The ONLY good thing about Apple Intelligence is base hardware gets increased RAM finally.
And I'm keeping AI turned off.
And I'm keeping AI turned off.
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.
M4 vs. M5 Chip Buyer's Guide: How Much Better Really Is M5?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Learn patience. Count your blessings. In a few months it will be fixed. Stop upsetting yourself.
I think it's faster but I don't believe Apple marketing anymore 
.
.
Apple needs to stop improving their silicon every year because it's angering too many people.
Lightroom uses GPU acceleration and for some tasks uses the NPU/Neural Engine (though it's been disabled for the Denoise function due to quality issues). The GPU improvements would be noticeable in the M5 over the M4, but both would be substantially faster than the M1 variants.
I mean that's how most chip manufacturers operate. When introducing a new generation, Intel and AMD usually release consumer or mobile chips first and the high end workstation Xeon and Threadripper chips are the last to be updated to new generation architecture because they're lower volume. TSMC likely doesn't have the fab capacity for Apple to update all of its chips and all of its devices at the same time, plus keep up with demand for A series chips.
Does anyone know what other "Tweaks" Apple makes for different products containing the M5 Chip?
I was fully expecting the Vision Pro with M5 to be VASTLY more powerful than the old original M2 version
And so it should be.
But I heard in a VR podcast that the story was that Apple have "Tuned" it more for power and efficiently as opposed to performance in the upgraded Vision Pro.
No-one actually has it yet, so we don't have any comparison reviews, but does anyone know any more?
An M5 may not just be an M5 because it's an M5 perhaps?
I was fully expecting the Vision Pro with M5 to be VASTLY more powerful than the old original M2 version
And so it should be.
But I heard in a VR podcast that the story was that Apple have "Tuned" it more for power and efficiently as opposed to performance in the upgraded Vision Pro.
No-one actually has it yet, so we don't have any comparison reviews, but does anyone know any more?
An M5 may not just be an M5 because it's an M5 perhaps?
Honestly, I wish something like the M series ran vSphere. I bet it would really kick ass.
The gains from M4 to M5 are, CPU-wise, just incremental. Not bad either with the best single-core score in GB6 of any consumer CPU, and a multi-core score near the M1 Ultra, which is a lot. But compared to the M4, it’s just an incremental 15% aided by the new process (N3P) and the new efficiency cores (e-cores), with much improved performance now.
However, while the changes in the CPU are good but just incremental, the improvements in the GPU side are much more significant. I suspect this is a whole new GPU core architecture, because both the graphics in general and Ray Tracing in particular are much improved.
But the biggest, dare to say almost disruptive upgrade is in AI operations, which apparently now will make an extensive use of the GPU thanks to the tensor cores. Now more than ever it is really tempting to equip our M5 machine with the maximum amount of RAM available…
However, while the changes in the CPU are good but just incremental, the improvements in the GPU side are much more significant. I suspect this is a whole new GPU core architecture, because both the graphics in general and Ray Tracing in particular are much improved.
But the biggest, dare to say almost disruptive upgrade is in AI operations, which apparently now will make an extensive use of the GPU thanks to the tensor cores. Now more than ever it is really tempting to equip our M5 machine with the maximum amount of RAM available…
The M4 is already very good but if you want to make a little long-term investment, I’d go with the M5. Not only it has much better graphics performance but also will handle any hypothetical future AI features much, much better.
Here’s a better answer from a fellow MR user.
Last edited:
Video encoders are "junk" -- unless you need them for your workflow.
GPUs are "junk" for many people.
NPUs are "junk" for people who don't use apps that access them.
Matrix-multiply accelerators are "junk" -- until you start needing matrix manipulation.
Efficiency cores are "junk" -- unless you want to conserve battery life when doing background tasks.
Apple solves many problems with each of their processor architectures. So does Qualcomm. So does ARM. So does everybody! Everything is a blend. If you obsess about single-processor speed, you would be better off inventing a time machine and going back to before 1999 when there were no multi-processors or GPUs. Or you could design and fab your own #!$$ processor. Good luck with that.
It is pointless to grouse about "wasted" real estate on any Apple general processor architecture. The M5 architecture does an amazing diversity of tasks amazingly well, and you have no idea when your needs may shift to another application. Appreciate the brilliance of its design and architecture. If you don't appreciate it, you can always go elsewhere.
I agree that most of us don’t even need to upgrade every five years. My base model M4 MacBook Air replaced a 10 year old Dell desktop. The M4 is amazing, but for my routine of web browsing, spreadsheets and letter writing, I don’t see any huge productivity increase.
I waited for the M4 because I wanted two external monitors and to be able to use the built in screen at the same time. Speed is a bonus. The microphone is probably a bigger deal for me than speed. Calls on my Mac are fantastic quality. It would be hard to go back to a pc.
Do I have a youtuber for you...
He does exactly what you are asking for. Compares M series SOC`s performance for lightroom etc.
This is the guy, Artisright, you want to checkout in regards to Apple and Lightroom performance. Hands down the best analysis in my opinion of the M series processors year over year.
https://youtube.com/@artisright?si=UGV0yFhoZV1ChVZA
What I would like to see is Apple collaborating with different actors to harness all the new capabilities of the M5 and future SOCs
- Apple collaborating with Local AI content generation platforms like Stable Diffusion and LLMs.
- Apple collaborating with EPIC to improve Unreal Engine performance on Mac.
- Apple collaborating with more Game Studios.
- Apple collaborating with Local AI content generation platforms like Stable Diffusion and LLMs.
- Apple collaborating with EPIC to improve Unreal Engine performance on Mac.
- Apple collaborating with more Game Studios.
My M4 Max studio never gets warm when compiling or heavy programming tasks. Even gaming with Cyberpunk (heavy GPU) or other games like FrostPunk 2 (heavy gpu and CPU) it barely gets warm. I have never heard the fans once.
Even when I was being super extra and pushing heavy duty local LLMs while compiling it got slightly warm but no fan noise ever. Pretty nuts
the age of single threaded performance improvements is largely over. Even when you look at M3 to M4 improvements whose single-threaded score increased 25% the scores were skewed by large improvements in a handful of AI tests making use of new matrix instructions like object detection (+114%) whereas code compilation was only up slightly (+14%). The M5 continues this trend where outside of AI tasks half the meager single-threaded improvement is explained by clock speed bumps with very little improvement made by the architecture itself. This is not a criticism of Apple, who is actually scaling single threaded CPU performance better than other chip designers ... this is just the end of an era.
1. Apple pushes a lot of open source code for local AI, see MLX: https://github.com/ml-explore, most LLMs have a MLX variant including stable diffusion. Check our LM Studio to very easily find them and try them out.
2. Unreal Engine 5 natively supports Metal now and games by default are set to target macs, unfortunately game studios often disable it.
3. Yes, please.
The new AVP is pushing 10% more pixels and maxing out at 120hz instead of 90hz. Also running AI workflows and non-AI apps at twice the speed of the M2 it is replacing. That said improving power and efficiency is important because the M2 ran hot on the AVP, something that goes on your face.
Yes-ish. Apple for os26 gave developers the option to use the built in AI models in macOS to locally do things. Any app that takes advantage of this will run much faster behind the scenes. So the answer is depending on developer support.
For example we are currently experimenting with using Apple's built in AI models to parse scanned in medical paperwork so we don't have to deal with privacy risks and HIPAA compliance issues.