Apple Silicon M1: benchmarks and review for Mac Mini, MacBook Air and MBP 13 M1 and comparison with other brands

[bill-p]

Nah, that's too easy. Tell him to run 10 loops of Cinebench and report the final score. Should take about half an hour. That'll prove/disprove thermal throttling.

 

[scatopie]

As @leman predicted

Great consistency so far. This is pretty good news in that the throughput is not held back by lack of power source!

 

[Henk van Ess]

Nah, that's too easy. Tell him to run 10 loops of Cinebench and report the final score. Should take about half an hour. That'll prove/disprove thermal throttling.

Can you take over my shift? Check out

https://twitter.com/i/web/status/1328474158687162369

and ask him and post it here

 

[titan4]

Jeez, guys. Just let the guy enjoy his new macbook. I am sure that somebody, who actually cares, will benchmark the crap out of these devices very soon. ?

 

[bill-p]

Can you take over my shift? Check out

https://twitter.com/i/web/status/1328474158687162369

and ask him and post it here

Haha, I was kidding. I'll do that when I have my own MacBook Pro. It should be here some time this week according to tracking.

Let the poor kid get back to school and enjoy his computer. I think we owe him that much.

 

[Henk van Ess]

Haha, I was kidding. I'll do that when I have my own MacBook Pro. It should be here some time this week according to tracking.

Let the poor kid get back to school and enjoy his computer. I think we owe him that much.

Goodnight! Dreaming of the M5.

 

[thingstoponder]

.

As for these benchmarks (so far), they are competitive with Renoir, but they don’t dominate AMD’s offerings. Of course they are doing this a lower power draw (I assume). But this is to be expected, they are using a brand new 5nm process, where as AMD is using a 2 year/generation old process node, combined a generation old core. Zen 3 clocks higher, has much higher IPC than Zen 2 did. M1 is very powerful, but it does not seem to dominate X86 as some seem to believe. Apple has excellent engineers, but so does AMD. I am not mentioning Intel because their issues with their process node prevents them from being able to compete with Apple and AMD.

Don’t get me wrong, I love what Apple is doing with their CPUs. I held off getting a laptop, just to wait for this new silicon, but I think some of us should temper of expectations. I can’t wait to get my MacBook Air to compare with my Intel/AMD systems. I suspect those huge caches, along with Safari with be some of the fastest web machines you can buy since Javascript loves cache.

Let’s wait for actual rigorous testing before jumping to conclusions.

Zen 3 wasn’t that big of a jump in IPC. Relatively small jumps in x86 are seen as a huge deal. Apple’s IPC is far ahead of either company btw.

 

[M1 Processor]

.

Let’s wait for actually tests before jumping to conclusions.

Zen 3 wasn’t that big of a jump in IPC. Relatively small jumps in x86 are seen as a huge deal. Apple’s IPC is far ahead of either company btw.

Now notice through all that prior post, I put the emphasis of “so far“ and I am ”assume” so basically I was making a point of not jumping to conclusions. I will know tomorrow when I have access to the silicon... The IPC increase from Zen 2 to Zen 3 was actually higher than the IPC increase from the A13 to A14, per Anandtech. In addition, AMD managed to increase clocks within the same process node without and increase in power. In two generations AMD increased IPC by around 40 percent, how these jumps “relatively small”? AMD said Zen 4 will feature a similar IPC boost, and that’s not even including the move to 5nm. I think you when you refer to X86, you must be referring to Intel. ?. AMD has been improving by leaps and bounds in recent years, all with a budget that dwarf’s Intel/Apple. We should give credit, where credit is due. It kind of shows that x86 is far from dead, as some people think.

Lastly, I’m well aware that the IPC is higher in Apple silicon, but that doesn’t mean too much if you’re unable to scale clocks high enough. Are you aware that these wide-pipelines with multi-stage decoders designs are inversely related to clock speeds?

And btw I’m not an x86 apologist. I rolled the dice on M1 instead of getting a G14 with its 4900HS/2060 Max-Q combo because I prefer macOS. That’s despite fact the G14 will probably curb-stomp the M1. The M1 seems to be awesome, but we should be reasonable about its performance.


I think we can all agree now, that when Apple was bragging being double the “latest PC chip”, they sure as hell wasn’t talking about AMD. Maybe some 8th gen Intel crap. ?

 

[Sanpete]

I think we can all agree now, that when Apple was bragging being double the “latest PC chip”, they sure as hell wasn’t talking about AMD. Maybe some 8th gen Intel crap. ?

