Metal and OpenGL: Best Settings for MacBook Pro 2020?

[A_B_C]

Hi everyone,

yes, I am aware that there are quite a few threads on the difference between Metal and OpenGL. But maybe some kind soul out there could nonetheless provide an answer to my questions.

• What are the preferred settings for Display and Hardware Acceleration in Preferences for a 16" MacBook Pro 2020, baseline model with AMD Radeon Pro 5300M?
• Should I use Metal wherever available?
• Are there any benefits of using OpenGL? What is the advice “Switch to OpenGL if you encounter issues” supposed to mean? Why can you expect that switching to OpenGL will cure any issues that may arise?
• Is there a difference in power consumption between the two modes?

Somehow this is still a bit puzzling to me. Thank you in advance, 

Alex 😀

[Lee D]

There isn't really any preferred settings, the apps will default o particular settings based on the system.

Ideally if Metal is available yes, if should be enabled, the sae goes for most apps if they have an option.

OpenGL and Metal uses different ways to render and process, so if one isn't performing as good it's worth switching to the other and comparing, see this FAQ article we have on Performance related issues for more information.

Power consumption is handled by the OS, but using a hardware based acceleration may consume more power than using software.

[A_B_C]

Thank you, Lee. So I’ll leave Metal on for everything, and we’ll see how it goes … 😀

[lepr]

On 8/6/2020 at 3:03 PM, Lee D said:

OpenGL and Metal uses different ways to render and process, so if one isn't performing as good it's worth switching to the other and comparing,

Surely you would need to do the comparison first so that you could determine that "one isn't performing as good", and then choose which to use.

[James Ritson]

On 8/3/2020 at 4:50 PM, A_B_C said:

Hi everyone,

yes, I am aware that there are quite a few threads on the difference between Metal and OpenGL. But maybe some kind soul out there could nonetheless provide an answer to my questions.

• What are the preferred settings for Display and Hardware Acceleration in Preferences for a 16" MacBook Pro 2020, baseline model with AMD Radeon Pro 5300M?
• Should I use Metal wherever available?
• Are there any benefits of using OpenGL? What is the advice “Switch to OpenGL if you encounter issues” supposed to mean? Why can you expect that switching to OpenGL will cure any issues that may arise?
• Is there a difference in power consumption between the two modes?

Somehow this is still a bit puzzling to me. Thank you in advance, 

Alex 😀

Hi @A_B_C, for a modern Mac system with a reasonably powerful GPU that has sufficient VRAM, you should absolutely use Metal Compute (hardware acceleration) where possible.

The main hurdle to overcome is graphics memory. macOS will use a certain amount of VRAM to simply drive the display—the amount is dependent on your screen resolution. Previous 15” MacBook Pro models would use a resolution of 3360x2100 I believe (1680x1050 logical resolution) but the 16” models use a slightly higher resolution. Be aware of this memory usage if you plan to use an external display at 4K or 5K—even more so if you run an external drive whilst keeping the lid open to use the internal display as well.

In addition to this, the Affinity apps will demand graphics memory for the Metal view context. With Metal Compute, this memory requirement increases because the GPU is also being used for compositing, and this is typically where you may run into slowdown if you max out the VRAM—you’ll then be eating into swap memory which incurs a performance penalty.

Under everyday editing circumstances, you may never run into any memory issues. If your editing experience suddenly becomes very slow and laggy, however, you may have run into the limitations of your graphics card. I know you can configure the 5500M/5600M models with 8GB, but am I right in thinking the 5300M can only have 4GB? In which case you will simply have to work in your usual manner and try for yourself!

As an example, I do a lot of work with 32-bit OpenEXR multichannel documents—I typically have about 20-30 layers in a document that averages 5000-6000px in width (and usually similar in height). On top of all that, I run a 5K monitor. This combination is a little extreme, but I can edit comfortably with 8GB VRAM most of the time. I previously had a MacBook with a 4GB VRAM GPU however, and I did find it wasn’t quite sufficient—I ended up borrowing an external GPU enclosure with a card that had 8GB VRAM.

For most editing, including compositing in 16-bit with multiple layers, I think you will typically be fine with 4GB—again, though, this is also dependent on your screen resolution and document resolution.

The advice about switching to OpenGL is typically just a troubleshooting step, but it has more meaning for Photo since you have to disable hardware acceleration to use it—thereby alleviating the significant GPU memory requirement and potentially speeding up older systems where the GPU’s capabilities would not be meaningful enough anyway.

