Mac Mini M2

[SpyderTracks]

So Apple have just announce the Mac Mini running on the M2 chip.

Believe it or not, this thing outpowers the Mac Pro which is still running on Intel and can run you over £50,000 for a fully configured system.

This is definitely what I've been waiting for, it's insanely overpowered for my needs but should last me happily for 10 years or so running Logic ProX

It's actually $100 cheaper for the base model as well, which quite frankly blows me away given inflation, runs at $599 for the base model M2 standard.

For the M2 Pro which includes a major step up in the graphics department (which the Mini has never had before the Apple Silicon models) will run you at $1299

 

[TonyCarter]

Be careful on the SSD configs though, as the smaller SSD configs have a single stick vs the dual-stick in the older M1 machines...and are showing les than 50% of the performance. For example my Studio Max can deliver 7000MBps read and 5000MBps write, whereas the 256/512GB configs in the M2 are about half of that (info from a small sample admittedly, but all using the same tools Crystal/Amporphous Diskmark and
BlackMagic Speedtest).

Not a big problem for someone who's just using it as a home machine, but where the SSD is used for 'virtual RAM', then it may show some slowdowns.

MacStudioMax M1 512GB...

MacMini M1 256GB

MacMini M2 256GB

MacBookAir M2 1TB

 

[SpyderTracks]

Be careful on the SSD configs though, as the smaller SSD configs have a single stick vs the dual-stick in the older M1 machines...and are showing les than 50% of the performance. For example my Studio Max can deliver 7000MBps read and 5000MBps write, whereas the 256/512GB configs in the M2 are about half of that (info from a small sample admittedly, but all using the same tools Crystal/Amporphous Diskmark and
BlackMagic Speedtest).

Not a big problem for someone who's just using it as a home machine, but where the SSD is used for 'virtual RAM', then it may show some slowdowns.

MacStudioMax M1 512GB...
View attachment 35875

MacMini M1 256GB
View attachment 35876

MacMini M2 256GB
View attachment 35877

MacBookAir M2 1TB
View attachment 35878

Understood. I won't be buying for a little while, as with any new design (I believe the logic board is entirely redesigned for this mac mini), there will be hiccups on the first iteration.

Hopefully it's a software issue rather than hardware, it seems a really major area to skimp out on when the other components will undoubtedly be able to saturate that bandwidth.

I'm hopeful it's a hardware / first generation issue.

I also wouldn't count the mac book air in test comparisons as it has massively tied down PCIe lanes aside from other limitations due to the thin design.

 

[TonyCarter]

Understood. I won't be buying for a little while, as with any new design (I believe the logic board is entirely redesigned for this mac mini), there will be hiccups on the first iteration.

Hopefully it's a software issue rather than hardware, it seems a really major area to skimp out on when the other components will undoubtedly be able to saturate that bandwidth.

I'm hopeful it's a hardware / first generation issue.

I also wouldn't count the mac book air in test comparisons as it has massively tied down PCIe lanes aside from other limitations due to the thin design.

From the teardowns it seems it could simply be the single vs dual SSD stick...and in the dual-stick config Apple may simply be using them in parallel...so, for example 1 SSD = 8 DRAM chips, or 2 SSDs = 16 DRAM chips, so twice as many channels?

...and as these are proprietary sticks (i.e. similar to 2280 NVMe) with the controller being part of the Apple Silicon board, and using proprietary firmware, then there's no simple way to interrogate them in another system.

 

[SpyderTracks]

From the teardowns it seems it could simply be the single vs dual SSD stick...and in the dual-stick config Apple may simply be using them in parallel...so, for example 1 SSD = 8 DRAM chips, or 2 SSDs = 16 DRAM chips, so twice as many channels?

...and as these are proprietary sticks (i.e. similar to 2280 NVMe) with the controller being part of the Apple Silicon board, and using proprietary firmware, then there's no simple way to interrogate them in another system.

The testing methodology above isn't very sound, you've got different formats of devices and different program versions.

Really needs to be more standardised.

Plus the program they're using hasn't been updated for M2 chips as yet, so there's possibly something that needs updating at a software level.

Until those things are addressed it's not a fair comparison.

 

[TonyCarter]

These synthetic benchmarks never give the complete picture, and performance for a 10GB single file will be completely different than a 10GB folder of 10,000 smaller files.

Luckily, most of my stuff is bundled into 1-5GB folders containing a max of 100 smaller files (collected fonts, vector artwork and images) along with the 2 or 3 large working files.