M1 Macs Reviewed

[randomhkkid]

Summary

Reviews just dropped, M1 mac lineup of Macbook Air, Macbook Pro and Mac Mini Reviewed.

 

From Anandtech where they say the M1 in the Mac mini behaves like a 20-24W TDP chip at peak loads.

Spoiler

 

CPU performance:

Spoiler

 

 

GPU performance:

Spoiler

This is under Rosetta 2 emulation. Sounds like since this was previously a metal enabled title only the CPU side is under Rosetta 2 with GPU running metal natively. I'd guess it's translating OpenGL to Metal much like how dxvk does dx to vulkan.

 

 

Quotes

Quote

"The performance of the new M1 in this “maximum performance” design with a small fan is outstandingly good. The M1 undisputedly outperforms the core performance of everything Intel has to offer, and battles it with AMD’s new Zen3, winning some, losing some. And in the mobile space in particular, there doesn’t seem to be an equivalent in either ST or MT performance – at least within the same power budgets." - Andrei Frumusanu Anandtech

Quote

The Verge Summary 9.5/10

GOOD STUFF

• Fast
• Intel-based apps work well
• Excellent battery life

BAD STUFF

• Awful webcam
• iOS apps are a whiff
• It’s time to admit Macs would be better with touchscreens

 

My thoughts

Damn. This is super impressive stuff, I was skeptical about the non-native ARM performance but either Rosetta 2 is magic or these M1 chips are absolute monsters. I'm glad that the 10hour+ laptop is back to being a thing too!

 

Edit: GPU performance is equally as interesting. Looks like Rosetta 2 performance is just below the Nvidia 1650 and above the AMD 560X.

 

Edit 2: I'm also personally disappointed to see each of these new Macs only support 2 displays total so only 1 external display for the Air and Pro

 

Sources

https://www.theverge.com/21569603/apple-macbook-air-m1-review-price-specs-features-arm-silicon

 

https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested

[GravityHurts]

This looks pretty good for apple, but I'm unfamiliar whether or not Geekbench is a good benchmark.

[JoseGuya]

I'm still very conflicted. Do I switch my aging 15 inch late 2013 MBP to the M1 MBP, or to the 2020 16 inch Intel? Do I wait for the 16 inch with Apple Silicon? WHAT TO DO

[A51UK]

Having look up geekbench, it look like it be really bad benchmark for years. We should wait for more benchmarks, It best to not go by one benchmark anyway. I hope we find out today or tomorrow, also be careful with youtube channel you look at for review, it good idea not to go on pro-apple or pro-pc but more in the between like Linus or Game Nexus. people who love computer but not pro anything. 

[ColeWorld]

It looks like Linus' early impression and skepticism to the announcement video wasn't echoed by other reviewers because they had the M1 already in hand being tested. Dave Lee sorta hinted at that in the WAN show last week too. This rift between LMG and Apple just puts LMG a bit behind on reviews for Apple products. 

[ColeWorld]

20 minutes ago, GravityHurts said:

This looks pretty good for apple, but I'm unfamiliar whether or not Geekbench is a good benchmark.

Geekbench isn't a good benchmark. However other benchmarks are also showing great performance for the M1 Chip.

[Commodus]

4 minutes ago, A51UK said:

Having look up geekbench, it look like it be really bad benchmark for years. We should wait for more benchmarks, It best to not go by one benchmark anyway. I hope we find out today or tomorrow, also be careful with youtube channel you look at for review, it good idea not to go on pro-apple or pro-pc but more in the between like Linus or Game Nexus. people who love computer but not pro anything. 

It's not really a bad benchmark, just one that only tells you so much about a given CPU. Some of the additional testing appears to bear out claims that the M1 is very fast -- what surprises me is how elegantly it handles multiple apps at once.

[ColeWorld]

7 minutes ago, JoseGuya said:

I'm still very conflicted. Do I switch my aging 15 inch late 2013 MBP to the M1 MBP, or to the 2020 16 inch Intel? Do I wait for the 16 inch with Apple Silicon? WHAT TO DO

Definitely don't switch to the Intel mac now. If you can wait till they release a 16inch Apple Silicon next year, you definitely should. Unless you don't require that extra sustained power for your workload. There's also rumored to be a design change next year with mini-led displays and a new 14inch and 16in MBP.