Hope that helps!

[A_B_C]

That was extremely helpful, James! Thank you so much! 😀

I think the 5300M can indeed only have 4GB. When just using the internal display of the MBP 16", I haven’t run into any performance issues so far. And since my external EIZO screen still has one of the low-density resolutions, I haven’t had the impression of any performance drops here either. The real-time processing of live filters and blend modes with Metal Compute enabled is just stunning. You guys did an amazing job in this area. 😀

(I actually decided to get the baseline MBP 16" model, as the new Apple-Silicon-based computers were already looming at the horizon, so I wasn’t sure whether it would make sense to invest in a top-performance model at this moment in time.)

[R C-R]

On 8/14/2020 at 4:14 PM, James Ritson said:

The main hurdle to overcome is graphics memory.

The new 27 inch 2020 iMac model that comes with an 8 GB Radeon Pro 5500 XT as standard can be factory configured instead with either an 8 GB Radeon Pro 5700 or a 16 GB Radeon Pro 5700 XT.

So, aside from the difference in graphics memory between the two 5700 options, for use with the Affinity apps can I expect to see any better performance with one of them vs. the standard 5500XT?

[James Ritson]

On 8/15/2020 at 10:38 PM, A_B_C said:

That was extremely helpful, James! Thank you so much! 😀

I think the 5300M can indeed only have 4GB. When just using the internal display of the MBP 16", I haven’t run into any performance issues so far. And since my external EIZO screen still has one of the low-density resolutions, I haven’t had the impression of any performance drops here either. The real-time processing of live filters and blend modes with Metal Compute enabled is just stunning. You guys did an amazing job in this area. 😀

(I actually decided to get the baseline MBP 16" model, as the new Apple-Silicon-based computers were already looming at the horizon, so I wasn’t sure whether it would make sense to invest in a top-performance model at this moment in time.)

If your Eizo monitor is 2560x1440 (or 2560x1600 for 16:10) then you'll be in an even better position performance-wise! Probably a wise choice given the ARM MacBooks. I have an almost maxed-out 16" MacBook for work—it performs well but the thermals are terrible, the fans are almost constantly active if I'm working in Photo or anything else for that matter. I've had previous MacBooks with the older Polaris GPU architecture (and CPUs with fewer cores!) and they were much quieter. Cannot wait for the ARM models..

On 8/15/2020 at 11:19 PM, R C-R said:

The new 27 inch 2020 iMac model that comes with an 8 GB Radeon Pro 5500 XT as standard can be factory configured instead with either an 8 GB Radeon Pro 5700 or a 16 GB Radeon Pro 5700 XT.

So, aside from the difference in graphics memory between the two 5700 options, for use with the Affinity apps can I expect to see any better performance with one of them vs. the standard 5500XT?

I believe compute performance will increase going from 5500XT - 5700 - 5700XT, since each model has a greater number of compute units and stream processors. Whether the performance increase is meaningful is another matter—a 5500XT will likely be sufficient for most workflows since even older mobile Polaris GPUs (e.g. the 455) could easily hit over 100fps when compositing, say, a live motion blur filter.

It's dependent on your screen resolution, typical document bit depth and your workflow however. If you work with lots of high resolution layers, you may want to opt for a more powerful GPU, especially if it has more VRAM. If you just develop RAW files, add a few adjustments and filters then call it a day, you would be fine with any modern baseline GPU—even integrated solutions.

It's worth noting that not every operation is hardware accelerated as of yet—vector layers like a Curve path or quick shape for example. If you use these in conjunction with complex live filters, you'll be bottlenecking the render pipeline with CPU operations, therefore losing the advantage of compositing purely on GPU. It will still be faster than software, just not excessively fast—in this sense, you can't really "brute force" performance with a faster GPU, so don't be too concerned with spending large amounts on a top of the line model.

[A_B_C]

Thank you again, James! Your insights are highly appreciated. 😀

Yes, it will be interesting to see the first benchmarks of the ARM-based models. I’m sure you guys have already received one of those transition models of the Mac Mini, but as far as one could read on the internet, this model is probably just meant for testing out software on the new architecture, less intended for doing actual CPU- and GPU-intensive tasks. As far as I’ve read, the processor seems to be throttled and not be using all of the available cores. So the first ARM-based Mac will surely be an interesting reveal.