m1 MacBook Air M1 - Benchmarking & Questions

[exhaustedbrit]

Apologies if this is the wrong forum category, I figured the results in here will be 'news' to some degree.

Today my MacBook Air with the new M1 chip arrived. I've linked some images as proof > Image 1 - Image 2

This thread was spurred by various discussions elsewhere in the forums and aim to consolidate my initial benchmarks - more importantly I wanted to open a channel for anyone to request benchmarks and/or answers to questions you may have. I have available various hardware, including a few Intel based Macs and quite a few Windows machines which I can go into more detail about later.

 

Please let me know any benchmarks or questions you have and I'll update this post to consolidate them as soon as I can.

Technical Observations

Just gonna dump any extraneous ramblings here.

 

Power Draw

I let the laptop sit on charge until it was firmly at 100% and then measured wattage at the wall using my wattmeter with the included 30W USB-C PD charger. Take these measurements with a pinch of salt as the included power adapter will not be 100% efficient, and my meter is only supposedly rated for ± 0.5W of precision. Where I've put variability (±) below this is just based on guesstimate looking at the numbers vary for a brief moment.

At idle with panel brightness at 0% I measured 2.1W ± 0.2W.

At idle with panel brightness at 100% I measured 7.1W ± 0.2W.
Running CB23 multi-core for 10 minutes with panel brightness at 100% I measured 29.0W ± 2.0W - this dropped to 24.4W ± 1.0W after about 9 minutes.

Running CB23 single-core for 10 minutes with panel brightness at 100% I measured 13.7W ± 0.4W consistently throughout the entire 10 minutes.

Playing back 4K60HDR content on Youtube at 1x speed with panel brightness at 100% I measured 9.1W ± 0.2W.

Playing back 4K60HDR content on Youtube at 2x speed with panel brightness at 100% I measured 9.7W ± 0.2W.
Playing League of Legends at 2560x1600 High with Anti-aliasing I measured 16W ± 2W.

 

Benchmark Results

Cinebench R23
After updating macOS & possibly less background stuff going on the results seem to be a bit better.

 

Multi 7724, Single 1517 (Minimum Duration: Off) Result of 1 run after installing macOS 11.0.1

Multi 7261, Single 1508 (Minimum Duration: 10 Minutes) Result of 1 run after installing macOS 11.0.1
 

Multi 7573, Single 1476 (Minimum Duration: Off) Result of 1 run before installing macOS 11.0.1

Multi 7020, Single 1468 (Minimum Duration: 10 Minutes) Average of 3 run before installing macOS 11.0.1

Shadow of the Tomb Raider (Installed via Steam)
All results running under emulation at 100% resolution scale with TAA enabled, HDR off and the 'High' graphics preset.
2560x1600: Min 12 / Max 25 / Avg 14 / 95%n 12 (See results)

1920x1200: Min 18 / Max 39 / Avg 23 / 95%n 19 (See results)
1440x900: Min 25 / Max 60 / Avg 33 / 95%n 28 (See results)

All results running as above but with AA turned off and the 'Medium' graphics preset.

2560x1600: Min 13 / Max 29 / Avg 16 / 95%n 14 (See results)

1920x1200: Min 20 / Max 45 / Avg 25 / 95%n 21 (See results)
1440x900: Min 28 / Max 67 / Avg 37 / 95%n 30 (See results)

 

All results running as above but with AA turned off and the 'Low' graphics preset.

2560x1600: Min 20 / Max 48 / Avg 26 / 95%n 22 (See results)

1920x1200: Min 29 / Max 71 / Avg 39 / 95%n 33 (See results)
1440x900: Min 41 / Max 102 / Avg 56 / 95%n 46 (See results)

 

League of Legends (Installed via Steam)
2560x1600 High with Anti-Aliasing: 60-90FPS depending on busyness.

1440x900 High with Anti-Aliasing: 80-130FPS depending on busyness.

This is running under Rosetta, Activity Monitor shows League of Legends as 'Intel' under the architecture column.

 

World of Warcraft

2560x1600 with settings at '10' (max) and 2xAA running around goldshire: 30-40FPS.
The above is before the upcoming patch, so it's actually running under Rosetta. Tomorrow supposedly blizzard add support for a native binary so we can compare results.