[ColeWorld]

7 minutes ago, A51UK said:

Having look up geekbench, it look like it be really bad benchmark for years. We should wait for more benchmarks, It best to not go by one benchmark anyway. I hope we find out today or tomorrow, also be careful with youtube channel you look at for review, it good idea not to go on pro-apple or pro-pc but more in the between like Linus or Game Nexus. people who love computer but not pro anything. 

Definitely, more reviews the better. However other reviews are starting to come out, and it's looking like this thing basically outperforms the top 16in Intel MBP on optimized apps and almost matches it in translated apps.

[Aidoneus]

Anandtech benchmarks show the M1 GPU pushing out 40 FPS in Rise of the Tomb Raider, 1080p Very High. Competitive with entry level dedicated cards. https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested/3

 

These chips look seriously impressive for what they are. 

[ColeWorld]

https://www.macrumors.com/2020/11/17/apple-silicon-mac-reviews-videos/ 

 

This post has a lot of the youtube reviewers and their reviews listed.

[randomhkkid]

30 minutes ago, GravityHurts said:

This looks pretty good for apple, but I'm unfamiliar whether or not Geekbench is a good benchmark.

I was skeptical but Anandtech made a good case for it recently in this article https://www.anandtech.com/show/16226/apple-silicon-m1-a14-deep-dive/4

Quote

There’s been a lot of criticism about more common benchmark suites such as GeekBench, but frankly I've found these concerns or arguments to be quite unfounded. The only factual differences between workloads in SPEC and workloads in GB5 is that the latter has less outlier tests which are memory-heavy, meaning it’s more of a CPU benchmark whereas SPEC has more tendency towards CPU+DRAM.

 

[ColeWorld]

 

Source: YT Dave Lee

[Commodus]

18 minutes ago, JoseGuya said:

I'm still very conflicted. Do I switch my aging 15 inch late 2013 MBP to the M1 MBP, or to the 2020 16 inch Intel? Do I wait for the 16 inch with Apple Silicon? WHAT TO DO

How well is it running? Do you want a large screen?

 

If it's still handling tasks well enough, hold off until there's a 16-inch MBP using Apple Silicon. If you're struggling and need a Mac now, though, let the screen size be your guide — the 16-inch Intel MBP is still a beefy system. But if  you want to go smaller, spring for the 13-inch M1 MBP and don't look back.

[Curufinwe_wins]

Happy to see much more details in this piece from Anandtech. Looking pretty dope for a first attempt. Best efficiency and performance in the <35W market.

 

The only notable worry/question I have is that the phone based A14 at 6W is less than 10% behind in frequency (and around 10-15% behind in performance) compared to M1 at somewhere between 20-25W. That is NOT a good look for scaling up to desktops at all.

 

Of course, many user tasks will be fast enough at the lower frequency (xeons were mostly low frequency for almost a decade), and the extreme performance edge at low tdp means scaling wider should be plausible, but don't give up on x86 just yet. MT results are still heavily favored (as they should be) for the ultra mulitcore chips now on the market, and at the unlimited power state, Z3 just seems to eek things out.

 

 

Rosetta performance is quite impressive, 70-80% of native (when it runs, since some extensions dont seem compatible, but those are not common cases) is extremely impressive stuff. Much better than I was expecting, and almost certainly past the "good enough" stage for again a hyper majority of Apple's userbase.

 

 

A great "first" effort, and hopefully lights a fire under the rest of the industry as well.

[randomhkkid]

3 minutes ago, Curufinwe_wins said:

The only notable worry/question I have is that the phone based A14 at 6W is less than 10% behind in frequency (and around 10-15% behind in performance) compared to M1 at somewhere between 20-25W. That is NOT a good look for scaling up to desktops at all.

To be fair it's not just the same chip with a frequency bump. The M1 has 2 more Big cores, double the GPU as well as a load of things like TB3 controller built in.

 

I think you're right that they're definitely near the limits of the efficiency curve for this base design but I'm expecting them to have a completely separate chip for their higher end models.

