They should do that once they have the specs of the MBPs. And what we are really waiting for is to see what the M4 Pro chip can do. Is it closer to a full “pro” chip like the M2 pro. Or is more like a consumer, but better chip like the M3 Pro.
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.
M3 vs. M4 Chip Buyer's Guide: How Much Better Really Is M4?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Good to see the improvement. But not going to notice the difference in day to day usage.
Interesting. I've always heard that we cannot really perceive a difference less than 2x, so upgrading before then is just a waste (ie 8 to 16GB RAM, or 2K to 4K screens, or 30 Hz to 60Hz, or a doubling of CPU power, which is why Moore's law focuses on that metric), so I'm curious how this 4x metric translates to real world product cycles.
From Geekbench:
M1 MBA Single Core: 2343
M4 (from this article: 3864
1.65x improvement: Indicates upgrading from an M1 to M4 is not recommended.
I replaced my 2015 Macbook Pro Retina with an M2 15" Air.
1174 vs 2595
2.2x improvement: At least is a perceptible difference (>100%) but nowhere near 4x improvement
The 2015 notebook was fast enough, and had tons of memory and storage, so I could have held out for a few more years, but the screen failed. I don't replace my devices frequently, if they still work (I am using an Iphone 8!) but I think technology products simply aren't made well enough to last until a new model is 4x faster than the old one.
I would target a 2x to 3x increase in specs before you upgrade--with an absolute minimum of 2x. But I don't see how you can practically hold out for a 4x (maybe with desktops you can, if you upgraded them with RAM and storage so they will work longer).
Looking at Geekbench, you have to go back about 10 years to find a Mac with single core scores 1/4 of the current M4. Non-desktop hardware made these days just won't last that long. (I am far more careful opening and closing my Macbook lid as the screen is probably the first part to fail).
Looking at phones:
Iphone 8: 1033
Iphone 16Pro 3391
That's only a 3x improvement and it's hard to still run an Iphone 8 today!
I think it’s possible to use Macs for a very long time. I have a MacBook 12 from early 2016 and it was my main laptop for anything that didn’t involve graphic design or video until this Summer, although a few months ago I needed to do some light illustrator work (some vector graphics posters for an event) and I was on a trip so that’s the only thing I had and it managed! Indeed, it wasn’t lightning fast, but it did a pretty good job. it is still capable of doing light work and I have used this laptop every day.
Same goes for iPads, they last. However, I feel like iOS/iPadOS age the hardware faster than MacOS. And it’s especially noticeable with iPhones.
What sucks is that no one puts m1 max on thier comparisons anymore. M2s were hot garbage, and m3 was made with even weaker comparisons.
But i havent seen anything on m3 vs m1 max or studio.
The worst thing tho..no noteworthy way to judge apples INTEGRATED, NON DISCRETE, SHARED RAM graphics. 8 cores. 10 cores. Mine has 24 cores. Its a great way to say nothing about performance. These dont hold up to typical pc amd/nvidia chipsets. Oh..but efficiency..what a ruse. Efficiency wont help you in after effects.
But i havent seen anything on m3 vs m1 max or studio.
The worst thing tho..no noteworthy way to judge apples INTEGRATED, NON DISCRETE, SHARED RAM graphics. 8 cores. 10 cores. Mine has 24 cores. Its a great way to say nothing about performance. These dont hold up to typical pc amd/nvidia chipsets. Oh..but efficiency..what a ruse. Efficiency wont help you in after effects.
[About the GPU]
It's apparently clocked 10% faster than the M3. That may be the limit of gains other than the dramatic RT improvement.
It's talking about the part of the chip that actually interacts with the monitor (or internal display). The display engine is a major part of Apple's "secret sauce" - they spend multiple times more transistors on it than do chips from Intel and AMD, in exchange for far better efficiency. That's why you can drive an external monitor from your MacBook Air for hours with no problem, while x86 laptops collapse.
The engine has received a major upgrade in the M4 generation. The M3 and previous were capable of driving two displays, including a laptop's internal screen. If you wanted more screens, you needed to move up to the M3 Pro. But now in the M4 you can drive two additional screens besides the internal display. Or, presumably, 3 external screens on the Mini, or laptops with their screens closed, though we won't know about the mini until later today, and the laptops until they're tested since Apple doesn't talk about 3 screens.
No, not even close. I wish...
Much of the M3->M4 performance gain is not IPC but clocks. This doesn't make the M4 any less impressive, but we should be accurate. Actual IPC is a fungible number since it depends on what you're measuring. Most of the improvements in the M4 come from the implementation of SME. I don't favor excluding measurements that include SME code when publishing benchmarks, but if you want to talk about how fast "regular" apps will go, the improvement is on the order of 5%, IIRC, maybe less.
M3 was pretty much the same story, sans SME.
Some general rules of thumb are quite useful. But the above is both wrong and facially ridiculous.
First of all, reductio ad absurdum: I can damn well tell the difference between a task that takes half a day and a task that takes an entire day. But that's not really "absurdum" - plenty of tasks (transcoding, scientific computing, AI, and many more) take on the order of hours (or days, or even more).