 

Speedometer (Running on Safari 14.0.1)

MacBook Air (Late 2020, M1): 224 runs/minute ± 3.8

MacBook Pro 15” (Late 2019, i9-9880H): 115 runs/minute ± 4.2

MacBook Pro 13” (Mid 2018, i5-8259U): 117 runs/minute ± 1.7
iPhone XS Max: 141 runs/minute ± 5.0

Questions & Answers
Q: Can it playback 4K60 videos on YouTube at 2x speed without skipping frames?
A: It plays back smoothly 4K60HDR YouTube content at 2x playback speed. It claims to drop half the frames likely due to it being a 60Hz panel.

 

Notes

For reference I have access to the following machines for comparisons:

• 2018 MacBook Pro 13" (Core i5-8269U, 8GB, 256GB, 4-port model with 2 fans)
• 2019 MacBook Pro 15" (Core i9-9880H, 32GB, 1TB, Vega 20)
• 2020 MacBook Air 13" (Core M1 8-core GPU, 8GB, 512GB)

I also use two Razer Core X Chroma enclosures with Vega 56s which sadly does not work with the M1 macbook. I have various Windows based machines too, including server hardware.

 

 

Edited November 17, 2020 by exhaustedbrit
Added shadow of the tomb raider & world of warcraft results.

[Spindel]

What's the round time, of late game Civ6 huge map with 12 civs and all city states, like?

[saltycaramel]

What’s its score in Speedometer

 

https://browserbench.org/Speedometer2.0/

 

(and the Intel machines for comparison)

[Kilrah]

Probably want to update the OS before running those, believe there's a 11.0.1?

[randomhkkid]

Exciting! I'd be interested in verification of geekbench running in Rosetta vs native performance.

 

Also can it playback 4k 60 youtube videos at 2x speed without dropping frames? (weird use case but I do this a lot)

 

Thanks @LAwLz turns out I can't read today. It also look like your scores differ significantly from leaked 990/4530 reported here https://hexus.net/tech/news/cpu/146878-apple-m1-cinebench-r23-benchmark-scores-revealed/

[LAwLz]

10 minutes ago, randomhkkid said:

It also look like your scores differ significantly from leaked 990/4530 reported here https://hexus.net/tech/news/cpu/146878-apple-m1-cinebench-r23-benchmark-scores-revealed/

It's explained in the edit of the post. 

They didn't benchmark the M1. They benchmarked the A12Z. 

 

 

Bits and chips is a garbage website/Twitter. Don't give them the attention. 

[exhaustedbrit]

24 minutes ago, randomhkkid said:

Also can it playback 4k 60 youtube videos at 2x speed without dropping frames?

Yes it can - seemed very smooth to me though it's a 60Hz panel which might explain the debug metrics showing half of all frames dropped. Here's a video.

49 minutes ago, saltycaramel said:

What’s its score in Speedometer

Have updated with speedometer results for all my macs and an iPhone XS Max for good measure. All run the same Safari version except perhaps the iPhone.

56 minutes ago, Spindel said:

What's the round time, of late game Civ6 huge map with 12 civs and all city states, like?

Installing Civilization 6 however this might take longer to get results for.

 

26 minutes ago, randomhkkid said:

Exciting! I'd be interested in verification of geekbench running in Rosetta vs native performance.

Will try and give this a go if I can get it working - Steam uses Rosetta 2 and seems to behave fine FWIW.

30 minutes ago, Kilrah said:

Probably want to update the OS before running those, believe there's a 11.0.1?

In the process of installing, will re-run CB23 once installed and finish other benches after it's installed.

[randomhkkid]

20 minutes ago, exhaustedbrit said:

Yes it can - seemed very smooth to me though it's a 60Hz panel which might explain the debug metrics showing half of all frames dropped. Here's a video.

Thanks! Looks like every second it's dropping 30 frames out of every 120 (60fps x 2) so playback is 90fps. Looks like my i9 VRM modded MBP 16 still has some dignity preserved 

[exhaustedbrit]

8 minutes ago, randomhkkid said:

Thanks! Looks like every second it's dropping 30 frames out of every 120 (60fps x 2) so playback is 90fps. Looks like my i9 VRM modded MBP 16 still has some dignity preserved 

At least from my perspective sitting in front of it, it appeared to be entirely smooth so I assumed that the dropped frames was every other frame given the youtube render likely uses `requestAnimationFrame` for compositing and there's only gonna be 60 of those a second since it's a 60Hz panel. Could be wrong though.

