Gone are the days of the Mac sitting in the PC’s shadows of video render performance in Adobe Premiere Pro. I switched to PC a few years ago after tiring of the long render times of 4K video on the Mac. We wanted to stick with Premiere Pro because it was cross-platform. Switching to Final Cut Pro would mean we would always have to be on a Mac as we often access old projects for our videos and video work we do for clients.
Recently, I purchased a 2018 MacBook Pro with the i9 processor, 32GB of RAM, and a 1 TB SSD with hopes that connecting an eGPU to the Mac would give me comparable performance to a PC laptop. Once I learned that the Mac could support multiple eGPUs connected to its four Thunderbolt 3 ports, I had to try connecting as many as I could. The results surpassed what I could have hoped for beating out my $4200 Ryzen Threadripper PC build.
This video is that test.
Gigabyte RX 580 Gaming Box eGPU: https://jer.fyi/rx580gamingbox
Here are the results of rendering out the above video (17:32 length – YouTube 4K Ultra HD Render Settings) on all configurations shown in the video:
MacBook Pro No eGPU: Coming soon
MacBook Pro 1 eGPU: Coming soon
MacBook Pro 2 eGPUs: Coming soon
MacBook Pro 3 eGPUs: 19 minutes 32 seconds
Ryzen Threadripper 1950x PC: Coming soon
Threadripper PC Parts listing coming soon.
Hey what’s up, it’s Jerad with State of Tech, and today I’m here for scientific purposes basically. I have a MacBook Pro, and I’ve really been messing around with these eGPUs, trying different things, trying to see if I can get the kind of performance out of my MacBook Pro that I’m able to get out of a tower, like a full computer, whether it be a PC or maybe even a more powerful Mac.
What originally started this experimentation, is that I had a eGPU for my MacBook Pro. So this is my MacBook Pro 2018 edition, this is the i9, it’s the fully maxed out edition except for the hard drive, I only went with the one terabyte on that.
These here are eGPUs from Gigabyte. These are RX 580 eGPUs. They come as a kit already put together with a cool little carrying case, and they’re relatively small.
Now prior to this, I had a eGPU that was a razor core version two. Great eGPU, very well built, but it was just a little big and being on my desk I just wanted something smaller. So I thought, well I’ll give these a try just to see if there are any performance difference. I was able to determine that there was very very little performance difference between the razor core. You would think maybe the razor core because it’s bigger, and you can fit a … This is just a single fan GPU, it’s not a very big GPU, that maybe the bigger of the GPUs would perform better. I was only really able to get a couple of seconds of better performance out of the razor, and that could go either way. That’s not even definitive.
What really got me thinking, was when I took one of these. I got one of these, and I plugged it in to my MacBook Pro along with the other eGPU. Now I had read online in a forum that the MacBook Pro will support up to four GPUs being plugged into it. And I thought to myself, well why would they do that unless there was a purpose for performance, like maybe either with gaming or with video editing.
So for me video editing really is what I care about most, and so we’re going to do some tests today. Now I have the MacBook Pro sitting here. I have a project that’s on my hard drive, and I used this project as a kind of a benchmark, it’s not a very long project, but I know that the results are going to speak for themselves.
I’m going to go ahead and open up this project, which this is actually going to be a little different of a test for me because the last time that I ran this test was about a week ago, and Adobe dropped their 2019 edition of their software of their Creative Cloud, and there’s actually even been some improvements made in Premiere Pro for the Mac specifically. They said there’s lots of performance boosts that they’ve made, and I’ve noticed this. I’ve edited a couple of projects earlier this week, and I can tell that there is some performance benefits to upgrading to the latest version, the 2019 edition.
I’m going to go ahead and kind of just get things set up here, and we’re going to jump back in. We’re going to start by running this test, rendering out this project, which is about a minute and forty in length. And we’re going to render that out just on the laptop itself with no power hooked up to it. Then we’re going to hook power up to it. Then we’re going to hook one eGPU up to it, two, and then three eGPUs, and we’re going to test this out and see what the difference is.
Now like I said earlier, it could support up to four eGPUs, but the reason I’m only going to run three is because I still need one of my ports, one of my USBC ports to go the monitor. I run this little adaptor here, which is a USBC to display port adaptor to my monitor. I found that when I plugged in my monitor to the eGPU that there was just weird performance issues and weird bugs with the system. When it came to disconnecting the eGPU from your Mac, there was weird issues there. And then of course when I was rendering and doing things that required the performance of the eGPU, the performance was de gradated because the eGPU had to also run the monitor.