So the 2x rule is utter BS, even for time. As for other quantities, you completely fail to take into account, well, reality.
For example, for RAM, adding more RAM may have no impact, or it may speed a task up by 10x, 100x, or more. It depends on the "working set" of the application and use case. Your app/task needs a certain amount of RAM to function; if you have less than that free, you're going to spill to disk ("paging", often incorrectly called "swapping"). Depending on how often the entire app RAM footprint is touched, that paging could slow you down a few percent... or a few hundred times (basically, the difference between access speeds of RAM and SSD). And that break point could be anywhere. So taking your machine from 8GB to 16GB, or from 64GB to 128GB, could have no effect, or a huge one.
You are historically incorrect about Moore too.
Coming back to the time aspect though: Human perception is way finer than you think because it's not limited to immediate effects. If you use your computer for work and you're throughput-bound, well, time is money. Saving 10% of your time on frequent compute-bound tasks will save perceptible amounts of money, which will directly drive a decision to upgrade or not.
I have a feeling the M4 Ultra will be a true beast!
Really thinking of replacing my M1 Ultra, next summer.
Really thinking of replacing my M1 Ultra, next summer.
About the Iphone - the 8 was and is a great device. But htere are other points to think about; 5G, faster wlan etc. About Computers: you are absolutely right. If it is a little maxed out, it is enough for many years. And with the mbp's... yes - we have to be very careful with them or in case of some damages parts, be smart and get stuff from craigslist or ebay to replace them. I have a older thinkpad and the same problem with the screen - but thanks to linux and hdmi i still cam use them. The only remaining problems are just the support of software and getting parts to repair them.
People doing these tasks are probably running ARM on Linux. Macs just make trouble when dealing with servers and big data.
Okay, I see you have video transcoding in there (I assume video). Yes, but even video editors spend more time editing their projects than rendering them. The big issue is does the computer allow the editing software to keep up in real time while cutting, and make the UI feel responsive.
There's special media engines in the M1 Max and M1 Ultra which take care of the basics of playback, FCPX and Davinci Resolve are completely responsive. The last big step up was to an M2 Ultra as it bound the GPU units much more efficiently than the M1 Ultra. The M1 Ultra is no slouch either. I have one, and there's literally no need for an M2 Ultra or an M3 or an M4 to get work done.
This is all just specs chasing. If you own an M1 Max, you probably don't need to upgrade. The M1 Max is the best deal in terms of CPU + GPU power. Unfortunately, the refurb deals seem to have petered out.
PS. Somebody needs these computers. That somebody is someone who does not now have a modern and fast Mac. Maybe someone who bought an early M1 (not Max), or someone on Intel. Anyone on M1 Max or M2 Pro doesn't need an M4. We do need one more computer, I'm looking at these to see if 32GB memory version Mini will be a cost-efficient substitute for what would normally be an M1 Max work station. If you use 64GB or more memory, then an M1 or M2 Max with the memory you need would be miles better than an M4 limited to 32GB.
Your inability to imagine any workflows other than ones you're familiar with is preventing you from understanding this very basic concept. To repeat:
"Human perception is way finer than you think because it's not limited to immediate effects. If you use your computer for work and you're throughput-bound, well, time is money. Saving 10% of your time on frequent compute-bound tasks will save perceptible amounts of money, which will directly drive a decision to upgrade or not."
Computers for home/hobby use can have lifetimes of 5-10 years (though security is becoming a serious issue for machines that no longer get OS updates). Work computers are a different story. For throughput-bound tasks, the computers will be replaced when the value of the human time they save exceeds the expense of replacement.
I'm stuck now that the M4 chips are coming to the iMac, Mac Mini, and presumably tomorrow to the MacBook Pro's. I have and love my 2020 iMac tricked out with an i9, 64GB RAM and a 4TB SSD. I do lots of Lightroom and Photoshop and am struggling whether to get an M2 Ultra Studio with 64GB RAM and a 2TB SSD (4 was ridiculous) or await the M4 Studio in 2025.
I just read all that's currently available on the M4 Mac Mini and I'm tempted. The M4 Pro with 64GB RAM and a 2TB SSD comes in well under $3k. The M2 Ultra Studio is just over $4k.
I've even thought about using an M4 MacBook Pro docked to my NEC PA271 monitor.
Too many choices and always wanting that which is just over the horizon.
I just read all that's currently available on the M4 Mac Mini and I'm tempted. The M4 Pro with 64GB RAM and a 2TB SSD comes in well under $3k. The M2 Ultra Studio is just over $4k.
I've even thought about using an M4 MacBook Pro docked to my NEC PA271 monitor.
Too many choices and always wanting that which is just over the horizon.
The best deal right now is either the M4 Mini with 32GB of memory and 256GB drive (you can keep the internal as just a backup boot drive and boot from an external NVME 4TB drive) at €1149 or if you want to go up market, the M2 Mac Studio 64GB 60 core GPU 1TB drive for ~$4K. An ultra is much more powerful than any of these M3 or M4 Pros. The M4 Mac Studio Max is probably what would suit you best as a long term computer but it's six months away.
