Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
What we’ve seen with the A19 gives me hopes for the M5 to be more than just a minor refresh with higher clock speed. I was considering waiting for the M6 macs because of the big change in transistors and the technology behind the 2nm TSMC silicon but now… maybe the jump is quite substantial with the M5.

Let’s unpack what we’ve seen with the A19:

- GPU up to 40% more powerful per core
- GPU with better AI capabilities
- CPU with bigger cache memory
- Enhanced security at the silicon level

I think all of this can be potentially used to build the M5 which, despite being still manufactured with a 3nm process, offers a series of enhancements at the architectural level.

I missed John Srouji yesterday at the A19 presentation…
 
The precision heat pipe allows 40% more heat to be extracted from an SoC - bodes well for Apple’s future large SoCs being both quiet (low air speed) and powerful (high wattage Ultra with less bottlenecking at higher >5? Ghz)
 
A large flat vapour chamber over the SoC just means a large mm^2 area of rapid heat dissipation.

M5 : If Apple couple this with box tube with fins - they can create a laminated airflow in a pyrolytic (graphite) structure : Very quiet for a modest air flow.

Speculation : If Apple’s can quietly extract 700W of power from an M5 Mac Studio / or 250W Mac Mini enclosure - it would give Apple many years of ‘M’ iterations with speed bumps. If Apple release a 500+ Watt M5 Mac Studio Ultra - would there be much of a market for a full sized Mac Pro?
 
Last edited:
A large flat vapour chamber over the SoC just mean a large mm^2 area of rapid heat dissipation.

M5 : If Apple couple this with box tube with fins - they can create a laminated airflow in a pyrolytic (graphite) structure : Very quiet for a modest air flow.

Speculation : If Apple’s can quietly extract 700W of power from an M5 Mac Studio / or 250W Mac Mini enclosure - it would give Apple many years of ‘M’ iterations with speed bumps. If Apple release a 500+ Watt M5 Mac Studio Ultra - would be much of a market for a full sized Mac Pro?
Even a year ago I'd have said no, but if they aren't ever doing an "Extreme" then that does probably make sense.

I think the Extreme is coming eventually, at least I hope so, but the Studio being over engineered is nice and bodes well for the future.

The use case for Graphics cards is dead since they aren't compatible, and even Pro Audio is moving away form PCIe cards now, most new audio devices are external thunderbolt even from Avid and with thunderbolt Optical cables you can have high-speed runs across the studio to where your audio racks live, far away from any fan noise etc.

Adding storage seems like the only real use case left, which isn't nothing, but probably isn't worth a $2,000+ upcharge since you can get an entire external SSD array for that much money.

If 3D XPoint didn't get discontinued I'd make the case that there could be some really cool things done with that and ML/AI tasks since you have basically unlimited writes and doing that very fast could open up some opportunities, but alas.
 
For multi-core, Apple A19 Pro is as fast as M2, if not faster.

Geekbench 6.5: 4019 / 11054

15% faster vs 16 pro on ST (3500 vs 4000) on 6% higher freq (4.0 vs 4.26). 25% improvement on MT.

Nicely done.

Lets say M5 uses same freq boost and similar arch improvement, and keeps core count, could mean 4300 ST and 18700 MT for the 10 core.
 
Last edited:
  • Like
Reactions: Antony Newman
The M5 looks better and better… however, the M6 is supposed to come with a much higher performance jump while maybe reducing heat/power consumption, because of the jump to 2nm and GAA transistors…

I really don’t know if it’s worth the wait. If the next generation of M5 macs comes with a base storage of 512GB or a minimum of 24GB of RAM, without increasing prices, I will definitely pick one.

And even if that doesn’t happen, I think it will be a good time to upgrade because with 2nm prices are probably going to get up, the first batches of a new manufacturing process usually don’t have as good yields, there may be more “binning”… and of course, the eventual invasion of Taiwan is always a factor to count on.

So yeah, I think I’ve waited enough since the M2 times that I planned to upgrade my 2014 Mac mini. I was tempted by the M2 but I wanted to wait for the 3nm. There wasn’t an M3 Mac mini, I tried the M4 Mac mini and it was already quite awesome but I wanted a bit more… and with a mature enough N3P (or maybe apple is using N3X) I think the M5 will be the perfect time to upgrade.
 
  • Haha
Reactions: Basic75
The M5 looks better and better… however, the M6 is supposed to come with a much higher performance jump while maybe reducing heat/power consumption, because of the jump to 2nm and GAA transistors…

I really don’t know if it’s worth the wait. If the next generation of M5 macs comes with a base storage of 512GB or a minimum of 24GB of RAM, without increasing prices, I will definitely pick one.

And even if that doesn’t happen, I think it will be a good time to upgrade because with 2nm prices are probably going to get up, the first batches of a new manufacturing process usually don’t have as good yields, there may be more “binning”… and of course, the eventual invasion of Taiwan is always a factor to count on.

