Blackmagic eGPU maximum power to your Mac computer

The official support for external graphics accelerators thanks to macOS High Sierra opened new frontiers and possibilities to all Macs: we enter a time when virtual reality is going to enter with force, and we may need to improve the options that we have in — for example — our laptops, when a new technology appears. No need to switch computers and even taking advantage of laptop mobility while on the road, but turning them into a more powerful workstation when we get home or office.

From High Sierra, macOS is compatible with this type of accelerator. There are already various options on the market, but they are usually composed of external connection boxes – often similar to the size of a semi-tower PC – that integrate some of the graphics cards that can be found on the market. For all this to work, the box has to have a specific chip that makes it compatible with macOS, and in addition, there must be the graphics controllers that we are assembling. Just to put you in context, some of the most used graphics cards in these external custom boxes are being the Radeon Pro Vega, since the latest macOS already include their drivers since the iMac Pro exists.

Apple could be expected to support a brand with a more defined product, complementing the output of the new generation of MacBook Pro that appeared in mid-July: this way Blackmagic arrived and presented us with its exclusive eGPU for Mac, designed in collaboration with Apple itself and they only sell exclusively through their stores. Presented almost as an accessory to further enhance the new portable range, what performance does this new device perform? For this, we tested it together with the 15.4″ MacBook Pro that we analyzed a few days ago, although it can work with any Mac that has Thunderbolt 3 connection via USB-C.

-Blackmagic and Apple have worked together to design this eGPU specifically designed for Macs with Thunderbolt 3-

Blackmagic eGPU
Blackmagic eGPU

A design that accompanies the Pro range

The first surprise we find this product is its physical appearance. Beyond the external boxes as we can find them on the market, which are more functional than colorful, Blackmagic’s design tells us two things: first that it is not randomly created, and second that it has much more of Apple than it seems.

Those who have seen a Mac Pro (2013) will be able to find the matches: a format that looks close to the cylinder (in this case an octagon, rather), made of a single piece of anonymized aluminum, in color «space gray». Does that ring a bell? Not only is the look and feel similar to that of professional laptops or Apple’s iMac Pro, they practically mimic its appearance and can be considered a further extension of the set.

It doesn’t have a button or connection: it will be activated when we connect it to power and a Mac. On the back, it has a very good variety of connection ports as it acts as a hub: this is one of the big differences of this product with the custom facts of the market, its direct Thunderbolt 3 USB-C connection to the Mac and also to monitors which also allows to be used as HUB.

The design uses a thermal grille that cools by convection, i.e. the cold air is collected with a large (and unique) fan at the bottom and pushed upwards cooling the device, ejected by the upper ventilation area. It’s exactly how the Mac Pro’s cylindrical design works on a big touch, and thanks to this it has an impressive drag on the heat, reaching only about 18 dB of noise when the ventilation system is activated. I’ve barely heard it in my trials, it’s really quiet.

-The octagonal design will remind the Mac Pro: its convection cooling system is also very similar (and silent)-

Only with a connection cable to your Mac will you have data, power and four extra USB-A ports behind the device, which makes it very useful when you want to pin it and unpin it with a single cable. No more settings, once connected we will see in the top menu bar an icon indicating that the Blackmagic eGPU is mounted:

We can connect or disconnect it therefore without problems hot. If we want to unpin it, we can safely disassemble it as if it were one more external drive, from that menu bar. To check the use we are giving to the new system graphics card, we will use the Activity Monitor by viewing the floating window «GPU History»:

It’s important to add that the external drive speeds up the graphics of the external monitor, not the graphics of the laptop monitor itself or the iMac Pro if you have it connected to it. For handymans, this limitation can be skipped: there is a Unix Shell that we can launch on our Mac to set the laptop screen as an accelerated screen, but it is not official or has a graphical interface to install it easily, so it is not a feature designed for it – though it works.

Another point that generates more doubt with this product is its compatibility with Windows using Boot Camp. In theory, it is possible to speed it up in Windows 10 Pro if we have the right drives, but as they comment on one of the best forums about eGPUs, the easiest way to do it causes constant errors – at least with what we have today. There is another safer way to change the EFI loaded by an external drive, but it is not without problems either and of course it is absolutely unofficial.

This Blackmagic product is intended to be used exclusively on Mac, as we leave it very clear already in the product box. It is likely that in a few months we will have more stable drives designed by the community (I take advantage to recommend this great page about it), but at the moment it is not stable for the day to day.

Synthetic testing: for the differences between OpenCL and Metal

With macOS Mojave, Apple will begin the transition to its own Metal graphics library. Currently, it is also used in conjunction with OpenCL and OpenGL, open cross-platform libraries. The shift towards Metal makes a lot of sense if we think of Apple and its ecosystem: on iOS Metal is used by the acceleration facilities provided by the operating system, and it is logical that greater cohesion between iOS and macOS at this point makes sense for the future of all to the company’s platform – especially considering the control between software and hardware that they have.