Maybe we'll agree if you reveal what laptop chip you're talking about, since that's what Apple was talking about.

 

[Joe Dohn]

We should give credit, where credit is due. It kind of shows that x86 is far from dead, as some people think.

People in the computing world (not just x86, but everywhere in the tech industry) believes that old technology = outdated. And that's not necessarily true. If you do enough research, you can still refine an old technology and make it competitive.

 

[hower1k]

These comparisons are great and all but they aren't actually that helpful to typical Apple buyers. What people should be comparing the new Macbook Pro, Macbook Air, or Macbook Mini the new M1 models are replacing.

I know Geekbench isn't the greatest tool to use as a comparison but the last intel entry-level Macbook Pro had a single-core score of 905 and a multi-core score of 3746 and the new M1 MacBook Pro has a single-core score of 1726 and a multi-core score of 7580.

Yes, Apple threw out that it's better than 98% of laptops sold or whatever. We all know that the person buying a $300 windows machine from Microcenter isn't going to be in the market for a MacBook Air or Pro so that's a nonsense remark, but it's Apple and that's what they do. I think the best comparison one could make is how does the New M1 Macbook Air compare to the newest Surface Laptop 3.

Apple has doubled its scores in less than a year and increased battery life and doubled the battery life without raising prices. For the average Apple buyer that is what matters, not that it is not as good as some maxed-out Windows gaming laptop or a desktop rig.

 

[M1 Processor]

Maybe we'll agree if you reveal what laptop chip you're talking about, since that's what Apple was talking about.

Well AMD’s latest PC laptop chips are Renoir, and if you look at Apple’s chart the “latest“ chip TPD its 30-40w. If this was an AMD chip it would be the AMD Ryzen 9 4900HS, a 35w TDP Chip. And no, the M1 isn’t double the speed of 4900HS. May actually be slower, but we shall see. They are comparing Intel chips trust me, and they are being a bit disingenuous at that. Like when they compared the Mac Mini, they used the i3 version, and kept the i7 because the M1 probably won’t be hugely faster than the i7 version.

 

[name99]

Now notice through all that prior post, I put the emphasis of “so far“ and I am ”assume” so basically I was making a point of not jumping to conclusions. I will know tomorrow when I have access to the silicon... The IPC increase from Zen 2 to Zen 3 was actually higher than the IPC increase from the A13 to A14, per Anandtech.

The IPC increase of the A8 was also very disappointing. Followed by the STUNNING performance increase (both IPC and frequency) of the A9...

Everyone who follows Apple closely understands exactly what happened.
The priority of the A8 was to halve the CPU power. That was achieved.

The priorities of the A14 were
(primary) get everything lined up for the mac transition. This means a few new instructions (for better emulation) changes to the TLB (again to handle x86), changes to the IO setup, adding a hypervisor, stuff like that.
(secondary) reduce power which had clearly gone a little beyond what made sense with the A13

Apple succeeds by setting LIMITED priorities and achieving them. These add up, year after year. The alternative is to set a dozen goals and achieve none of them because you're unfocussed.

This year's goals were achieved. Next year's goal, I assume, include
(primary) allow the design to grow to supporting something at least matching the 28 cores of the largest Mac Pro (and its dGPU), which likely means 64 large cores (though who knows? maybe 48?). This may be achieved by chiplets, by large dies, by packing four of the next generation SoC in a single package, who knows?
(secondary) take advantage of the density that 5nm provides to boost the CPUs with another large kick in IPC, not least via SVE/2.

nongoal, I strongly assume, is to boost the frequency except perhaps by 100MHz or so, whatever falls out of the other optimization work.
If you imagine that A14's minimal bump in IPC reflects anything except this year's particular schedule,you;re going to be very surprised next year.

 

[M1 Processor]

These comparisons are great and all but they aren't actually that helpful to typical Apple buyers. What people should be comparing the new Macbook Pro, Macbook Air, or Macbook Mini the new M1 models are replacing.

I know Geekbench isn't the greatest tool to use as a comparison but the last intel entry-level Macbook Pro had a single-core score of 905 and a multi-core score of 3746 and the new M1 MacBook Pro has a single-core score of 1726 and a multi-core score of 7580.

Yes, Apple threw out that it's better than 98% of laptops sold or whatever. We all know that the person buying a $300 windows machine from Microcenter isn't going to be in the market for a MacBook Air or Pro so that's a nonsense remark, but it's Apple and that's what they do. I think the best comparison one could make is how does the New M1 Macbook Air compare to the newest Surface Laptop 3.

