For a compute-bound job, I just want it to run faster. As promised by Apple.
A few more watts won't matter.
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.
MBP M1 Max vs. Core i9
- Thread starter chars1ub0w
- Start date
- Sort by reaction score
A few more watts won't matter.
Where did Apple specifically promise that this one weird bad benchmark would run faster?
And yes, I am willing to claim that it's a very bad benchmark. It's an extremely narrow part of Perl, and unlikely to predict performance in other tasks - probably not even common regex-based loads. You can easily find benchmark suite results which prove that the M1 Max is lots faster than the 2018 i9 MBP, both single and multithreaded.
MacBook Pro (16-inch, 2021) vs MacBook Pro (15-inch Mid 2018) - Geekbench
browser.geekbench.com
You should know regex is a core feature of Perl. You classify it as "bad" because it doesn't conform to Apple's claims (see Apple's diagram). I showed an independent regex engine, that of Python, also doesn't show that the M1 is superior.
Bottom line is if the battery is dry the notebook is a dead weight. While the raw performance is comparative the efficacy is not and that's where Apple Silicon excels.
Q-6
Q-6
No.. You haven't proven any such thing. There are countless reviews and benchmarks proving Apple's claims.
It's not even up for debate.
While true, compute heavy operations are usually run with the A/C adapter in. At least, when I'm running bigger jobs, I make sure I'm always plugged either on the Intel in my office or the M1 Max at home.
And I've observed plenty of times when I've noticed my M1 Max is not faster than my Intel MBP despite Apple's claims. So I tested a bunch of code, and yes there are plenty of such cases. You can run my Perl and Python code to confirm. Sure there are plenty of times when I've noticed the M1 Max is faster, e.g. at detecting objects in Photoshop etc.
I did a quick test with my 2017 MBP Core i7 and the results are similar to the above:
n=5: 0.042
n=10: 0.044
n=15: 0.044
n=20: 0.095
n=25: 1.667
n=30: 51.977
n=35: didn't test
this test probably has an artificial performance ceiling at rather low CPU requirement, unless i9 is also not better than the 2-years older i7.
Again your annecdotal feelings and some random scripting examples are not comprehensive evidence of the claims you are making.
Sites like anandtech have run industry standard benchmarks like SPEC etc and have proven Apple's claims.
I strongly suspect the code you are running will run the same on any modern CPU and not show a difference because the code itself is the problem and not the CPU itself.
Last edited:
Not quite, Stockfisch is actually real software that does something 😁
I ran this perl script on 3 machines.
2013 Core i7 15" MBP
2019 Core i9 16" MBP Intel Core i9-9880H
2021 M1 MAX 16" MBP
With n= 25
2013 Core i7 15" MBP = 4.7 sec
2019 Core i9 16" MBP = 3.75 sec
2021 M1 MAX 16" MBP = 2.7 sec
With n =30
2019 Core i9 16" MBP = 2 mins 1 sec
2021 M1 MAX 16" MBP = 1 mins 32 sec
So you can see the M1 Max is faster.. however the interesting thing is in power consumption.
n=30 run
Core i9 32-35W Core Power
M1 Max 4.6 W Core power
This is a really silly test.. even then the OP's conclusion is grossly incorrect. It is very clear the M1 is faster and consumes a whopping 6x less power.
Once the i9 gets heat soaked the n=25 runs end up being 4 secs.. the M1 Max consistently does 2.6 - 2.7 seconds with no slow downs.
2013 Core i7 15" MBP
2019 Core i9 16" MBP Intel Core i9-9880H
2021 M1 MAX 16" MBP
With n= 25
2013 Core i7 15" MBP = 4.7 sec
2019 Core i9 16" MBP = 3.75 sec
2021 M1 MAX 16" MBP = 2.7 sec
With n =30
2019 Core i9 16" MBP = 2 mins 1 sec
2021 M1 MAX 16" MBP = 1 mins 32 sec
So you can see the M1 Max is faster.. however the interesting thing is in power consumption.
n=30 run
Core i9 32-35W Core Power
M1 Max 4.6 W Core power
This is a really silly test.. even then the OP's conclusion is grossly incorrect. It is very clear the M1 is faster and consumes a whopping 6x less power.
Once the i9 gets heat soaked the n=25 runs end up being 4 secs.. the M1 Max consistently does 2.6 - 2.7 seconds with no slow downs.
Last edited:
And here is the same regex performance using R's built in regex engine. Didn't run it on the i9. This is just to make clear that you are measuring pathological performance.
The R code was
C-like:
dumb_regex_bench <- function(n) {
s <- paste0(rep("a", n), collapse = "")
r <- sprintf("(a?){%s}a{%s}", n, n)
grepl(r, s)
}You should know that benchmarking just one feature of Perl, a language with a very high feature count, is a terrible benchmark. Even if that feature is a very important Perl feature (which of course regexes are, nobody disputes that), you're being far too narrow. Especially when you then proceed to make sweeping generalizations based on this result.
Hell, I wouldn't trust your test to accurately predict general regex performance. It's a weird regex to care about even if literally all you use a computer for is to plow through data sets with regex.
If you actually are an academic, you know full well you are being the exact opposite of rigorous in your approach to finding out whether M1 is faster than your i9, and in your responses to criticism. I think you're just here to troll.
Last edited:
But you are not measuring the performance of the perl regex engine. You are just measuring its behaviour on a single pathological case where the regex engine breaks down. Again, this is like evaluating performance of different cars by drawing them in mud. This is not a useful comparison.
If you want to measure the performance of perl regex on different architectures, why not take a complex non-pathological regex that will use many of the perl features, for example this one, and run it on a corpus of several thousands or million of email addresses.
P.S. I will even make it easy for you, here are some sample email addresses I randomly generated, along with a complex regex to match them
Last edited:
Of course it won't predict general regex performance. This regex is a prime example of how not to write regexes. What's the point in benchmarking a regex that no sane person will ever use? The only utility is for the regex engine developer, to identify pathological cases and fix them.
In addition I'd keep the Intel Mac if I want a free winter heater.
I think that ideal is over sold. True for laptops that are on battery that sort of performance is awesome and its one of the best features that I enjoy on my mac, BUT when doing hard core, or intensive work, most everyone is doing that while plugged in.
As a for instance, the performance per watt is really meaningless when you're using an iMac/Mini/Studio. The more important metric is performance,
If you're not sufficiently impressed by Apple Silicon's performance, please watch this!
And remember, Apple wasn't even in the chip business 15 years ago.
And remember, Apple wasn't even in the chip business 15 years ago.
Last edited:
In places where the kWh is 2x what you're used to and median household income is 2x less than what you make then power consumption is a concern whether it be desktop or laptop.
Some of use use our Macs 12hrs/day for 3,650 days in these places.
Not to mention our carbon footprint using 14nm Intel chips?
Last edited:
iPhone made it possible. If Apple stuck to Macs as their main profit driver then they'd be a die shrink or 2 behind Intel until now.
Last edited:
I concede your point. I looked at the cost of electricity in the EU and the cost difference between 2020 and 2022 is horrifically high and I can see it being a major incentive having a laptop sipping electricity.