Thanks. Your remark, the one I quoted to post that response, made no mention of a MBA from 2014.
That said, it would have been useful to learn how long a 2020 Intel MacBook Air (or Pro) would have taken for those tasks.
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.
Other Starting to appreciate some bits of the Intel-Mac era.
- Thread starter TheLion01
- Start date
- Sort by reaction score
Thanks. Your remark, the one I quoted to post that response, made no mention of a MBA from 2014.
That said, it would have been useful to learn how long a 2020 Intel MacBook Air (or Pro) would have taken for those tasks.
I know it's something you feel passionately about, but come on, if someone steals your MacBook they're not going to be kind enough to unplug the Time Machine drive you have plugged in an leave that with you...
“There’s no endemic issues” addressed that. Reading comprehension is fundamental.
Again, in the previous post (not the one you quoted), I mentioned that all benchmarks (Cinebench, Geekbench, Passmark etc.) show a significant performance gap in both single-core and multi-core (which is the one that is relevant in my case) between M-series and latest Intel Macs. Naturally, the gap is larger when considering Pro and Max variants as well as when considering later M-series generations as opposed to earlier ones, since each generation brings along some performance improvements over the previous one. For example, to compare AS macs with Intel Macs: Geekbench 6 Mac benchmarks.
There is really no comparison... Of course, one could argue that Geekbench favors AS over x86 architectures.
OK then. To compare AS macs with current and older Intel and AMD CPUs used in PCs and macs: Cinebench 2024 multi-core...
CPU Performance Rankings Cinebench 2024 Multi Core - Last updated Apr 2024
Latest CPU Performance Rankings: Including Cinebench R23, Passmark, and Geekbench 5 Blender Passmark single-core and multi-core benchmark scores
www.topcpu.net
...and Cinebench 2024 single-core:
CPU Performance Rankings Cinebench 2024 Single Core - Last updated Apr 2024
Latest CPU Performance Rankings: Including Cinebench R23, Passmark, and Geekbench 5 Blender Passmark single-core and multi-core benchmark scores
www.topcpu.net
As you can see, even M2 Max beats Intel Core i9 13900H and 13900HK in multicore (M3 Pro 12C is slightly faster, 1060 instead of 1050) and M3 Max 16C beats the HX versions of Intel and AMD! Also, in single-core, M3 Max is #1 and M3 #3.
Passmark suggests that M3 Pro 12C is on par with Intel Core i7-13700H and M3 Max 16C beats most HX CPUs from Intel and AMD!
Conclusion: M3 Pro/Max and even M3 are on par or better in many cases than equivalent-tier, latest Gen CPUs from Intel and AMD and certainly much faster than older Gen Intel CPUs.
And here's another one where the reviewer compares 27 laptops (one is a MBP with M2 Max) both on battery and plugged-in. Again, M2 Max is better than Intel Core i9-13700H in terms of multicore performance when plugged-in and almost up there with Intel Core i9-13900HK (about 5.5% slower); other than that, it loses from the HX CPUs, which are in the M3 Max territory. What is interesting, though, is that when on battery, M2 Max preserves its performance and comes #2 in total, whereas Intel and AMD laptops drop it by 30%-40% (old news, of course, but it is nice to see it validated via testing).
Laptop CPU Performance in Cinebench 2024 - Jarrod's Tech
This is my collection of Cinebench 2024 single and multi core scores from my laptop reviews. You can use this benchmark data to get an idea of the CPU performance to expect from a particular laptop / mobile processor. Testing Methodology “Plugged In” results were tested with the charger...
jarrods.tech
LPDDR4x/5 is always soldered and not available in DIMM form from anyone. There is a technical reason for this, although I don’t have that answer readily handy right now. It might be trace lengths or need to be close to the CPU/SoC for maximum benefit.
Can you do stuff like create a bootable usb windows installer for x86 systems off it? I'm in IT, and have all sorts of weird things that I need to take care of. I guess I could just google this!
Mojave was my favourite OS since Snow Leopard - I did like the visual updates in Big Sur (although it took a while), I found macOS peak performance was Mojave in recent years.
The hard part for you will be the throw away mentality of Macs now in the Silicone era. They are all sealed, all bound together, so upgradeability is out the door, which happend once they stopped making Retina MBPs.
If you can get your hands on a 2013 Mac Pro and get some bits for it for expandabilty, it was a wonderful machine for me.
Good to know, and Apple may adopt this standard by the M4, as this is way too new to have been engineered into the M3 or anything prior. This is actually great news as this would make a 1.5TB capable Mac Pro an actual product. I can see some devices remaining soldered with higher end devices getting expandability.
It may be at least another 1-2 years before this becomes a real possibility though.
EDIT: If the total max memory is 128GB as suggested by the article, then it’s a non starter for the Mac Pro, Studio and the same issue remains.
They don't mention it in that article, but while CAMM modules would be replaceable, it wouldn't be as easy as replacing SO-DIMMs. They're screwed in, and the boards they're screwed to may not allow jumping from 16 to 32 GB, for example.
I believe they can be pre-ordered now (pre-orders were due to have begun), but I can only see many companies jumping on the bandwagon if it's cheaper to manufacture than soldered RAM. Apple are more likely to pay more to manufacture devices we can't tinker with, or deliberately use CAMM boards that prevent upgrades.
This was introduced on September (a couple of months prior to the commercial launch of M3 series macs) and only by Samsung AFAIK .
Modular LPDDR5 RAM is still not a commodity. Maybe next year.
Although I don't think that Apple will adopt this solution; it will rather stick to the full SoC implementation (incl. RAM). But that's another story.
