blastdoor

About

Username
blastdoor
Joined
Visits
338
Last Active
Roles
member
Points
6,913
Badges
1
Posts
3,878
  • M4 24-inch iMac vs M3 24-inch iMac compared -- Apple's iconic Mac gets a speed boost

    zeus423 said:
    The 27” iMac is iconic. This one, not so much. 
    It really does seem like Mac mini + Studio Display is the replacement for the 27" iMac. The Mac Studio + Studio Display is the replacement for the 'iMac Pro'
    watto_cobra
  • Apple is researching smart glasses with an internal study, report claims

    Rick601 said:
    Wonder how many focus groups Steve Jobs used to develop the iPhone?  Asking for a friend. 
    Yeah… 

    Using an employee survey to design a new product sounds like something HR at my employer would do. To be clear, I don’t intend that as an endorsement of the idea. Our HR couldn’t design their way out of a paper bag. Maybe that’s because my employer would be surveying dimwits like me — maybe a survey of apple employees would yield better results. 


    watto_cobra
  • First M4 Max benchmark tears apart the M2 Ultra Mac Studio

    How does this compare to what’s available in the PC world, whether Intel/AMD or the more comparable ARM based Snapdragon?
    The M4 Max score is quite a bit ahead of Intel CPUs on both single and multi-core performance on Geek Bench 6. 

    https://browser.geekbench.com/processor-benchmarks

    I'm sure NVIDIA GPUs are still faster, but it's a bit more difficult comparing them since they don't really run the same benchmarks. It's also pretty hard to compare AI engines for the same reason. 
    It’s also ahead of the fastest non-server AMD chip, the 9950x:

    https://browser.geekbench.com/v6/cpu/compare/8593852?baseline=8608283

    the only sub-test where the 9950x wins is ray-tracing, because the 9950x has 32 logical (16 physical) cores, and ray tracing scales incredibly well with more threads. But ray tracing is an outlier in that regard. Otherwise, the m4 max wins big.

    note that for multithreaded workloads that scale well, AMD’s threadripper and Epyc products are far ahead of Apple because they have a ton of cores. But they are also incredibly expensive and use a lot of power. For example: 

    https://system76.com/desktops/thelio-mega-r3-n2/configure

    you can easily configure a system that costs over $20k with a 96 core cpu. 
    CalamanderwilliamlondonAlex_Vwatto_cobra
  • First M4 Max benchmark tears apart the M2 Ultra Mac Studio

    Effectively two M4 Max chips with an interconnect and therefore double the cores, the score should also be about twice as high as the M4 Max.
    This is almost certainly wrong. Many of the sub-tests in geekbench do not scale well across cores (though some do). If you look at M2 Max vs ultra, you don’t see a doubling of the overall score.

    That’s appropriate for many users whose workloads don’t scale well with more cores. But people buying an Ultra presumably do have workloads that scale well with more cores, so the headline GB score isn’t helpful for them.

    Alex1Nwatto_cobra
  • Early Geekbench scores show M4 Pro may outpace Mac Pro with M2 Ultra

    When you look at the subscores there are many where the M2 Ultra leads. Part of the issue is that not all components of the GB suite scale perfectly with more cores. That’s fine because that’s reality — some things scale well with more cores and some don’t. 

    But it means you need to really understand the details to know what to buy. Presumably people who buy the Ultra have workloads that scale well across many cores. For those folks, the M2 Ultra might still make more sense. 

    I think the comparison to the m1 ultra is more interesting because the m4 pro is at worst tied (more or less) on some tests but leads by large margins on others. So, at least for cpu, the m4 pro clearly beats the m1 ultra. That’s what really impresses me.
    ForumPostlotonesbeowulfschmidtwatto_cobra