[captain_to_fire]

1 minute ago, Curufinwe_wins said:

compared to M1 at somewhere between 20-25W

Apple revealed that the M1 has a TDP of 10W

 

[randomhkkid]

2 minutes ago, like_ooh_ahh said:

Apple revealed that the M1 has a TDP of 10W

 

FullSizeRender.mov

Only 10W in the Air, other devices have more headroom. Mac Mini is 20-25W

[Curufinwe_wins]

6 minutes ago, like_ooh_ahh said:

Apple revealed that the M1 has a TDP of 10W

 

FullSizeRender.mov

That rating can basically screwoff, no offense. I didn't say that the A14 was a sustained 1-2W chip. And M1 in the mac mini where Anandtech is running would be classed as around a 20W part by comparison to other systems. We all know that makers are intentionally obfuscating tdps these days with boosts and such, so I'm not going to talk or report on that metric.

 

7 minutes ago, randomhkkid said:

To be fair it's not just the same chip with a frequency bump. The M1 has 2 more Big cores, double the GPU as well as a load of things like TB3 controller built in.

 

I think you're right that they're definitely near the limits of the efficiency curve for this base design but I'm expecting them to have a completely separate chip for their higher end models.

That is a fair point. I guess it will be more useful to compare to the mb air solution at sustained loads and see what roughly half the power is really getting relative to the mac mini.

[JoseGuya]

17 minutes ago, Commodus said:

How well is it running? Do you want a large screen?

 

If it's still handling tasks well enough, hold off until there's a 16-inch MBP using Apple Silicon. If you're struggling and need a Mac now, though, let the screen size be your guide — the 16-inch Intel MBP is still a beefy system. But if  you want to go smaller, spring for the 13-inch M1 MBP and don't look back.

Struggling at running iOS Simulator, Android emulator,  archiving at Xcode, with VS Code, chrome and slack all running, it gets to a crawl.

 

I'm mostly working at home connected to an ultra wide right now, so the size of the screen is not an issue, but those 2 thunderbolt ports are really just not enough. Waiting for the 16 inch version with apple silicon will be tough 

[ne0tic]

This is looking really impressive! To get this performance and efficiency is huge. Of course, it's just a first generation product so I feel like most people should wait for the second generation to be safe in terms of compatibility. But it looks like Apple are off to a very promising start with Apple Silicon in Mac's. This could be huge for the whole industry in the coming years!

[n0stalghia]

Wait, the Mac mini has higher FPS in Tomb Raider even while emulating the game through Rosetta than a mobile 560X? On 20W TDP? what

[dbcooper]

5 minutes ago, n0stalghia said:

Wait, the Mac mini has higher FPS in Tomb Raider even while emulating the game through Rosetta than a mobile 560X? On 20W TDP? what

It's insane lol

[Aidoneus]

11 minutes ago, n0stalghia said:

Wait, the Mac mini has higher FPS in Tomb Raider even while emulating the game through Rosetta than a mobile 560X? On 20W TDP? what

I’m not sure the emulation of x86 to AArch64 will be having all that much impact on that number. RoTR already had a native Mac x86 version, so presumably all the graphics APIs are Metal anyway, so no translation is necessary for M1. It’s only any CPU calls that would need translating, and I don’t think games like RoTR are traditionally very CPU heavy. 

[LAwLz]

1 hour ago, GravityHurts said:

-snip-

This looks pretty good for apple, but I'm unfamiliar whether or not Geekbench is a good benchmark.

49 minutes ago, ColeWorld said:

Geekbench isn't a good benchmark. However other benchmarks are also showing great performance for the M1 Chip.

 

Geekbench is a good benchmark suite. You just have to understand what it is.

It is essentially a pure CPU benchmark. Workloads are very much real world. It is literally a compilation of commonly used programs such as the C compiler Clang, the PDF reader from Chrome and compression engine from 7-Zip.

The "problem" is that the workloads are very small in terms of memory size, so it doesn't stress the memory very much. This means that a really good CPU but with weak memory interfaces will look just as good as a really good CPU with a really good memory interface.

The thing with the M1 is that it has a REALLY fucking good memory interface. 

 

The 3900X has something like 20GB/s bandwidth for copies, and around 15GB/s for writes. I think the 5950X has similar results.

The M1 gets 61GB/s for copies and around 34GB/s for writes. And it does that with similar memory latency to Zen3 as well. It is fucking insane how good these results are.

 

So Geekbench is more than a fine benchmark. A lot of people who criticize it probably just heard that Linus Torvalds once said that it was bad, but then forget that comment was made like 8 years ago, and he didn't like it because of the small data sets.