So yeah, I think I’ve waited enough since the M2 times that I planned to upgrade my 2014 Mac mini. I was tempted by the M2 but I wanted to wait for the 3nm. There wasn’t an M3 Mac mini, I tried the M4 Mac mini and it was already quite awesome but I wanted a bit more… and with a mature enough N3P (or maybe apple is using N3X) I think the M5 will be the perfect time to upgrade.
I own a M4 Air and even if I would be tempted to upgrade just because the performance will be mindblowing, there is no real benefit for me. My production loads are often under one minute with the M4. Every once in a while 2-3 minutes. Cutting that to sub 2 min all the time… is not that much of a difference.

But supplementing the Air with a future mini M6 for redundancy with 2 great devices.. that could happen.

Edit. No news but the value of the Pro and Max for personal use is really not great, cpu-wise. The Max version is often matched by the base in two years. M3 Max binned scores 19000 on MT. If my little number play in the post before turns out true the base M5 will have that in reach. Like how the M4 base was able to reach M2 Max and M3 Pro. Apple engineers are on fire with their low-energy (and the power cores too) evolution.
 
Last edited:
I own a M4 Air and even if I would be tempted to upgrade just because the performance will be mindblowing, there is no real benefit for me. My production loads are often under one minute with the M4. Every once in a while 2-3 minutes. Cutting that to sub 2 min all the time… is not that much of a difference.

But supplementing the Air with a future mini M6 for redundancy with 2 great devices.. that could happen.

Edit. No news but the value of the Pro and Max for personal use is really not great, cpu-wise. The Max version is often matched by the base in two years. M3 Max binned scores 19000 on MT. If my little number play in the post before turns out true the base M5 will have that in reach. Like how the M4 base was able to reach M2 Max and M3 Pro. Apple engineers are on fire with their low-energy (and the power cores too) evolution.
But don't forget about GPU difference and extra encoders / decoders for video editing. My M1 Max probably is still not matched by base M4 for those workloads.
 
But don't forget about GPU difference and extra encoders / decoders for video editing. My M1 Max probably is still not matched by base M4 for those workloads.
Yeah, also there’s the form factor. A Mac Studio will always perform better, especially for sustained workloads, due to the much better dissipation.
 
For multi-core, Apple A19 Pro is as fast as M2, if not faster.

Geekbench 6.5: 4019 / 11054

I’m eager to see the first A19 and A19 Pro silicon analysis…

I wonder if the Neural Engine has been improved with a new architecture or packing more transistors, or they left the NE as it was because they prefer to focus on the GPU for the new AI features.

I would also like to know if regular A19 GPU cores have those ML accelerators as the A19 Pro, but I don’t think so, maybe they are binned GPU cores.

Then, the crucial aspect is if the M5 will inherit those AI/ML accelerators for the GPU (I hope it will), the increased cache memory, and everything that makes the A19 Pro so awesome.

And… maybe we’re talking about a two generation leap in the manufacturing process, if Apple decided to use N3X process node.
 
I’m eager to see the first A19 and A19 Pro silicon analysis…

I wonder if the Neural Engine has been improved with a new architecture or packing more transistors, or they left the NE as it was because they prefer to focus on the GPU for the new AI features.

I would also like to know if regular A19 GPU cores have those ML accelerators as the A19 Pro, but I don’t think so, maybe they are binned GPU cores.

Then, the crucial aspect is if the M5 will inherit those AI/ML accelerators for the GPU (I hope it will), the increased cache memory, and everything that makes the A19 Pro so awesome.

And… maybe we’re talking about a two generation leap in the manufacturing process, if Apple decided to use N3X process node.
I don’t understand. You aren’t suggesting A19 is a binned A19 Pro are you?

Anyhow, as there is a binned A19 Pro chip in the iPhone Air which like the A19 non-Pro has 5 GPU cores instead of the 6 in the regular A19 Pro, it will prove interesting to compare the iPhone 17 to the iPhone Air.
 
I don’t understand. You aren’t suggesting A19 is a binned A19 Pro are you?

Anyhow, there is a binned A19 Pro chip in the iPhone Air which like the A19 non-Pro has 5 GPU cores instead of the 6 in the regular A19 Pro, so it will prove interesting to compare the iPhone 17 to the iPhone Air.
Nah, forget about that part. I was completely wrong, I just read that the regular A19 also has machine learning accelerators.
 
And… maybe we’re talking about a two generation leap in the manufacturing process, if Apple decided to use N3X process node.
There is no chance N3X, the second generation of the first-generation N4X (blog), will be used in a macOS device. These are not part of the regular 5nm N5-N5P-N4-N4P and 3nm N3-N3E-N3P optimization sequences. They are a different branch of the family: high-power for maximum clock frequencies.
 
  • Like
Reactions: Populus and EugW
There is no chance N3X, the second generation of the first-generation N4X (blog), will be used in a macOS device. These are not part of the regular 5nm N5-N5P-N4-N4P and 3nm N3-N3E-N3P optimization sequences. They are a different branch of the family: high-power for maximum clock frequencies.
Thank you for the clarification, I didn’t know they were different branches, and at 100W of TDP, there’s indeed no chance of N3X being used for the M5.
 
Sorry if this was already shared, but the Geekbench AI GPU test results are interesting

17 Pro (Max?) iPhone Air vs 16 Pro Max

Screenshot 2025-09-13 at 18.33.48.png


Huge gains across most sub tests. Can't wait to see this new architecture scaled up in M5!

Edit: corrected the post, this is iPhone Air vs. 16 Pro Max.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.