EMBER

Why Ember

I wanted Mac gamers to have more than one choice.

Here is why I started Ember and what I am working toward.

It started with a simple frustration

I wanted to work and play on the same MacBook. But when I looked for ways to run Windows games on Mac, there was effectively one practical option, and it was not cheap.

That is not a criticism of the tool. I just wished there were more choices.

Competition moves things forward

Windows programs can run on Mac thanks to Wineand the people who have spent over 30 years improving it together. Projects such as DXVK, DXMT, FEX, and MoltenVK, and the companies and individuals building tools on them, also deserve credit. Ember is built on these projects too.

More choice helps every developer improve faster. I hope Ember can become one of those choices. We publish our changes to open-source code.

Making full use of the hardware

I wanted games to make full use of Apple Silicon. The goal is to reduce bottlenecks in the runtime and graphics translation and make full use of the hardware’s performance.

That is why Ember supports D3DMetal, DXMT, and DXVK for graphics, and both Rosetta and FEX for code translation. The app chooses the fastest path for each game, then adds Ember Flow frame generation, Ember Forge detail restoration, and Ember Boost game mode boosts.

Without complicated setup

I am building Ember to handle installation, launching, and cleanup after a game closes. You should not need to edit configuration files or memorize commands.

If something goes wrong, you can ask Ember. It finds the cause, adjusts the settings, and keeps a record so you can undo the changes.

Performance we measured ourselves

I want to back up claims of speed with numbers. These measurements are from this Mac (M4 Pro). Results vary by scene and settings.

49.7→37.8 msDXMT · synthetic GPU load test · median input→screen latency
24→14 msEmber Flow · Rosetta D3DMetal Smooth · Stellar Blade FHD presentation latency (rounded)
+7%Lost Ark · game fps in a crowded scene · improved GPU readback
23–29
→17–20 msLost Ark · before/after realignment · median of 1-second maximum frame times
10.2→5.6 msValheim · FEX DXMT · FHD · 60 fps cap · internal latency with frame generation off
29 vs 46 msCS2 · Rosetta · DXMT vs D3DMetal · median input→screen latency (rounded)

10/18–20 smoothjob, spinless, fpslat, and realign comparison records. Internal latency and input→screen latency use different measurement methods. See conditions and sources. No fixed latency is guaranteed.

I will also list games and features that do not work. Showing only the successes would not help you make an informed choice.

— Ember’s creator