M4 Pro Unbinned is good, but I'm keeping my M1 Max MBP

[daneoni]

I wasn't planning to, but i'm doing the opposite. M1 Max -> M4 Pro.

The M4 Pro got significant upgrades this year, and now beats the M1 Max in pretty much everything besides those specific workflows that heavily rely on the Media Engine.

M4/M4 Pro mini and MBP are the MVPs of this generation, and will easily handle most people's needs.

Plus economically, the value of the M1 macs are now notably downtrending with consequent drop in trade in/sell values

So i'm likely gonna upgrade, and then wait for a redesign

 

[raythompsontn]

Even for 3-4 minute exports from Lightroom the fans are completely silent

My experience with Lightroom is that the CPU barely gets above 20% when exporting a thousand images. Importing itself is not CPU intensive. Building previews during, and after, the import is when my CPU is maxed out. The fans come on but are barely audible on my M4 Pro.

 

[ILoveCalvinCool]

I have no idea where this “unbinned”stuff seems to be coming from on the Mac side. But it makes zero sense and is extremely confusing nomenclature.

I get the impression it's a British thing. Whenever I click on the profile of someone who uses it, they're in the UK. I've had it explained to me several times and it seems like each explanation is different. I still have no idea what it means, other than "binned" seems to mean the same thing as "base model", and "unbinned" is anything else. I don't think the people who use the terms even agree on their meaning.

 

[callumrd1]

My experience with Lightroom is that the CPU barely gets above 20% when exporting a thousand images. Importing itself is not CPU intensive. Building previews during, and after, the import is when my CPU is maxed out. The fans come on but are barely audible on my M4 Pro.

Not my experience at all. Here's a screenshot of an export with 98% cpu usage:

And the fans never rose above their inaudible lowest speed.

Edit: this was exporting ~250 30-45 megapixel RAW photos to 4k wide jpegs.

 

[LogicalApex]

What's your memory pressure % during regular work use? I like to keep it in the green (at less than 50%) for regular usage, so there's headroom for intense use... and future proofing.

It stays within acceptable range for me of yellow or green the majority of the time and rarely in the red.

Why would you want to keep it in the green? Wasted RAM isn't all that useful so making use of it or not paying for excess is a better choice IMHO.

People say future proofing a lot around here, but I'm not convinced it makes any real sense honestly. Apple upgrade pricing is intense. Meaning, they are effectively charging you the same amount of money in advance to "future proof". Meanwhile, you're overspending on RAM and storage when you can't future proof the CPU, GPU, and NPU improvements that are massive YoY...

For instance, if I wanted to go up on RAM which would be nice for running even larger local LLMs I would need to spend another $1K to do so. $1K goes a long long way toward a new MacBook Pro in 1 or 2 years if I feel the improvements are worthy. If you consider the ladder jumps from the base model pricing you can quickly get to a point where you're buying the 2024 MBP multiple times over today to "future proof".

Buy what you need today and save the money you'd spend on "future proofing" and apply it to what you need when those needs materialize.

Just my $.02

 

[TechnoMonk]

It stays within acceptable range for me of yellow or green the majority of the time and rarely in the red.

Why would you want to keep it in the green? Wasted RAM isn't all that useful so making use of it or not paying for excess is a better choice IMHO.

People say future proofing a lot around here, but I'm not convinced it makes any real sense honestly. Apple upgrade pricing is intense. Meaning, they are effectively charging you the same amount of money in advance to "future proof". Meanwhile, you're overspending on RAM and storage when you can't future proof the CPU, GPU, and NPU improvements that are massive YoY...

For instance, if I wanted to go up on RAM which would be nice for running even larger local LLMs I would need to spend another $1K to do so. $1K goes a long long way toward a new MacBook Pro in 1 or 2 years if I feel the improvements are worthy. If you consider the ladder jumps from the base model pricing you can quickly get to a point where you're buying the 2024 MBP multiple times over today to "future proof".