The optimal solution for me would be a combination, but I don't know if this is possible: unified memory integrated on the SoC by default (with different options, e.g. 18/36/48/64/96/128GB) for faster memory access and two RAM slots to enable expandability of the memory later on by strictly using RAM of the same type as the one in SoC (e.g. LPDDR5-6400 or, in the future, LPDDR5x-8533).
Also at some point, Apple should introduce dedicated GDDR6 VRAM for the iGPU, to fully unlock the potential of its recent advancements in thus department. Again, I am not sure at all that this will happen.
It was introduced by Dell in 2022, but Samsung announced they'd be making it in September '23. It's not proprietary, and is expected to be the de-facto standard before long. I sadly don't see Apple adopting it, as I said, even if it cost them less and worked as well. I also don't see them providing expansion slots, or dedicated VRAM.
A: 100% wrong that it is only/primarily intended for servers.
B: 100% wrong. It should be cheaper soon.
C: It's complicated.
Thought I'd resurrect your ridiculous unsubstantiated claims, after doing some reading on socketed LPDDR5 today.
Happy to reciprocate. I disliked how you smugly claimed nonsense, and doubled down on it, boorishly.
I’m aware of the benchmarks. Quite, thank you.
What standard benchmarks do is reveal how well the tested system performs on the algorithms of that standard benchmark. Those algorithms are not a substitute for everyday application.
That is precisely why I asked you whether you had tried your specific, non-benchmark-oriented tasks on a late Intel Mac, ideally one just preceding the switch to Silicon — that is, tasks applied in real-world, everyday situations. That ends up, over time, revealing performance quirks, bottlenecks, etc., which don’t always get dinged or scored by the usual benchmark utilities.
That is partially true. Partially because, for example, Cinebench runs a full-load test for 10 mins to consider thermals as well and also tests the GPU. On the other end, Passmark tests operations like integer and floating-point math or tasks like compression etc. which are indicative of CPU performance.
Also only partially true because both systems (Intel and AS macs) are tested under the same benchmark and their evaluated accordingly, it is not that we use Benchmark A for Intel and B for AS; it is apples-to-apples comparison. And AS macs constantly outperform Intel Macs in all benchmarks and comparisons. That is the objective reality. Now, what is the real-world situation which would work against AS macs but, magically, in favor of Intel Macs, despite that all benchmarks strongly suggest otherwise? Please give me at least one example.
The 2014 MBA is the latest Intel Mac I have. That's why I wrote that this particular comparison isn't much useful (although we did not need anyway that comparison to validate the obvious performance gap in favor of AS).
Why don't you buy an AS Mac and compare it with a similarly-spec'ed 2019 Intel mac? 😎
You already answered my inquiry a few replies back, cheers.
I don’t use Matlab and Mosek, nor do I have your reference/testing work files for those, so this (unhelpful) suggestion is kinda moot.
Moreover, I just plain avoid T2-equipped Macs.
You could test it based on your workload, programs and files, just to validate that the Intel CPUs used previously in macs are indeed slower than AS. It would be ike reinventing the wheel, of course, since this info is already out there and has been given in quantifiable metrics.
In summary: AS macs are much faster than Intel macs. On the downside, they don't provide upgradeability and are limited in repairability. Everyone can weigh those factors as he/she believes and choose accordingly-different options for different persons is a good thing. What is not a good thing, is to present the selection of an Intel mac over an AS one as a no-brainer, like the AS macs have no gap to fill in or reason for existence; their impressive performance gain over their predecessors alone is a good enough reason for some people. It is also certainly not cool to use characterizations like SillyMacs which, implicitly, extend to their users as well, simply because you want to feel good with your purchase decisions.
New macs have pros and cons, same as old macs. This idolization of the past becomes boring really fast.
This thread would be more interesting and productive if it was centered around the ridiculous upgrade prices for RAM and storage or the business segmentation which pushes users that don't know better to buy overkill machines or, in the other extreme, to give a ton of money for an underspec'ed machine with a RAM bottleneck and limited storage capacity. Instead, we are talking about the objective reality of AS being faster than older Intel chips, when it is faster than many high-end 13th Gen Intel CPUs and Zen 4 AMD CPUs.
Cheers.
If someone steals my MacBook it is going to be from somewhere other than my secure office that the Time Machine drive is located at.
The two are only in the same location while I’m at work.
Yes it’s just writing to an external USB. That works just fine
For the extreme vast majority of computer users, the delay is between the user's brain and the keyboard, and not anything with the computer itself. Almost all Intel era machines (except the butterfly era) are wonderful general purpose office tools to write e-mails, letters, form letters, mailings, light excel spread sheets, etc. They are also acceptable for very casual personal photo editing (think cropping, or changing colors in Photos).
Is there value in owning an AS machine? Absolutely! But owning an early intel machine (or a PPC machine, or an 680x0 machine) is a joy onto itself, to get it to work, like an older car. Sure, it might not be fast, or flashy, but there's a certain joy that exists with the machine that no longer exists with AS.
Is there value in owning an AS machine? Absolutely! But owning an early intel machine (or a PPC machine, or an 680x0 machine) is a joy onto itself, to get it to work, like an older car. Sure, it might not be fast, or flashy, but there's a certain joy that exists with the machine that no longer exists with AS.
That reminded me of this test. It shows that some modern systems are many times slower than old ones when it comes to the most basic functionality: text entry. Using a cheap-as-chips Windows machine at work, I can certainly appreciate it!
I really (really!) wish a company would develop a word processing "computer" with very low latency. Get it as fast as an Apple 2, and I'll be happy.