Apple has doubled its scores in less than a year and increased battery life and doubled the battery life without raising prices. For the average Apple buyer that is what matters, not that it is not as good as some maxed-out Windows gaming laptop or a desktop rig.

I agree, this is impressive as f**. These CPU are hugely impressive, but some people drunk the Apple cool aid and thought they are faster than they really were. Like twitter is furious with the leaked benchmarks when in fact, they are pretty damn fast.

 

[sultanq]

Well AMD’s latest PC laptop chips are Renoir, and if you look at Apple’s chart the “latest“ chip TPD its 30-40w. If this was an AMD chip it would be the AMD Ryzen 9 4900HS, a 35w TDP Chip. And no, the M1 isn’t double the speed of 4900HS. May actually be slower, but we shall see. They are comparing Intel chips trust me, and they are being a bit disingenuous at that. Like when they compared the Mac Mini, they used the i3 version, and kept the i7 because the M1 probably won’t be hugely faster than the i7 version.

Apple never said the M1 is the world’s fastest laptop chip for multi-core performance. They did say it’s the fastest single core, and has double the performance per watt of the “latest PC chip”. The 4900HS roughly matches the M1 in multi-core, and is much slower single core, while requiring roughly 3x the TDP. The M1’s GPU also trounces the integrated Vega 7 of the 4900HS.

 

[M1 Processor]

The IPC increase of the A8 was also very disappointing. Followed by the STUNNING performance increase (both IPC and frequency) of the A9...

Everyone who follows Apple closely understands exactly what happened.
The priority of the A8 was to halve the CPU power. That was achieved.

The priorities of the A14 were
(primary) get everything lined up for the mac transition. This means a few new instructions (for better emulation) changes to the TLB (again to handle x86), changes to the IO setup, adding a hypervisor, stuff like that.
(secondary) reduce power which had clearly gone a little beyond what made sense with the A13

Apple succeeds by setting LIMITED priorities and achieving them. These add up, year after year. The alternative is to set a dozen goals and achieve none of them because you're unfocussed.

This year's goals were achieved. Next year's goal, I assume, include
(primary) allow the design to grow to supporting something at least matching the 28 cores of the largest Mac Pro (and its dGPU), which likely means 64 large cores (though who knows? maybe 48?). This may be achieved by chiplets, by large dies, by packing four of the next generation SoC in a single package, who knows?
(secondary) take advantage of the density that 5nm provides to boost the CPUs with another large kick in IPC, not least via SVE/2.

nongoal, I strongly assume, is to boost the frequency except perhaps by 100MHz or so, whatever falls out of the other optimization work.
If you imagine that A14's minimal bump in IPC reflects anything except this year's particular schedule,you;re going to be very surprised next year.

The point about the A13/A14 IPC was simply a reply to a user who said Zen 3 wasn’t a big IPC jump and that x86 users are used to small jumps. I corrected him, that was the point of talking about the A13/A14, that’s it.

Well I won’t be surprised at all, if I did not believe in Apple I would not have bought the MacBook Air. ? I believe in this chip, but some people need to temper down. I have no dog in the fight between Apple and AMD/Intel, I’m simply reporting things as I see it. Like you said you can simply add cores and provide a major boost in performance, but its simply not that easy, the M1 is already a massive chip (16 billion transistors?), how large will a 64 chip be? In addition, a part of the M1 power saving is probably limited I/O, and having a desktop class I/O will increase power as well. They will also need to compete with AMD and Nvidia in the GPU space with their desktop class processors. AMD has been sitting on 64 core processors for a while now too.

I have to semi-disagree with you on the A14 being lower power than the A13 as well. The iPhone 12 seems to love to throttle, but that I could be the GPU, Idk.

 

[Sanpete]

Well AMD’s latest PC laptop chips are Renoir, and if you look at Apple’s chart the “latest“ chip TPD its 30-40w. If this was an AMD chip it would be the AMD Ryzen 9 4900HS, a 35w TDP Chip. And no, the M1 isn’t double the speed of 4900HS.

Hard to say. Apple claims the M1 gets up to 2x the performance and can match the peak performance of the PC chip at a fraction of the power. That first claim is probably about performance at particular wattages, maybe at 10W specifically. The second claim is the one harder to live up to, probably, if they were talking about a 35W chip. What laptop is the 4900HS in that Apple could have used to compare to in October? We can compare and see.