[randomhkkid]

2 minutes ago, exhaustedbrit said:

At least from my perspective sitting in front of it, it appeared to be entirely smooth so I assumed that the dropped frames was every other frame given the youtube render likely uses `requestAnimationFrame` for compositing and there's only gonna be 60 of those a second since it's a 60Hz panel. Could be wrong though.

Yeah I think you're right, 90fps presents as 60 due to the panel limitation. Was just curious to see if there was fixed function hardware for 4k 120fps VP9 decode, was cpu usage high during playback? I'm assuming it could either be pure cpu or a hybrid approach if we're dropping frames.

[Spotty]

1 hour ago, exhaustedbrit said:

Apologies if this is the wrong forum category, I figured the results in here will be 'news' to some degree.

Moved to Laptops & Prebuilts. There's already several threads in Tech News discussing the announcement, leaked benchmarks, etc for the new macbooks. If you want to discuss the news around the new laptops then do so in the existing thread(s) instead of creating new ones. Your post seems to be more a discussion about the laptop itself though, so Laptops section.

[randomhkkid]

2 hours ago, exhaustedbrit said:

Benchmark Results

Cinebench R23

Multi 7573, Single 1476 (Minimum Duration: Off) Result of 1 run

Multi 7020, Single 1468 (Minimum Duration: 10 Minutes) Average of 3 runs

 

 

Cinebench is impressive. On my MBP 16 i9 2.4Ghz I just ran a Multi 8221, Single 1192 (Minimum Duration: 10 Minutes) Average of 3 runs

[exhaustedbrit]

44 minutes ago, randomhkkid said:

Was just curious to see if there was fixed function hardware for 4k 120fps VP9 decode, was cpu usage high during playback? I'm assuming it could either be pure cpu or a hybrid approach if we're dropping frames.

I'd assume there's hardware decode going on, I put some power draw measurements above, increasing playback speed to 2x from 1x for that footage resulted in only a 0.6W jump in power draw. Activity monitor showed negligible CPU usage, except that of the monitor itself. I don't think it's breaking a sweat trying to view the content.

16 minutes ago, randomhkkid said:

Cinebench is impressive. On my MBP 16 i9 2.4Ghz I just ran a Multi 8221, Single 1192 (Minimum Duration: 10 Minutes) Average of 3 runs

I let the dust settle and re-ran the benchmarks after updated macOS with a power meter and it's edging ever closer with 7261 in multicore and 1508 on single core with a 10 minute run on wall power.

[SupaKomputa]

It beats Zen 2 single core, ON A LAPTOP.

Pretty scary.

[randomhkkid]

29 minutes ago, exhaustedbrit said:

I'd assume there's hardware decode going on, I put some power draw measurements above, increasing playback speed to 2x from 1x for that footage resulted in only a 0.6W jump in power draw. Activity monitor showed negligible CPU usage, except that of the monitor itself. I don't think it's breaking a sweat trying to view the content.

Just indicates that the decode block doesn't handle more than 90fps for the VP9 codec I guess. Not that it's an issue given macOS's poor support for higher refresh rate displays anyways.

[exhaustedbrit]

I've updated the post to include Shadow of the Tomb raider benchmarks and League of Legends performance numbers. All appear to be running under emulation showing 'Intel' as architecture in Activity Monitor.

[CarlBar]

Do you have access to 3dmark at all? Not a perfect benchmark but it's another one we can find comparative data for easily.

[exhaustedbrit]

6 minutes ago, CarlBar said:

Do you have access to 3dmark at all? Not a perfect benchmark but it's another one we can find comparative data for easily.

Not all iOS apps are available it seems and I could only run the iOS version of Ice Storm Unlimited, it scored 145,572 but each test completed in what felt like less than a second so I really doubt how useful the results are.

I downloaded GFXBench Metal and ran the suite, results (Screenshot 1|Screenshot 2):

Apple M1 - MacBook Air
Car Chase Offscreen (1080p): 176.9 FPS

Aztec Ruins Offscreen (1440p High Tier): 77.37 FPS
Manhattan 3.1 Offscreen (1080p): 272.0 FPS

Seems to be comparable to the Radeon Pro 570X found in the 27" iMacs and or the GTX 980M.

