The closest is the AMD U series of chips, but even more efficient and thus able to be in a tablet.
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 M1 CPU & GPU speed is very disappointing
- Thread starter Appletoni
- Start date
- Sort by reaction score
- Status
The closest is the AMD U series of chips, but even more efficient and thus able to be in a tablet.
Interesting Passmark benchmark results mirror Stockfish results so the workload must be similar.
https://openbenchmarking.org/test/pts/stockfish
Hi!
I was not saying M1 IS a Core Y. I was saying that it fills the equivalent space in Apple's lineup that the Core Y does in Intel's. Nothing to do with performance but more in terms of what slot in the product mix it occupies.
I’d agree that Y is not a bad analog. But I would also agree with the other posters that it is difficult to do direct analogs not only because of the M1’s particular performance and power characteristics but also since AMD and Intel segment their product line up far more than Apple is likely to do with its chips.
Of course. M1 is disruptive precisely because it is a 15W part that is HIGHLY performant and either beats or stands toe to toe with parts drawing 6-7 times the power. It has created the weirdness of units like the MacBook Air being in the same performance class as specced out MBP 16s.
I don't know on that, but I use passmark as pretty much my only benchmark I look at as it fits more with my own experience, and I really only use it as comparison between PC's I'm purchasing.
Passmark is known to be one of the least reliable benchmarks in the industry. The array of tests they perform is extremely primitive, does not stress the CPU enough and their methodology is crap as well. If that is the test you primarily rely on, then some of your opinions start making more sense to me.
As you can see if Apple increases the performance cores from 4 to 8, the M2/M1X (20M) will be slower than the 4700U.
You didn't read all of what I said -- I told you when I use it, and that's the only times I use it -- I hate benchmarks with a passion and only go by my own experience normally.
No, I got it, and I understand what you mean. Passmark is still pretty much non-informative though. It is already very difficult to make informed purchases and I don't see how passmark's arbitrariness makes it any better...
I find it a good source for comparison shopping, and that's it, and always using the same one keeps things consistent. It seems to work out well for me, and your description actually makes sense why -- it's all I'm looking for, just a general idea of how fast it's going to feel when doing normal tasks. For our office PC's, graphics based benchmarks make no sense, and neither does single thread or stress benchmarks. (Though I might run any of those types on a specific PC itself if I suspect some kind of intermittent problem)
And I mean general, a passmark of 15000 isn't enough different than a passmark of 13000 for that to be the determining factor, so something else will make the difference.
It's a game that comes with every Mac. Have a try, maybe you'll defeat the game when running on M1 due to such low performance on Chess ?
And...? What is the point of your posts?
Are they to show that the M1 is not universally great at running every single application or benchmark on the planet, irrespective of whether they have been designed to run well on the M1? OK - it's a valid data point, so, err...thanks for your input. You still haven't demonstrated whether this is due to a failure on the part of the developers to create code optimised for the platform, or some inherent flaw in the processor, which seems to be the argument you are putting forward...
These results only matter if these are the applications you need to run, in which case, the information would be useful to you.
Fortunately, for the vast majority of existing MacOS applications, the M1 Macs run extremely well and most users are delighted with the performance, low temperatures, inaudible fan noise, and excellent battery life for the laptops.
[Addendum: to be fair, these kind of tests, when performed without some ulterior motive to re-enforce a subjective prejudice, are valid and useful. In my career, I have run specific benchmark tests against hardware to determine its suitability to run the required workload, .e.g. running Java benchmarks against a server to be used for running Java middleware. It doesn't matter if it runs other applications better or worse - it only matters if it runs what I need with acceptable performance.
These tests are becoming even more relevant with the growing availability of ARM-based cloud services. Some applications run extremely well on ARM and the price-performance ratio is compelling. Others are not so good, especially those that use features only found in specific Intel / AMD CPUs. You need to test before deploying in production so that you know how well things will run until real loads.]
Last edited:
No, I get it. It gives you certain peace of mind. I have no problem with that.
For more serious use though, Passmark massively overestimates the multi-threaded performance of most CPUs because the tests are too simple and too short to be of much relevance. What they measure is the multi-core performance under full turbo. It is trivial to verify too: for example the multi-core score for i5-11600KF is almost exactly it's single core multiplied by the number of cores (3409*6 ~ 20500 vs. 20103). You will never get this kind of scaling in a real world workloads that last more than a split second...
And that's the case for almost every CPU on the list, except the server parts with very high core count (as their max turbo is severely limited compared to consumer CPUs). This leads to weird artifacts like AMD mobile CPUs being on par with the desktop CPUs.
I've never used any benchmarks for purchasing server class hardware. Our main server is not even in the same class as anything else, it runs a Power9 processor and for what it does, it's about %1000 over-provisioned even though it's the lowest end version. Nice database server!!
That's not what I was trying to say. My point is that Passmark massively overestimates multi-core performance for most CPUs because the tests it runs are too short to take into account that most CPUs will significantly drop the core clock mere milliseconds after a demanding workload has started. It's certainly ok for eyeballing the general performance class for a CPU, but it's misleading for everything else.
And that's all I use it for. Benchmarks really don't mean anything to me other than that as it's just not important for the job I do. Think of me as a systems programmer for the predecessor of that power9 machine, that was the last man standing after the last RIF so now my main job is to keep users happy and everything running.
Single-core on 11th gen had a huge improvement in single-core and the i5 has six cores so no surprise it would beat M1 in multicore. The single-core scores on 11th gen are very close to M1. The thing to remember is that the M1 is the weakest CPU that Apple will make in the Apple Silicon line and they are poised to double processing power for most of their range and provide quadruple and even double quadruple for their high-end systems.
Here is a different picture in SPEC2006 and 2017 beating Ryzen 9 5950X, i7-1185G7 and i9-10900K in several tests.
"Apple’s performance is extremely balanced across the board, but what stands out is the excellent 502.gcc_r performance where it takes a considerable leap ahead of the competition, meaning that the new Apple core does extremely well on very complex code and code compiling."
The 2020 Mac Mini Unleashed: Putting Apple Silicon M1 To The Test
- Status