*** Apple event 10th November @ 6pm - One More Thing: Apple Silicon Macs Reveal ***

[Deleted member 209350]

The score may not be important but surely running 2 CPUs through the same benchmark gives us an idea of how they compare?

Its okay to compare processors if they are iterations of each other, like A13 vs A14 or 5900x vs 3900x. But when you compare a 3900x to an A14 thats when Geekbench becomes a terrible comparison tool because its nowhere near accurate enough

 

[Armageus]

The score may not be important but surely running 2 CPUs through the same benchmark gives us an idea of how they compare?

No because it's sadly the only cross platform benchmark, and no one has any idea how accurate it is when comparing different platforms

 

[englishpremier]

No because it's sadly the only cross platform benchmark, and no one has any idea how accurate it is when comparing different platforms

FCPX exists on all Mackbooks and Apple computers. That's why i said earlier I'd like to see a BruceX or similar test. If the export is in a codec that the T2 chip doesn't support, then you get pure results of the CPU and graphics card, for a true comparison.

 

[HACO]

People really need to stop using Geekbench in 2020. Its so **** as a benchmark and not an accurate representation at all

It's actually a very decent benchmark, comparing it with SPEC, it has a R^2 of over 0.99. The only issue is that it's short and therefore not for sustained performance, so it doesn't take into account thermal throttling.

Read more:
https://nuviainc.com/blog/performance-delivered-a-new-way-part-2geekbench-versus-spec

Also from Anandtech:
https://www.anandtech.com/show/16226/apple-silicon-m1-a14-deep-dive/4

There’s been a lot of criticism about more common benchmark suites such as GeekBench, but frankly I've found these concerns or arguments to be quite unfounded. The only factual differences between workloads in SPEC and workloads in GB5 is that the latter has less outlier tests which are memory-heavy, meaning it’s more of a CPU benchmark whereas SPEC has more tendency towards CPU+DRAM.​

Maybe you want to say SPEC is also isn't a decent benchmark, at that point the question becomes why you'd think that, and what do you consider a good general-purpose multi-workload benchmark?

 

[ChrisD.]

People say things on forums that they read from random poster on other or the same forums.

 

[Armageus]

Maybe you want to say SPEC is also isn't a decent benchmark, at that point the question becomes why you'd think that, and what do you consider a good general-purpose multi-workload benchmark?

Not necessarily a case of it not being a "decent benchmark", but is it relevant? A Single threaded benchmark isn't relevant for anything in this day and age.

 

[Somnambulist]

Are you all secretly students or is there a dastardly way into the edu store? HMU.

A few years ago I signed up to some Windows course via HotUKDeals I didn't even do but entitled me to a 3 year NUS Extra card for. Got a student discount on my iPad Pro from that (and 10% off every time I shopped at Co-Op, Amazon Prime student discount etc). Think I need to look for another one of those soon as it ran out in September!

 

[HACO]

Not necessarily a case of it not being a "decent benchmark", but is it relevant? A Single threaded benchmark isn't relevant for anything in this day and age.

Many workloads still are (and likely forever be) single threaded. For example, Javascript performance (for web browsing) is 100% single-threaded. Most programs also have a main thread that relies on single-threaded performance and just assign workloads to new threads as necessary.
https://en.wikipedia.org/wiki/Amdahl's_law

But again, multithreaded benchmarks will also say the same thing as the single-threaded ones on all modern platforms (Cinebench single and multithreaded scores have a R^2 of over 0.99 once corrected for the number of cores, same as Geekbench). All of these platforms handle multithreading very well. This really becomes a concern when you scale upwards of 100 cores on shared memory architectures. Not an issue at all on consumer computers.