Buy what you need today and save the money you'd spend on "future proofing" and apply it to what you need when those needs materialize.

Just my $.02

If it’s hobby generic device it makes sense. In my experience, my workloads at some point or the other keep using more RAM. I maxed out my M1 Max at 64 GB RAM, I know folks who upgraded because they went with 16 GB or 32 GB. My next MBP, I am hoping 256GB, I already push to 170GB RAM+Active swap. With unified memory being non uogradeable, I am not wasting money upgrading every time the RAM requirement goes up in years 2-5.

 

[ILoveCalvinCool]

Buy what you need today and save the money you'd spend on "future proofing" and apply it to what you need when those needs materialize.

But what I need is as much RAM as I can possibly get. I’m running LM Studio and Stable Diffusion. A base model is not an option.

 

[RumorConsumer]

what's your memory pressure looking like when working?

Totally nominal. Under 20%

I wasn't planning to, but i'm doing the opposite. M1 Max -> M4 Pro.

The M4 Pro got significant upgrades this year, and now beats the M1 Max in pretty much everything besides those specific workflows that heavily rely on the Media Engine.

M4 Pro mini and MBP are the MVPs of this generation and will easily handle most people's needs.

Plus economically, the value of the M1 macs are now notably downtrending with consequent drop in trade in/sell values

So i'm likely gonna upgrade and then wait for a redesign

this is exactly what I did, but the point of my post is that there was no appreciable performance difference that I experienced doing the stuff that I do.

 

[honcho]

It will depend on what your usage is like, but I'd suggest looking at Activity Monitor for your current workloads to get an idea of how much memory pressure you're experiencing. For most users, the lower RAM amounts are more than enough. Apple Intelligence currently isn't very taxing on RAM. Locally running LLMs are, but few users are running these enough to justify paying for the extra RAM to Apple versus using online platforms like ChatGPT or Claude. It is mostly the developer crowd like myself who are doing this.

While running a larger local LLM model you can see the memory pressure showing, at least for this model, I'm right sized memory wise for my M4 Pro.

View attachment 2455062

With the local LLM closed off and 20 tabs loaded in FF and over 30 tabs with multiple windows open in Safari, among other things, you can see my memory pressure is extremely low and I could get by easily for that workload on the 24GB base RAM offered on my 16" MBP.

View attachment 2455070

Thank you so much for the helpful advice. At the moment, I’m only really using my MacBook for general tasks and DJ software (Traktor/DJ.Studio) and occasional light DAW use. Looking at my current memory use, it seems to top out around 22GB. Memory pressure is green in Activity Monitor. However, that’s with 32GB memory and a discrete GPU. Although, I know MacOS will use more RAM if it is available.

My gut says to go for 24GB and I think your observations seem to confirm this. Given the generational advances of Apple Silicon, I imagine that I will be getting a new MBP in three years, so I am less keen on speculative upgrades. My workflow doesn’t really warrant it. As I value minimal fan noise, I am thinking that the 12-core M4 Pro would work well. The only extra I’ll need will be 1TB storage. That said, some of my software is GPU-accelerated so maybe the 14/20 is a better choice, as long as it doesn’t mean significantly more noise.

 

[RumorConsumer]

Thank you so much for the helpful advice. At the moment, I’m only really using my MacBook for general tasks and DJ software (Traktor/DJ.Studio) and occasional light DAW use. Looking at my current memory use, it seems to top out around 22GB. Memory pressure is green in Activity Monitor. However, that’s with 32GB memory and a discrete GPU. Although, I know MacOS will use more RAM if it is available.

My gut says to go for 24GB and I think your observations seem to confirm this. Given the generational advances of Apple Silicon, I imagine that I will be getting a new MBP in three years, so I am less keen on speculative upgrades. My workflow doesn’t really warrant it. As I value minimal fan noise, I am thinking that the 12-core M4 Pro would work well. The only extra I’ll need will be 1TB storage. That said, some of my software is GPU-accelerated so maybe the 14/20 is a better choice, as long as it doesn’t mean significantly more noise.

