. The mini 6pin won’t automatically pump 120W to the card. E.g. if you connect dual mini 6pin to single 8pin, then each mini 6pin should only deliver up to 75W.
It’s almost impossible for a modern GPU shutdown due to overheat. The GPU can heavily throttle itself to avoid thermal shutdown. And due to the throttling, you should able to feel significant performance drop before thermal shutdown.
For sudden shutdown, I suspect your way to power the card doesn’t work well, and the card draw too much from one of the source which makes them cMP SMC command a shutdown to protect the logic board.
Here is the funny part- the computer doesn’t shut down, just the gpu powers off. The card is connected to a 850w psu running only the GPU. Never seen anyone else mention this problem, which is why I am so curious.
Complete cMP shutdown has happened to me with the FE. Pixlas mod solved the problem. It’s gotten hot but never overheated or shutdown by itself (while the Mac stayed on).
Today I tested what I had in mind. I‘m actually using all four HDD SATA connectors on the mainboard.
I merge HDD1 + HDD2 to one 6pin and merge this 6pin with one of the PCI-E 6pin connector from the mainboard to form one 8pin. Similarly I merge HDD3 + HDD4 to another 6pin and merge this one with the second PCI-E 6pin connector from the mainboard to get the second 8pin.
Result is that the card is running absolutely stable now. Maximum power draw if I add all absolute spikes together (not taking into account that it is unlikely that these spikes happen on all wires at the same moment) is 22A at 12V = 264W
I had 2,5A maximums at each of the HDD 12Vs. And 6A maximums on the PCI-E connectors.
So here we go, I‘ve got a Reference Sapphire RX Vega 56 in a Mac Pro. Flashed with the Sapphire Nitro+ RX Vega 56 BIOS. Powered without Pixla mod. And it‘s working flawlessly now.
Glad I bought that thing for 500€ when it was new, and stuck to it until this point.
Thank you all for your valued input!
EDIT: some pics
Inside it's admittedly not so pretty anymore.
Unigine Heaven completes it run-through now.
iStat while running Luxmark:
Last edited:
How interesting. I'm glad it's working out. Although it looks like it's taking about the maximum power from those HDD SATA ports so I don't think it's exactly the best solution.
The Vega 56 is working well in 10.13.6 beta 1, except the HDMI audio driver isn't loading like in 10.13.5. Can't tell if there are any changes with the Vega driver in this release. However, there is a change in the Polaris driver. It lists a new card with model number 113-C94002A1XTA and a corresponding "Yelcho" framebuffer. It lists Radeon Pro 580 after that. A Google search didn't turn up anything for that card so perhaps it hasn't been released yet. Perhaps it's in a new iMac or MacBook Pro.
Does anybody else have problems with their rx580 running unreasonably hot at idle? Mine revs its fans to full every minute and a half. It's extremely annoying. The heatsinks are quite warm to the touch, just sitting on idle. I'm considering an RMA, but the more I read, the more it seems likely that it's just Apple's crappy drivers and a replacement card isn't likely to be any different.
It's ironic because I opted for this card over a 1060 because I didn't want to deal with driver issues.
It's ironic because I opted for this card over a 1060 because I didn't want to deal with driver issues.
I stick to AMD cards for many years, due to prefer OOTB support. Then I realise...
Yes, it's handy. But if Apple provide bad driver, then nothing we can do about it. It actually happened on the HD7950 Mac Edition card.
10.9 OpenGL performance bug
10.10 Safari freeze (or micro freeze)
10.11 Finder freeze
From memory, 10.12 finally no major issue, but the annoying shuttering bug still there (this bug exist since this card launch back in 2013). In fact, I tested this card few months ago with High Sierra, the shuttering bug still there.
Finally, I switch to Nvidia 1080Ti. Compare to AMD, it has much less room for errors. If I accidentally perform a PRAM reset, all I have is just a black screen on the next boot. I even can't access recovery partition. If fact, if there is anything wrong during OS upgrade, it can be very messy.
But once the OS is installed, then I have the choice to pick the good driver version to use.
It actually happened back in 10.12.3. The V387 Nvidia web driver may shows colour error if transparency is ON. But I can roll back to V378 driver and be a happy user again. (Nvidia eventually fixed the bugs after few weeks)
If this kind of error happen on AMD card. I can either wait for next upgrade, hope Apple will fix it. Or downgrade the whole OS, which means recover from the backup. And the worst is, Apple never care about this kind of bug report. On the other hand, Nvidia almost aways able to address the bug correctly within few weeks (just for bugs, but not necessary for other request).
And now, since I have a spare PICe slot. I simply install a GT120 at slot 4 to be the "Mac EFI". So that I don't even need to swap card if anything goes wrong. But just switch to the GT120 output.
On the other hand, I still have a stock Apple HD4870 as backup for emergency use.
Sadly, it's starting to seem like maybe the smartest move is to buy a Windows license and be done with Apple. There doesn't seem to be any way to win anymore.
That's my current project now. I am switching to Hackintosh. Just leave myself the ability to use macOS, but more focus on Windows.
I prefer macOS mainly because my wife is an Apple person. But even she want to switch to Microsoft Surface Studio now. If she don't need macOS anymore. May be it's really time to leave macOS.