Cinebench R23 is already ported to Aarch64 on macOS, so we'll see the results soon enough. Maybe once those results are out, the "Geekbench is bad" crew will turn into "Cinebench is bad" crew

 

[Deleted member 209350]

Many workloads still are (and likely forever be) single threaded. For example, Javascript performance (for web browsing) is 100% single-threaded. Most programs also have a main thread that relies on single-threaded performance and just assign workloads to new threads as necessary.
https://en.wikipedia.org/wiki/Amdahl's_law

But again, multithreaded benchmarks will also say the same thing as the single-threaded ones on all modern platforms (Cinebench single and multithreaded scores have a R^2 of over 0.99 once corrected for the number of cores, same as Geekbench). All of these platforms handle multithreading very well. This really becomes a concern when you scale upwards of 100 cores on shared memory architectures. Not an issue at all on consumer computers.

Cinebench R23 is already ported to Aarch64 on macOS, so we'll see the results soon enough. Maybe once those results are out, the "Geekbench is bad" crew will turn into "Cinebench is bad" crew

You just sound like a massive apple fanboy, do you own every product they dish out?

 

[ChrisD.]

You just sound like a massive apple fanboy, do you own every product they dish out?

He doesn't to me, he just seems to know what he's talking about when it comes to benchmark software and how CPUs are used in real life situations.

 

[Throrik]

I don't know much about CPUs and even I understand that a whole heap of things are still single threaded.

 

[HACO]

You just sound like a massive apple fanboy, do you own every product they dish out?

So I'm just going to assume that you didn't have an actual response to what I said about benchmarks, so you resorted to personal shots.

 

[Armageus]

Cinebench R23 is already ported to Aarch64 on macOS, so we'll see the results soon enough. Maybe once those results are out, the "Geekbench is bad" crew will turn into "Cinebench is bad" crew

I'd certainly be more interested in Cinebench results than anything shown so far

 

[GodAtum]

Photoshop wont be available for another year!?

 

[Throrik]

Photoshop wont be available for another year!?

You'll be able to run it via Rosetta2 until the Universal binary is available.

 

[HACO]

You'll be able to run it via Rosetta2 until the Universal binary is available.

The big big question is, whether Photoshop and other software run faster under Rosetta 2 compared to the Intel-based macs that they're replacing. As long as you're no worse off, there's nothing to lose by migrating to AS macs.

 

[Deleted member 209350]

So I'm just going to assume that you didn't have an actual response to what I said about benchmarks, so you resorted to personal shots.

It was more aimed at your last comment saying "Maybe once those results are out, the "Geekbench is bad" crew will turn into "Cinebench is bad" crew "

Which seems to insinuate certain things

 

[Broken Hope]

The big big question is, whether Photoshop and other software run faster under Rosetta 2 compared to the Intel-based macs that they're replacing. As long as you're no worse off, there's nothing to lose by migrating to AS macs.

Given the GPU is massively better than the integrated Intel ones, even under Rosetta anything that uses metal should be faster.

 

[HACO]

Given the GPU is massively better than the integrated Intel ones, even under Rosetta anything that uses metal should be faster.

But GPU performance isn't related to ISA as you only need to make sure everything is compatible, i.e. no emulation so I'm sure that will be fine. That's why Apple showing Tomb Raider on Rosetta during WWDC wasn't exactly a tough task for their chips.

CPU performance is what's mostly affected by emulation.

It was more aimed at your last comment saying "Maybe once those results are out, the "Geekbench is bad" crew will turn into "Cinebench is bad" crew "

Which seems to insinuate certain things

It's a common occurrence that people tend to dismiss any benchmarks that show ARM closing the gap with x86, or in this case, completely bridge it, as "not relevant", "not real", "not good", or "not representative" and they're just running out of benchmarks where x86 has any meaningful lead. In most cases it stems from either not doing the actual research into how much ARM-based chips have improved, or in some cases x86 (or Intel/AMD/PCMR) fanboyism that wants to pretend like only Intel and AMD can make "real" CPUs.

 

[Throrik]

The big big question is, whether Photoshop and other software run faster under Rosetta 2 compared to the Intel-based macs that they're replacing. As long as you're no worse off, there's nothing to lose by migrating to AS macs.

Hopefully the emulation is going to be good, but that is something that people will be able to see soon, hopefully some review units are out and reviews come out fairly soon as this will be a big thing, whether the emulation works properly as advertised. I wasn't around for the PowerPC to Intel days with Rosetta so don't know what to expect.