Actually that is specifically what Cinebench R23 on Apple silicon doesn't do.
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.
Cinebench R23 score of M2 Max.
- Thread starter Sam_The_Apple_Boy
- Start date
- Sort by reaction score
It's fine for comparisons between the same architecture. What I'm saying is that is not accurate for comparisons between architectures. If you use it to compare between x86-64 and Apple silicon you will get misleading results. Promoting it as a useful comparison to people who aren't aware of the differences in optimizations between platforms means that you are likely misleading them. I don't see why anyone would want to do that when there are better cross-platform tests.
I believe @maflynn is directing his comments at me because I once quoted him on it. I'm the #1 anti-Cinebench-as-a-general-purpose-CPU-benchmark commenter on the internet.
Cinebench should not be used as a general purpose CPU benchmark like what the internet is doing right now. I believe Cinebench is quite toxic to the CPU community because it forces Intel and AMD to optimize their CPUs for it even though it's a niche of a niche benchmark.
There was no one person I was directing it to, just my general observation that when cinebench is mention, there are people that seem to want to come in and poo-poo the idea of it.
As they should if Cinebench is mentioned as a cross-platform comparison. Trying to correct misleading information is a good thing and not something to be ridiculed.
How could the multi core performance reach upto 24-25k? Are you talking about geekbench?
Yes, we already know it will have up to 96GB (4 x 24GB).
No, you just don't get it. cinebench is simply not optimized for apple silicon, and therefore results are not meaningful, period. It is one thing to be able to run on apple silicon, it is a completely other thing to be optimized for it. an example, one clown on macforums claimed that Apple silicon was not as good at running Handbrake because he refused to use the optimized tools via Videotoolbox becuase they were hardware encoder/decoders, but what this clown didn't factor in is that Intel silicon also has embedded hardware encoders/decoders which he allowed to be used. so optimized for intel, not optimized for Apple silicon - whoah that is meaningful - Not, Lakshimash.Cue the remarks about Cinebench being a horrible product, 3,2,1
So that is the point about cinebench. Take the time, use some professionals who actually know what they are doing and use an optimized product for both venues, then the results are meaningful. Otherwise it is a little like complaining that your new racing bike is no good because when you walk it up hill it is no faster than your old mountain bike.
I very much doubt that Apple will go with less than the current M1 Max (32GB) but 64GB and 96GB is a given. Perhaps offering a 48GB option.
Siri will get a big update rather soon.
New technologies are being tested as we speak, some have already been available in limited form to the public for some time, like „smart“ replies, were Siri tries to know when to respond to what in what way. Also context understanding, like asking follow up questions, is being worked on and will probably soon be rolled out to the (english speaking) public.
Siri will receive at least one cornerstone update this year.
Also well get more profound Home Screen customisation with the next iOS release.
Leak or nah, you decide. But I am as certain as if my name was Kuo.
This comment will age better than whatever OP is posting.
It’s not hard to predict, but it’s hard to make up accurate information.
So don’t do it.
New technologies are being tested as we speak, some have already been available in limited form to the public for some time, like „smart“ replies, were Siri tries to know when to respond to what in what way. Also context understanding, like asking follow up questions, is being worked on and will probably soon be rolled out to the (english speaking) public.
Siri will receive at least one cornerstone update this year.
Also well get more profound Home Screen customisation with the next iOS release.
Leak or nah, you decide. But I am as certain as if my name was Kuo.
This comment will age better than whatever OP is posting.
It’s not hard to predict, but it’s hard to make up accurate information.
So don’t do it.
ah I see where it is from not sure how accurate some things are.
nanoreview.net
like same memory bandwidth really? after the normal M2 upgrade I would have expected it to be higher for example.
No LPDDR5X 🤔
Apple M2 Max: benchmarks and specs | NR
Performance tests in benchmarks and full specifications of Apple M2 Max with 12 cores, 3500 MHz.
nanoreview.net
like same memory bandwidth really? after the normal M2 upgrade I would have expected it to be higher for example.
No LPDDR5X 🤔
Last edited:
Agreed, but I'll up the ante. I am perfectly willing to say that Cinebench is a horrible benchmark. It measures one very narrow and mostly irrelevant thing: how fast Intel's Embree raytracing library renders exactly one canned scene (supplied as part of the benchmark).
People treat Cinebench as if it's an authoritative test of how fast your CPU will be at a wide variety of real world tasks, but it just isn't. Even if you set aside all the concerns about differing optimization levels, do you use your computer to raytrace all day and nothing else? With an Embree-based raytracing app? If you do not, you should not care about Cinebench results.
It's not even a great benchmark for Cinema 4D users. If you're an artist, rendering is not even close to 100% of what you use C4D for, and there's a decent chance you never even use the built-in CPU rendering at all. (C4D supports external renderers, including Maxon's own GPU-based Redshift renderer.)
Benchmark suites like SPECCPU or Geekbench CPU, which consist of many subtests covering a wide variety of tasks people ask CPUs to do, with a geometric mean scoring system, are far better general-purpose benchmarking tools. If you want one number, at least that number represents a wide variety of tasks, and if you want to drill down further and look at specific tests which correlate more closely to things you do every day, you can do that.
10 performance cores?
I'd be disappointed. Where's ARMv9 with SVE? Although to be fair, they'd probably use the Sawtooth cores from the A16.
This seems plausible. The M2 base is also 1398 MHz. And for memory, the M1 Max has the same LPDDR5 and the M2 Max supposedly does. However if it's the same clock speed, I'm worried it's not the ray-tracing GPU from the A16. They probably just copied the clock speed from M2, they couldn't measure it.
Not sure this is accurate.
I'd be disappointed. Where's ARMv9 with SVE? Although to be fair, they'd probably use the Sawtooth cores from the A16.
Last edited:
I wouldn't take anything from a site like that at face value. It could all be made up, how could you know?
But really I just wanted to mention that v9 does not imply SVE, it's still an optional extension. When Arm v9 comes to Apple's CPU cores, they could choose to not use it, same as they did in M1 and M2. (the SVE extension was already available at the time M1 was in development)