Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Silicon M1 MacBook Pro Earns 7508 Multi-Core Score in Cinebench Benchmark
- Thread starter MacRumors
- Start date
- Sort by reaction score
I don't know. I was going to say it's 10 W, but that's for the MacBook Air, according to https://www.apple.com/mac/m1/ For Geekbench, the thermal envelope difference between that and the MacBook Pro doesn't seem to matter, but for 10 minutes on Cinebench it probably will. I guess we'll have to wait for CineBench scores on the Air!
I agree, on the other hand, the M1 processor only has 4 high-performance cores, just like Tiger Lake. The other 4 cores are lower performance, according to Anandtech's SPEC2006 benchmarks at an average 25% performance of the high-performance cores. I don't know what the SMT yield of virtualized cores is these days, when Intel came out with their SMT implementation I remember it was around 25% on average, that's why the multithread comparison between M1 and i7-1185G7 seems relatively fair.
I agree, if they are in the same power envelope. For the MacBook Air, I realize we have no idea what the power envelope is. However, for the MacBook Air with a 10 W TDP that would currently be the Core i7-1160G7, Geekbench v5: ~1500 ST, 4450 MT, which is substantially lower than the M1's 1700/7500.
Maybe from Dell and HP, but overall 11th generation core processors make up only a tiny sliver of all notebook processors currently on offer in the notebook market. The vast majority is 10th or lower generation core processors. Plus, if I go e.g. to the Asus web store and search for i7-1165G7, they don't have any in stock currently. And that's without a 5-10% contingent that normally would have gone to Apple... .
So, your bluster is coming back to bite you and your response is to blame someone else because in your haste to bluster, you couldn’t remember your own post, or even bother to reread your own short post even when it had been quoted in full for your reading pleasure. Nice.
And to answer your question, no I have nothing vested in Geekbench, and in fact happen to find both the Geekbench and Cinebench benchmarks fun.
Now that we got the numbers out of the way, can we see real-world usage? I'm curious to see how smooth everything is in and around the OS. If the experience is the same as current Intel MacBook Pro's, than the M1 suddenly has less value to me - with battery life being the sole benefit and even then, 20 hours of battery life amazing but my workflow doesn't more than the approx 9 hours on my current MacBook Pro.
I am in no way ******** on the M1, but it's like debating between a Tesla or another EV. I don't care so much about range than I do total experience.
I am in no way ******** on the M1, but it's like debating between a Tesla or another EV. I don't care so much about range than I do total experience.
You’ll get extra longevity out of the battery due to less cycles/ lifetime.
When Apple releases a more powerful chip, then I'll discuss that chip. For now, the M1 is the best they've got. In fairly expensive machines, mind you.
You can talk about entry-level all day long, but the Macbook Pro don't have no entry-level price.
it's not meaningless; it's impressive; but yes a more interesting test will be monitoring temp/fan speed and running the benchmark in loop to see if it throttles and how much; and how noisy it is at that point. there we will se the difference in performances between the three machines.
Pretty impressive results.
Looking through the Cinebench R23 charts, it's pretty obvious that the 11th gen Intel i7-1185G7 is the best CPU Intel has ever released, by a good margin, and it much more directly competes against the M1 than Intel's desktop chip--it's got the same number of performance cores, and with a 28W TDP (14W "down-TDP" for thermally constrained form factors) it's targeting roughly the same thermal envelope when under heavy load. Basically, it's targeted to about the same classes of computer as the M1.
It's actually turning in very slightly better single-core results in Cinebench than the M1, which is significantly different from some other benchmark results. It's still, however, a solid 15% slower than the M1, which given Intel's decades in the business versus this being Apple's first "computer grade" mobile chip is still a very impressive feat.
The other chip to be comparing to is, I believe, the Ryzen 7 4700U, which has been out a bit longer and has twice as many cores with around the same TDP (25W). It falls way behind both CPUs in single-core, so isn't going to perform nearly as well in day-to-day tasks, but it does outperform the Intel part on multi-core while coming in a decent margin behind the M1.
Much less flattering for Intel are the results of the 10th gen 6-core i7 series with a bigger-laptop 45W TDP, which have only been on the market for a few months longer than the i7-1185, and are right in line with the M1 for performance.
...but a lot of "real world" computing doesn't involve sustained throughput, so such benchmarks may actually undersell the real-world "feel" of something thermally limited like the Air with a chip that can run very fast for short periods of time.
Reasoning: If I'm doing "everyday user" tasks, the CPU isn't sitting there maxed out for the vast majority of the time I'm using the computer. It will do a bunch of work loading a web app or rendering a complex page, then sit there quietly for a while. It'll to a bunch of work processing a image for a couple of seconds, then sit there quietly for the next minute while I touch up small parts of the photo. All modern CPUs can crank a single core way up in output to work with workloads like this, but if--and this is purely hypothetical--the M1 is better at doing that for a few seconds at a time in the tight thermal envelope of a MBA than an under clocked/thermally limited i7-1185G7, then it's going to behave much better in real world performance than sustained benchmarks will indicate.
This is one of the reasons that iPhones are demonstrably so fast--they can't sustain huge CPU throughput for more than a few seconds due to the thermal constraints, but many everyday tasks only require a few seconds of huge throughput. So while it won't render video as effectively as a laptop with a fan, it's faster--and feels faster--in a lot of everyday tasks.
It's also akin to the situation with SSDs; many M.2 SSDs can't sustain huge throughput for long periods of time due to thermal limitations, but they don't need to under many cases, since what they're generally tasked with is short flurries of heavy activity followed by long periods of idle.