My decision was, I would rather run three eGPUs and have my monitor going out, than have to use one of the eGPUs to power the monitor. Because then you can disconnect all three eGPUs and unplug them even, and because your monitor is connected to a separate USBC port I haven’t had any issues there. Who knows, maybe the MacBook Pro will have some updates later that will work the issues out of disconnecting an eGPU when the monitor is connected to it. But right now I find that running the monitor direct into the laptop where the monitor is being powered by the internal processing, the internal GPU that the MacBook Pro has, is the best option and then just using the eGPUs as power sources to increase performance.
All right, so I’ll be right back. We’re going to get set up here, and start rendering out this project.
Okay, so the first render just finished that’s under the MacBook Pro’s power by itself without any power plugged into it. Again, it’s a minute forty footage and it took three minutes and thirty one seconds to render out. So that’s quite a bit of time. Consider maybe rendering out a project that’s thirty minutes long, and that’s more than about more than double the time almost. So that’s twice the amount of time that you’re going to have to wait for a 4K project to render out.
So we’re going to go ahead and plug in our first eGPU over here. So I’m going to plug that in like so, and now we have power actually going to the MacBook Pro as well, and the fan has just spun up a little bit there. I’m going to go ahead and open Media Encoder back up, and we’re going to render out the exact same project but with one GPU. Let’s check that out.
Okay, so the first run with the first eGPU has finished. One of the interesting things that came as a change with Premier Pro and the 2019 edition here, is that the default in Media Encoder is set to GPU acceleration with Metal, that is of course Apple’s framework or platform that they’ve developed for their own systems. That’s something that’s not available on PCs and even older Macs, but OpenCL is typically what we would use. If you’re on a PC you would use the CUDA cores of your device, whatever GPU it is that was in your system.
But I had an interesting finding here, and that’s with rendering out in Metal with on eGPU is actually a couple seconds slower than the MacBook Pro was, even unplugged with no power going through it. So using Metal as the acceleration, the renderer, three minutes and thirty one seconds with the MacBook Pro unplugged, three minutes and thirty three seconds with one eGPU plugged in. But what I did was go ahead and set it to OpenCL just to see what we’re going to get, and it’s almost finished here and I think this is where we’re really going to see the performance.
I’m guessing that these GPUs are not optimized for Metal. They don’t operate that way, or something about them. That’s something I’m definitely going to have to research, because I’m interested.
Right now this project just finished, so with OpenCL and one eGPU, let me go into the log and I’ll make sure to show all of you the log as well. The log shows two minutes and fifty eight seconds. So that’s that kind of performance that I was looking for. So I’m going to notate that, two minutes and fifty eight on OpenCL, because that matters.
The renderer that you choose, that actually does matter. That’s something that I am now finding out, and it’s interesting information. So I’m going to go ahead and reset this render project. I’m going to quit out of Media Encoder. We’re going to plug in our next eGPU here. So we’ll go ahead and plug that in, and that should start spinning up here in a second and it will show now two eGPUs in that eGPU drop down that’s up in our menu bar.
So I’m going to go ahead and open Media Encoder back up. So what I’m going to do for this test here, is I’m going to show you all of the results for both rendering out using Metal, and using OpenCL. That way we can kind of see the difference, because obviously using Metal by itself with the Mac with no eGPU is better than using one eGPU under Metal. Of course if we go to OpenCL, then we save, what was that, we saved thirty three seconds essentially there on a render that’s only a minute forty long. So like I said before, imagine what that would be like on a project that’s like a half hour long.
So with two eGPUs I’m going to switch back to Metal. I’m going to render this out. So let’s go ahead and clear that out, and we will render this file out here, hit start. And we’ll see what kind of performance we get on Metal, and then I’ll switch over to OpenCL. We’ll render out with OpenCL. Then we’ll move over to our third GPU and see exactly what we had happen. So I’m going to go ahead and fast forward until we get through this next render, and I’ll check back in with you.
Okay so the numbers are really starting to speak for themselves. Now with two eGPUs, we saw a little bit better time rendering out in Metal with three minutes and seventeen seconds, but that really wouldn’t be that big of a jump considering you’d have to buy two eGPUs and you’re looking at like $800 or $900 of an investment right there. But when switching to OpenCL, it jumped down to two minutes and forty six seconds. That’s what I was seeing when I was doing this test initially with one of these Gigabyte eGPUs and the razor core GPU with the RX 580 that I had in it.
So now we’re going to add in a third one. So we’re going to see if performance can just continue to stack. I’m going to take the USBC cord for the third eGPU, and we’re going to plug that in here. Now adding three eGPUs to my MacBook Pro, and of course if I didn’t have this hard drive plugged in over here I could add a fourth and then just edit or use the screen of my MacBook Pro, but I do like to have that additional port there so that I can plug in an external monitor.