Don’t assume about three years. These things have seriously longevity. More than Intel ever did. I’d say really plan for 5. This is coming from somebody who always wants the best.

 

[Zirochka]

Yes, because of the manufacturing imperfections, different parts are placed in different "bins". Every part is put in the appropriate bin. They are all binned-- some in 12/16 bin and some in the 14/20 bin. No M4 Pro is "unbinned".

In other words, M4 Pro is not a defective chip? and what is the probability when you buy a binned chip but with a smaller number of cores (since they are binned but still on the board) that the remaining cores will also work well ? or some other core won't work?

 

[Alphazoid]

In other words, M4 Pro is not a defective chip? and what is the probability when you buy a binned chip but with a smaller number of cores (since they are binned but still on the board) that the remaining cores will also work well ? or some other core won't work?

The other cores will never be affected. The binned cores are physically turned off and won't interfere with anything. The remaining are rigorously tested before being put into products.

 

[Zirochka]

The other cores will never be affected. The binned cores are physically turned off and won't interfere with anything. The remaining are rigorously tested before being put into products.

Thank you!

 

[honcho]

Don’t assume about three years. These things have seriously longevity. More than Intel ever did. I’d say really plan for 5. This is coming from somebody who always wants the best.

It’s not the power that’s the issue. To be honest, if my Intel i9 weren’t so noisy or about to lose OS support, it would still be OK. It’s just that if the rumoured OLED MBP materialises in 2026/2027, with a new form factor, I know that I won’t be able to resist. M4 Pro now, M7 Pro in three years

 

[LogicalApex]

If it’s hobby generic device it makes sense. In my experience, my workloads at some point or the other keep using more RAM. I maxed out my M1 Max at 64 GB RAM, I know folks who upgraded because they went with 16 GB or 32 GB. My next MBP, I am hoping 256GB, I already push to 170GB RAM+Active swap. With unified memory being non uogradeable, I am not wasting money upgrading every time the RAM requirement goes up in years 2-5.

I'm not sure if you read my post, but my point is simple...

If you need the RAM or Storage or other upgrades in your workflow then they make sense to buy. That's not wasting money on future proofing that's meeting your needs with the device and "right sizing" to your workload.

Buying in anticipation of future needs though is a waste of money, but also suboptimal in general. You'll overpay for the hardware today and you'll have outdated hardware when that need arrives.

But what I need is as much RAM as I can possibly get. I’m running LM Studio and Stable Diffusion. A base model is not an option.

Again, if your workflow needs certain hardware then buying the hardware that meets that need is prudent behavior. That's not "future proofing" or wasteful.

I'm not sure I understand what's confusing here honestly.

Maxing out a M4 MBP, for instance, to have any future need met costs more than buying 2 base model MBPs for the 16" model. Meaning, instead of over paying of upgrades that user can instead upgrade their CPU twice as often and take advantage of the massive YoY performance gains Apple is adding to Apple Silicon instead. Which is more valuable than RAM that's sitting around doing nothing or storage that's sitting empty.

 

[RumorConsumer]

It’s not the power that’s the issue. To be honest, if my Intel i9 weren’t so noisy or about to lose OS support, it would still be OK. It’s just that if the rumoured OLED MBP materialises in 2026/2027, with a new form factor, I know that I won’t be able to resist. M4 Pro now, M7 Pro in three years

Ah. Same here.

 

[shenfrey]

I'm using an M2 Pro and it suddenly feels rubbish lol

 

[m1maverick]

keeping in mind reviews do show Max M4 ramps up fan much more but I doubt i'll even get close to what they're doing 5% of the time

I was disappointed to learn this. The M4 is a nice chip but it seems to be losing what made AS great...low power consumption (and corresponding lower heat generation). The M1 was amazing in this regard but subsequent chips are gaining their speed at the cost of power consumption / heat generation.