All of that is purely hypothetical at this point, and honestly is kind of hard to benchmark accurately even given ample time and tools, but it's worth remembering that while number-crunching through heavy sustained workloads are an important thing for some pro users, for a lot of use cases, it's the "burst-processing" capability that makes the biggest difference.
Looking through the Cinebench R23 charts, it's pretty obvious that the 11th gen Intel i7-1185G7 is the best CPU Intel has ever released, by a good margin, and it much more directly competes against the M1 than Intel's desktop chip--it's got the same number of performance cores, and with a 28W TDP (14W "down-TDP" for thermally constrained form factors) it's targeting roughly the same thermal envelope when under heavy load. Basically, it's targeted to about the same classes of computer as the M1.
It's actually turning in very slightly better single-core results in Cinebench than the M1, which is significantly different from some other benchmark results. It's still, however, a solid 15% slower than the M1, which given Intel's decades in the business versus this being Apple's first "computer grade" mobile chip is still a very impressive feat.
The other chip to be comparing to is, I believe, the Ryzen 7 4700U, which has been out a bit longer and has twice as many cores with around the same TDP (25W). It falls way behind both CPUs in single-core, so isn't going to perform nearly as well in day-to-day tasks, but it does outperform the Intel part on multi-core while coming in a decent margin behind the M1.
Much less flattering for Intel are the results of the 10th gen 6-core i7 series with a bigger-laptop 45W TDP, which have only been on the market for a few months longer than the i7-1185, and are right in line with the M1 for performance.
I take some issue with this "real world" claim. Yes, Geekbench is a very synthetic benchmark. And yes, Cinebench is a real-world task, albeit a very specific one. And yes, if you're doing heavy crunching, Cinebench is going to give you a much better idea of what the CPU can do within its thermal envelope (I'm assuming the Air will turn in measurably worse results than the Pro since it's limited to 10W).
...but a lot of "real world" computing doesn't involve sustained throughput, so such benchmarks may actually undersell the real-world "feel" of something thermally limited like the Air with a chip that can run very fast for short periods of time.
Reasoning: If I'm doing "everyday user" tasks, the CPU isn't sitting there maxed out for the vast majority of the time I'm using the computer. It will do a bunch of work loading a web app or rendering a complex page, then sit there quietly for a while. It'll to a bunch of work processing a image for a couple of seconds, then sit there quietly for the next minute while I touch up small parts of the photo. All modern CPUs can crank a single core way up in output to work with workloads like this, but if--and this is purely hypothetical--the M1 is better at doing that for a few seconds at a time in the tight thermal envelope of a MBA than an under clocked/thermally limited i7-1185G7, then it's going to behave much better in real world performance than sustained benchmarks will indicate.
This is one of the reasons that iPhones are demonstrably so fast--they can't sustain huge CPU throughput for more than a few seconds due to the thermal constraints, but many everyday tasks only require a few seconds of huge throughput. So while it won't render video as effectively as a laptop with a fan, it's faster--and feels faster--in a lot of everyday tasks.
It's also akin to the situation with SSDs; many M.2 SSDs can't sustain huge throughput for long periods of time due to thermal limitations, but they don't need to under many cases, since what they're generally tasked with is short flurries of heavy activity followed by long periods of idle.
All of that is purely hypothetical at this point, and honestly is kind of hard to benchmark accurately even given ample time and tools, but it's worth remembering that while number-crunching through heavy sustained workloads are an important thing for some pro users, for a lot of use cases, it's the "burst-processing" capability that makes the biggest difference.
A comparison with x86 Cinebench on R2 would actually be interesting. Regardless, I'm looking forward to reading the professional reviews to hear how the new AS M chip performs under normal usage conditions (i.e. a mix of nativeapps and x86 apps on R2).
It's obvious you have some emotional attachment to Geekbench and one thing I've learned is never to attempt to reason with someone who is acting off emotion. I'll continue to diss Geekbench because it sucks, your emotional attachment won't change that.
Which CPUs are you referencing from AMD and Intel that are as good or better? The Ryzen 4800U is not scoring as high as the M1 and nothing Intel has in the same TDP range as the M1 is even close. The AMD 4800U is all over the map in Geekbench Browser and the Core i7-1195G7 is more consistent, but neither of them is on par with the M1 in Single- and Multi-core. Both of these CPUs are consuming more power than the M1.
Well then you are doing it wrong.
If you compare it against the class of machine, and what class of CPU is in it prior to this update, then it is totally apples to apples and wins hands down.
Just because you make up your own rules doesn't make you right
I fully plan on doing it, but with the COVID situation, I am in no real rush to go out and test a M1 Mac. I am hoping someone will upload a video just to showcase usage rather than a benchmark that many won't even come close to utilizing day-to-day.
The 4800U scores over 10,000 in multicore compared to the M1's 7500. That's 33% faster. I don't care at all about TDP.
I am comparing them because the scores are nothing to write home about. My six core 2013 Mac Pro achieves 5,468 on this benchmark. That's 27% slower than the M1 processor. That's a 2013 Mac Pro based on 11 year old technology. Yes, the power consumption is lower for the M1 but I would expect as much given the amount of time which has elapsed.
I'm not trying to take anything away from these new systems but I think we should keep things in perspective.
Thanks. Quoted for posterity.
In any case, I would suggest you read the well written post above from @Makosuke. You might finally learn a thing or two about these benchmarks.
You are literally making up your own rules, with a bunch of different artificial qualifiers that suit the M1.
I'm comparing it to laptop chips that are found in similarly priced machines.
I don't know the TDP of the M1 yet.
Edit: But so far appears to be winning if the above "10W" estimate is correct.
Apple has a fully paid, unrestricted, unlimited ARM license. That means they can do with it whatever they like. ARM has no control about M1 whatsoever.