So we have three GPUs plugged in to the MacBook Pro, now that’s pretty crazy. I don’t even do this with a normal PC, because the performance doesn’t actually improve that much when you start stacking GPUs in a PC. So let’s go ahead and open Media Encoder back up. And after I get done with this test, I’m going to go back and render out this project in OpenCL using the MacBook Pro with nothing plugged into it, so that I can have that baseline OpenCL and Metal.
You’ll remember we started out with three minutes and thirty one seconds on Metal, adding one eGPU we actually lost a little bit of time, but adding a second eGPU we got about a little more than ten seconds of an improvement on that one minute and forty second file. So now we’re going to test out three eGPUs, because I want a big reveal at the end I’m going to start with Metal as the renderer, and then I’ll do OpenCL as the last test here with all these GPUs. So I’ll be back in a minute, let’s check it out.
All right, so my testing is complete. The numbers are pretty interesting. So with the MacBook Pro unplugged, just to recap here, rendering out under Metal it was three minutes and thirty one seconds, OpenCL was actually faster at three minutes and twenty six seconds. I thought that since maybe Premiere listed Metal as the first option now that that was actually going to be faster, but it turned out that it wasn’t. It was a little bit slower on this project.
Then adding in those initial GPUs, we didn’t see any real good performance gains except for OpenCL. We went from three minutes and twenty six seconds MacBook Pro only, to two minutes and fifty eight seconds. Then adding in another GPU, we got that down to two minutes and forty six seconds. So really only about twelve seconds of savings there, not really that much under OpenCL. Obviously the Metal numbers are not the numbers that we’re really going to be paying attention to, because they’re just higher numbers.
But then we added in the third GPU, the third eGPU, we dropped down to two minutes and three seconds under OpenCL. Keep in mind this file was a, this timeline was a minute and forty. So a minute and forty, and it took two minutes and three seconds to render out, that’s not too bad at all.
Now I typically would render out using my Threadripper PC that I built. My Threadripper PC has the first generation Threadripper 16 Core 32 Thread, it has 64 Gigabytes of RAM. It has a 1080 Ti GPU in it, and MVME hard drives inside of it. It’s an extremely fast computer, and that rendered out the project in two minutes and thirteen seconds with a monitor connected.
Now I say with the monitor connected because that 1080 Ti also had to power the monitor in order to render that out. When I disconnected the monitor and then just remoted into the computer, it actually rendered out that project in two minutes.
So I think the whole moral of the story here, is that rendering out on a Mac is very possible using Premier Pro and having a lot of power, and getting good render times. No more are the days where you have to go to a PC to get good render times. Now granted this is kind of a crazy setup here. By the time that you pay about $4000 for a MacBook Pro like this and then you pay $400 for each, so this is $1200 in GPUs sitting right here, you’ve got quite a bit of money wound up in this project. However, you can always take your laptop disconnected from these and go.
You can also take one of these with you easily because they come with a cool little carrying case. However, with my PC that is basically stationary. I’m not going to pick up my whole tower and carry that around with me as I work. And then I need a monitor, and a keyboard, and a mouse, and all that stuff, and it becomes quite a bit of a problem.
So also, with the PC I typically also would have a laptop. So I have about $4000, about $4200 wrapped up in that PC not including a monitor, and then if you add in buying a laptop for maybe $2000 you’re still ending up with a lot of money wound up in computers.
I think what I’m trying to say here, is that the MacBook Pro with eGPU support, and this is still early on I’m sure there will be more optimizations to come as Apple adds more ability to connect different GPUs and whatnot, and these GPUs start officially supporting Mac OS there will just me more and more performance gains there.
But OpenCL, three GPUs, connected to a MacBook Pro, out performed my Threadripper PC that has a GTX1080 Ti in it, and that is what’s going to make me sleep great tonight because that’s just such a cool thing to have happen in a world where Mac’s typically just don’t perform well in Premier Pro. Those days are over, and that was the job here that I was trying to achieve in plugging all of this stuff in.
It makes me really interested to see what happened if I plug in a fourth eGPU, but I had to buy all of these myself. I’m not sponsored by anybody, Gigabyte did not send these to me, but I’m sure they’re going to be pretty impressed with the results that I had that I show in this video.
Thanks so much for checking out this video. If you are a Mac enthusiast and you like to edit video, Premier Pro is once again something that you can use and of course have your productivity back. Use your Mac, because you’re not just going to have to let it sit there and render out videos all day anymore. I’m super excited for that.
So click on that subscribe button to this channel. Give it a thumbs up. If you want some eGPUs, I’ve got links down in the description below for you. Clicking on those links and making your eGPU purchases would help support the channel here. Like I said, all of this stuff I purchased myself to test out and do all these trial runs myself, because science right?
So go ahead and click on those links and support our channel here on State of Tech, and I hope to see you back here soon. So click that subscribe, give us a like, and we’ll see you back in the next one.