I was especially surprised when I watch a Max Tech video where the M1 was at a battery health of 83% and after all the tests was at 11% whereas the M4 was at 100% battery health and had something like 23% batter remaining. The M4 is fast but saving say, 35 seconds off of a 1:45 second task isn't going to save someone much time even if the percentage is 33% faster.

 

[RumorConsumer]

I was disappointed to learn this. The M4 is a nice chip but it seems to be losing what made AS great...low power consumption (and corresponding lower heat generation). The M1 was amazing in this regard but subsequent chips are gaining their speed at the cost of power consumption / heat generation.

I was especially surprised when I watch a Max Tech video where the M1 was at a battery health of 83% and after all the tests was at 11% whereas the M4 was at 100% battery health and had something like 23% batter remaining. The M4 is fast but saving say, 35 seconds off of a 1:45 second task isn't going to save someone much time even if the percentage is 33% faster.

Agreed on heat and I think they will find a way to mitigate it.

And the M4 series will save a lot more time than just a few seconds. On longer compiles, video, and image work people will save hours on more complex tasks.

 

[magbarn]

Is that what it sounds like? The specs say the M1 Max delivers 2300 in single core Geekbench while the M4 Pro returns 3940 (my machines number). That is an insane delta, and it was big enough that I wanted to know how it would feel if I upgraded. The answer - barely noticeable unless I was actually doing something with a progress bar which is infrequently.

The problem is Geekbench itself. My M4 mini blows away my prior M1 in single core. But for everyday tasks it feels just a bit quicker.

 

[TechnoMonk]

I'm not sure if you read my post, but my point is simple...

If you need the RAM or Storage or other upgrades in your workflow then they make sense to buy. That's not wasting money on future proofing that's meeting your needs with the device and "right sizing" to your workload.

Buying in anticipation of future needs though is a waste of money, but also suboptimal in general. You'll overpay for the hardware today and you'll have outdated hardware when that need arrives.



Again, if your workflow needs certain hardware then buying the hardware that meets that need is prudent behavior. That's not "future proofing" or wasteful.

I'm not sure I understand what's confusing here honestly.

Maxing out a M4 MBP, for instance, to have any future need met costs more than buying 2 base model MBPs for the 16" model. Meaning, instead of over paying of upgrades that user can instead upgrade their CPU twice as often and take advantage of the massive YoY performance gains Apple is adding to Apple Silicon instead. Which is more valuable than RAM that's sitting around doing nothing or storage that's sitting empty.

I do future proofing because historically the needs have gone up. I didn’t need 64 Gb when I bought my M1 Max, it replaced a 32 GB Intel MBP. For professional use I max out on RAM, it saved me from upgrading often. If it’s hobby, I could easily stick to base model. I know folks who were skimping on ram upgraded to M3 from zM1 because they didn’t have foresight for future growth.

 

[ILoveCalvinCool]

I'm not sure if you read my post, but my point is simple...

Your point being simple doesn't mean it's correct. Many people have to buy machines with an eye to their needs in a year or two. Good on you if you can buy a new Mac annually. Not everyone can.

 

[chars1ub0w]

I have a M1 Max 64GB 16" MBP. I really wish I had the M4 Max version. I'm estimating my program will take approximately 100-120 hours of M1 Max CPU. It has been running nearly two days straight now. If I can have the job finish a day or two earlier, that's a big win. And then I can run more computational experiments per month.

 

[magbarn]

I have a M1 Max 64GB 16" MBP. I really wish I had the M4 Max version. I'm estimating my program will take approximately 100-120 hours of M1 Max CPU. It has been running nearly two days straight now. If I can have the job finish a day or two earlier, that's a big win. And then I can run more computational experiments per month.

Geez, that's a lot of stress on a laptop, a M4 Ultra Studio might be a better use case for what you're doing.

 

[ILoveCalvinCool]

I'm using an M2 Pro and it suddenly feels rubbish lol

The new Studio cannot come soon enough. I can't believe how fast the M2 Pro went from godlike to agonizing.