Apple iMac 27" - Radeon Pro 570X

Car Chase Offscreen (1080p): 179.2 FPS

Aztec Ruins Offscreen (1440p High Tier): 80.2 FPS
Manhattan 3.1 Offscreen (1080p): 304.0 FPS

[CarlBar]

3 hours ago, exhaustedbrit said:

Not all iOS apps are available it seems and I could only run the iOS version of Ice Storm Unlimited, it scored 145,572 but each test completed in what felt like less than a second so I really doubt how useful the results are.

I downloaded GFXBench Metal and ran the suite, results (Screenshot 1|Screenshot 2

Apple M1 - MacBook Air
Car Chase Offscreen (1080p): 176.9 FPS

Aztec Ruins Offscreen (1440p High Tier): 77.37 FPS
Manhattan 3.1 Offscreen (1080p): 272.0 FPS

Seems to be comparable to the Radeon Pro 570X found in the 27" iMacs and or the GTX 980M.

Apple iMac 27" - Radeon Pro 570X

Car Chase Offscreen (1080p): 179.2 FPS

Aztec Ruins Offscreen (1440p High Tier): 80.2 FPS
Manhattan 3.1 Offscreen (1080p): 304.0 FPS

 

Fair. Thanks for checking though. Also love that profile image. Who's a good boy.

[Jonathan Heron]

10 hours ago, randomhkkid said:

Just indicates that the decode block doesn't handle more than 90fps for the VP9 codec I guess. Not that it's an issue given macOS's poor support for higher refresh rate displays anyways.

I'm pretty sure it can handle more than 90fps for VP9.

 

I don't have an M1 Mac, but I do have an iMac with a T2, which is based on the A10 and can be used for hardware video encode/decode. I did a test transcode of a 4K VP9 sample file into H264 and it ran at 113fps. I don't know if that was limited by decode or encode, but I'd feel safe saying the M1 isn't limited to 90fps VP9.

 

Update: Playing the same VP9 file in VLC at 120fps shows all frames displayed and no frames dropped.

[SupaKomputa]

4 hours ago, exhaustedbrit said:

Not all iOS apps are available it seems and I could only run the iOS version of Ice Storm Unlimited, it scored 145,572 but each test completed in what felt like less than a second so I really doubt how useful the results are.

I downloaded GFXBench Metal and ran the suite, results (Screenshot 1|Screenshot 2

Apple M1 - MacBook Air
Car Chase Offscreen (1080p): 176.9 FPS

Aztec Ruins Offscreen (1440p High Tier): 77.37 FPS
Manhattan 3.1 Offscreen (1080p): 272.0 FPS

Seems to be comparable to the Radeon Pro 570X found in the 27" iMacs and or the GTX 980M.

Apple iMac 27" - Radeon Pro 570X

Car Chase Offscreen (1080p): 179.2 FPS

Aztec Ruins Offscreen (1440p High Tier): 80.2 FPS
Manhattan 3.1 Offscreen (1080p): 304.0 FPS

They did that with a puny integrated GPU.

AMD and Nvidia should do more homework.

ARM is amazing.

[rikitikitavi]

Can the current state run Dota 2 60+ fps (on busy screen) with some decent visuals?

 

Or any iOS games (pc ports) like Planescape Torment or any other you might have in your library?

[exhaustedbrit]

1 hour ago, rikitikitavi said:

Can the current state run Dota 2 60+ fps (on busy screen) with some decent visuals?

 

Or any iOS games (pc ports) like Planescape Torment or any other you might have in your library?

I downloaded via Steam and also tried the Vulkan support DLC, the performance is much better with OpenGL. The game ran at 25-30FPS at 2560x1600 with everything maxed out, if I keep the resolution at 2560x1600 and render scale at 100% but drop all the settings to low/off frame rates are normally around 80-110 and dip towards 60s during the busy moments. If you want nice visuals though, you can't really stay above 60 it seems. I get away with leaving almost everything on High, turning graphics down or lowering resolution doesn't actually help that much considering how much worse it looks. I found if I just turn shadows, ambient occlusion and anti-aliasing off I am around 60fps most of the time.

If you do just run it at 1680x1050 and leave the settings turned up you'll get around 50-60fps, but IMO I'd rather play with settings all low/off but native resolution than it be blurrier but it can do it if you fiddle a bit.

[rikitikitavi]

Thats really nice. Did it consume much battery?

[exhaustedbrit]

3 hours ago, rikitikitavi said:

Thats really nice. Did it consume much battery?

It was sucking over 8W with powermetrics open, I reckon you can expect 5 hours usage at a stretch if you're not maxing out brightness on the display.