 

[Kung gu]

Well AMD’s latest PC laptop chips are Renoir, and if you look at Apple’s chart the “latest“ chip TPD its 30-40w. If this was an AMD chip it would be the AMD Ryzen 9 4900HS, a 35w TDP Chip. And no, the M1 isn’t double the speed of 4900HS. May actually be slower, but we shall see. They are comparing Intel chips trust me, and they are being a bit disingenuous at that. Like when they compared the Mac Mini, they used the i3 version, and kept the i7 because the M1 probably won’t be hugely faster than the i7 version.

I think Apple was covering intel latest laptop cpus, cause they never used AMD in their macs

 

[M1 Processor]

Apple never said the M1 is the world’s fastest laptop chip for multi-core performance. They did say it’s the fastest single core, and has double the performance per watt of the “latest PC chip”. The 4900HS roughly matches the M1 in multi-core, and is much slower single core, while requiring roughly 3x the TDP. The M1’s GPU also trounces the integrated Vega 7 of the 4900HS.

And where is your evidence it roughly matches the M1 in multicore? Geekbench? You are aware Geekbench isn’t the be all end all. I suggest you wait until the reviews are out tomorrow. You will probably be surprised. When referring to the “Latest PC Laptop Chip” they WERE referring to multithreaded performance and I quote:

1. Testing conducted by Apple in October 2020 using preproduction 13‑inch MacBook Pro systems with Apple M1 chip and 16GB of RAM. Multithreaded performance measured using select industry‑standard benchmarks. Comparison made against latest‑generation high‑performance notebooks commercially available at the time of testing. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.

It probably will trounce Vega 7, but you’re speaking of something in which you have no knowledge on because we have no comparisons as of this point.

 

[M1 Processor]

I think Apple was covering intel latest laptop cpus, cause they never used AMD in their macs

Exactly that’s what I was saying.

 

[nikidimi]

I am bit skeptical about those 4800U scores. They seem anomalously high compared to third-party reviews. I mean, it’s practically identical to a 45W 4800H above. If I understand correctly, these are user-submitted scores, right? Could be from overclocked systems or not entirely truthful.

Well, the 4800H and the 4800U are almost the same processor, but the U-series is clocked lower (when not boosting) and better binned. The scores are in line with other benchmarks I've looked for this CPU but that's probably a 25W laptop when boosting (which means ~35W for 15 minutes). And it will lose 1/4 of performance after the 15 minutes in most configurations. As I said earlier, the TDP is very confusing these days with all the possible boosts and configurations. So it'll be interesting to see a comparison of the performance degradation for the MBA, MBP and a 4800U-based laptop of a similiar size and weight, not only the benchmark score.

 

[M1 Processor]

Hard to say. Apple claims the M1 gets up to 2x the performance and can match the peak performance of the PC chip at a fraction of the power. That first claim is probably about performance at particular wattages, maybe at 10W specifically. The second claim is the one harder to live up to, probably, if they were talking about a 35W chip. What laptop is the 4900HS in that Apple could have used to compare to in October? We can compare and see.

Look for yourself. It seems pretty clear to me, that at less than 10 watts Apple claims in matches the “Latest PC Laptop Chip” in Multithreaded performance. This ”Latest PC Chip” its obviously a puny PC chip...

“4. Testing conducted by Apple in October 2020 using preproduction 13‑inch MacBook Pro systems with Apple M1 chip and 16GB of RAM. Multithreaded performance measured using select industry‑standard benchmarks. Comparison made against latest‑generation high‑performance notebooks commercially available at the time of testing. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.”

 

[M1 Processor]

I requested this person to run R23 on his MacBook Air and he was kind enough to perform the test. Spoiler alert: Its slower than the MacBook Pro, but still hella fast. I guess the MBA thermal throttles which is the be expected and the 10w config is slower than whatever wattage is in the MBP. 6304/1477

 

[Sanpete]

Look for yourself. It seems pretty clear to me, that at less than 10 watts Apple claims in matches the “Latest PC Laptop Chip” in Multithreaded performance. This ”Latest PC Chip” its obviously a puny PC chip...

“4. Testing conducted by Apple in October 2020 using preproduction 13‑inch MacBook Pro systems with Apple M1 chip and 16GB of RAM. Multithreaded performance measured using select industry‑standard benchmarks. Comparison made against latest‑generation high‑performance notebooks commercially available at the time of testing. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.”

I can't tell what I said that you're responding to. Do you know of a laptop with the chip you suggested that we can compare to?

 

[Admiral]

So the 16 Gb benchmark will probably be better

**** the 16GB benchmark. I want to see the 16-core benchmark. Holy schnikes.