To take full advantage of external acceleration drives like this, developers have to prepare their software in order to squeeze every last GPU calculation and not just the conventional calculation acceleration offered by the system Operating. With High Sierra we have started to see timid advances in this regard, but it will be with Mojave where we will see really drastic changes. For example, the current version of Final Cut Pro X is not yet ready to squeeze this eGPU (not even Apple itself takes it out of its product becnhmarks), so I haven’t included it in my tests – a future Version prepared for Mojave could fix it.

-It will be with macOS Mojave when we see the full power of these eGPUs: developers have to prepare their software to squeeze them to the fullest-

In any case, given the power of this Blackmagic solution, before practical testing if you wanted to have a good performance sampling of both controllers (OpenCL and Metal) also in synthetic tests. The most appropriate solution to test each of the cases, is Geekbench 4 Version 4.2.3 (401111) and its complete «Compute Test», which allows you to use this software to measure among other things machine vision patterns by sobel testing, process algorithms using Fourier transform, facial detection tests, particle physics…

For this I have measured each of the values of this test for all the graphics cards of the MacBook Pro of 15.4″ (2018) and that of this eGPU of Blackmagic, both in the OpenCL library and in Metal, resulting in the following table (in bold the best score per row) . By clicking on the link of each result, you will go directly to the full report of each of my tests:

As you can see, in gross score on synthetic tests, the Metal library works much better than OpenCL. This has nothing to do with the accelerator, as you’ll notice that the numbers also have a similar distance between the cards that MacBook Pro itself integrates. What we do start to see with these tests is the computing capability of the Blackmagic device: almost double that obtained without acceleration.

We continue testing using LuxMark, the popular benchmarking tool based on LuxCoreRender, but here we can only do the tests using the OpenCL library. We reviewed the result we got in the Analysis of MacBook Pro 15.4″ (2018): 1554 points. This test uses all the graphics cards in the system to get an overall score, making it a good measure to know how the sum of the eGPU to the laptop improves.

That is, the sum of the Blackmagic device improves the graphics capacity of the laptop by about 2.5x. This result is obtained we could call it a mixed test between synthetic calculation values and the rendering of a 3D scene (the most complex of the test), so it is also good value to contextualize what this eGPU is contributing to us as an improvement in the subsystem.

Practical tests: let’s look at those frames per second

Obviously a good product to measure the performance of the Blackmagic solution is to use the well-known DaVinci Resolve, in one of its best-known performance tests, the famous «Candle Test». This is the use of a high-definition image to which sets of distortion or noise-reducing nodes are applied simultaneously. These effects are applied in real time and the program shows us some fps (frames per second) that are worth it as a measure of performance.

This test is designed to stress our graphics equipment, which I make sure to select manually from the program settings while monitoring it with High Sierra GPU History. Before starting the test, I also make sure to raise the fps of the project to about 60 (we will not reach them with this accelerator, but it serves to be able to have a «ceiling» higher than the 24 fps that the default software has). The results are as follows:

The impression is that – of course – the company has optimized its software to take advantage of this external acceleration hardware. The numbers, despite being a very demanding test, come close to doubling the values without eGPU. The values I have obtained this test according to the rules of the Candle Test: if a test oscillates between two values, I always keep the smallest – despising decimals.

We move on to a couple of benchmarks that are starting to bring us closer to real 3D use: Unigine’s testing, a cross-platform game engine that is also used in virtual reality systems, high-load graphics games or virtual representation. Both tests I like because they are complete not only in the visualization, they are also interactive and very customizable.

With the default settings of both tests, we see that the values are practically double with the external accelerator, regardless of the test or the resolution settings that we have. I forced anti-aliasing filters to further stress the load, but they are good results which are fast enclosing that it is approaching to double performance. This may seem normal to you, but you have to keep in mind that the graphics card of this eGPU is a generation very close to the one that mounts the MacBook Pro 15.4″ (2018) top of the range, which we are using – on laptops with less power , like a 13″ or even from previous years, the jump can easily triple the results.

Let’s go with a look at the games, we’ll use the same from the analysis of MacBook Pro 15.4″ (2018), but elevating the complexity of the graphic effects. The goal is to see how playable these games are when they are required to power as much as possible. By having to play on the external monitor (I use the Asus Thunderbolt 3 I analyzed a few days ago), we will lower that resolution to suit the screen size: 2560×1440.

Although the numbers do not seem to improve as dramatically what we got with the MacBook Pro cards themselves, it should be remembered that in the case of eGPU testing all the graphics settings have been taken to Ultra or Elite modes. This improves the complexity and quality of the graphics as well as the appearance of the games, but does not make the fps duplicate. If we keep the parameters of the original test of the analysis of the laptop without going into the highest settings, with this eGPU we can easily reach 50 – 55 fps or 60 fps without sacrificing great graphic changes and is perfectly valid to play on 4K (for example, in Fortnite).