If budget is a concern the 10 core M4 Mini with 32GB of memory will get you a very long way. Or wait for the M4 Studio. Your iMac is a very nice computer. The only weak point for the iMac in creative work would be video editing, and even there it does fine for occasional edits.
The M4 Pro is probably what you want. Look carefully at benchmarks of your apps of choice (LR, PS) and the tools inside them (whatever filters, or other processor-heavy features you use a lot) to see if the many GPU cores on the Ultra will actually benefit you that means specifically: does the Ultra perform much better (not just marginally better) than the Max? Than the Pro? If they do, then the Ultra may still be worth getting, but you're giving up a lot in other areas.
If you get the M4, consider getting the 512GB SSD instead, and spending part of the the savings on a TB5 dock, like this one, and a nice 4-8TB SSD. Hopefully soon you'll be able to get a TB5 dock with an internal NVMe slot, but so far, only the Kensington dock is shipping.
You say that with such conviction! Even though it's mostly not so.
At 16P+8E, the Ultra *may* be faster at GB6 multicore than the M4 Pro at 10P+4E. I wouldn't bet a lot of money on it though- the Ultra scales much less well than the Max (or the Pro), and certain loads (like GB6MT) scale very sublinearly.
In fact, the M2 Ultra scores a bit over 21K, while the M4 (*not* the M4 Pro) scores over 15K. I expect the M4 Pro to do much better than that!
Of course the GPU story is different. At 60 or 76 cores, the M2 Ultra will crush the M4 Pro at some tasks. But not at others. The details of his workflow are of paramount importance. (If he were doing something with raytracing, the M4 would be a no-brainer... he isn't, but someone else might be.)
It depends. If someone does quite a bit of multitasking, an Ultra never slows down. A Max can get bogged down with multiple tasks. Plus as we agree, more GPU cores are great for video editing.
I have both a Max and an Ultra so I've had the chance to compare head to head.
I really don't see the Pro as a good option. It's as expensive as a low-end Studio and a lot less powerful, with fewer ports. If one wishes to go Mini, one should at least enjoy some economy with that. I don't think the expensive middle ground is a good place to be for value. Apple has finally offered a decent value in a computer (the 10 core Mini) for the first time in a long time – we should enjoy it. I'll probably pick one up for our office.
Any of these computers are fine for the original user. He should probably wait on the M4 Studios and then pick up a Max. If he's antsy, he should get the 10 core Mini now with 32GB RAM and 256GB drive and then sell it off to get the Studio when it drops.
I think the rule of thump mentioned (you only notice 2x differences in speed) is not meant for people who do intense work on their macs. It's probably a more useful rule for users like me: i use youtube, iPhoto with minimal editing, iMovie with minimal editing of videos up to 3 min or so, word, ppt, excel, the usual websites, sometimes scrolling through larger pdf's with pics (300 pages max), no games, word files with tables up to 150 pages is probably the max workload i ever see
. I'm working on a 2015 MBP and feel it's fast enough. So for me it makes sense to only consider 2x differences in processor speed. It's probably more effective to upgrade to a 15" M3 MB air with 16GB/2 TB than to a 14" M4 MBP with a 512 GB drive. Price is pretty much the same and screen size and SSD size is more important than 20% processor speed.However, is apple intelligence or AI in general a reason to consider 20% processor speed relevant?
I'd agree, if the M4 Studio were available. Once it is, you'll be right. Until then, it's situational.
I think that's good advice. But given his use profile, it may be smarter to just go with the Pro and not think about it again for 4-5 years.
Yes, that's exactly the point I was making.I think the rule of thump mentioned (you only notice 2x differences in speed) is not meant for people who do intense work on their macs. It's probably a more useful rule for users like me: i use youtube, iPhoto with minimal editing, iMovie with minimal editing of videos up to 3 min or so, word, ppt, excel, the usual websites, sometimes scrolling through larger pdf's with pics (300 pages max), no games, word files with tables up to 150 pages is probably the max workload i ever see
No. AI won't care about that.
I just ordered an M4 iMac. It'll be a nice speed boost from my 2017 i7 4.2. Going from a 27" screen to a 24" will be interesting. For the price, though, I could also buy a studio display and have 2 monitors, and that would still cost less than a similarly RAM'd and Storage'd M4 Mac mini (32GB RAM and 2TB SSD) with a studio display. Not as much processing power with the iMac, but the M4 10-core is more power than I need anyway. The most intensive work my computer sees is photo editing in LR and occasionally PS. I don't do any video editing. I'm sure that'll hold me for another 7 years.
Great breakdown @Hartley! Would be helpful to compile this with the original Apple Silicon guide for comparison as time goes on.
www.macrumors.com
Apple Silicon Buyer's Guide: Which Chip Should You Choose?
Choosing a new Mac often involves consideration of the many Apple silicon chips now on offer, so our comprehensive guide covers their generations,...
www.macrumors.com
There are still guys repairing Amigas from the 90s
Chris Edwards Restoration on Youtube
Find me someone, who can repair a macbook mainboard for a reasonable price 🤣. Of course, there are a few gurus - but finanially it makes no sense at all.There are still guys repairing Amigas from the 90s
Chris Edwards Restoration on Youtube