My MacBook Pro 16 M3 Pro Benchmark scores

[OrenLindsey]

In the M2 generation, Apple used a single NAND SSD chip for the base model SSDs. That's why speeds were slower. Now they've switched back to using two, so speeds are the same as the higher SSDs. Pretty good.
I do wonder though, why did they switch in the first place?

 

[GuruZac]

In the M2 generation, Apple used a single NAND SSD chip for the base model SSDs. That's why speeds were slower. Now they've switched back to using two, so speeds are the same as the higher SSDs. Pretty good.
I do wonder though, why did they switch in the first place?

My guess is supplier constraint? I can’t think of anything else other than cheaper for Apple to use single slower NAND.

 

[FrozenDarkness]

My guess is supplier constraint? I can’t think of anything else other than cheaper for Apple to use single slower NAND.

i think it's price. i don't think apple ships enough computers for there to be a constraint on NAND cells.

2 256gb chips is more expensive than 1 512gb in component and install costs probably.

 

[garethjs]

512gb SSD, so not bad IMHO, clearly better than the M2 Pro 512gb SSD's.

Same specs as yours. What r u getting with BM disk test?

It’s 4300-4500 write and 5300-5500 read with mine

 

[ori69]

People have been wondering about M3 Pro benchmarks, so here are mine, the 16" 36gb ram with the 12/18 M3 Pro.

By the way, the black is fantastic and really resists fingerprints


View attachment 2308827View attachment 2308828View attachment 2308831View attachment 2308830View attachment 2308829View attachment 2308832View attachment 2308833

Today I tested an MBP 16 M3 Pro in Logic. The tests came out very poorly.

 

[ggCloud]

Today I tested an MBP 16 M3 Pro in Logic. The tests came out very poorly.

View attachment 2309027

View attachment 2309029

How is the M1 slower than M2?

Very different results in Logic compared to other real world tests

 

[Tyler O'Bannon]

Also, FWIW, benchmarks were done connected to power (but were within a margin of error on battery) and the fans never came on, only slightly warm to touch.

This is as impressive as the benchmarks.

 

[Shirasaki]

Blackmagic scores (for the 512gb SSD, this is using the 1gb test)


View attachment 2308862

This is sequential speed right? The benchmarks of QD1 4K is far more important for most people than pure sequential.
Thankfully you also attached that and to me it’s not super exiting compared to other off the market SSD with regards to 4K random access.

 

[AgentMcGeek]

So your base M3 Pro GPU beats the M1 Max? That’s impressive!

 

[Unregistered 4U]

My guess is supplier constraint? I can’t think of anything else other than cheaper for Apple to use single slower NAND.

Oh, I think you’re right. Because it would require two of the half size chips, I think they couldn’t get a sufficient quantity of those at the time (of course, probably read here, so maybe just a rumor

 

[playtech1]

View attachment 2309029

Interesting results - and a bit disappointing for M3 Pro if it's a downgrade from M2 Pro in Logic. I see the Mini M2 was also materially worse than the M1 though - which seems odd given the similarity between M1 and M2. Are you sure these results stack up?

 

[Mac mini power user]

Today I tested an MBP 16 M3 Pro in Logic. The tests came out very poorly.

View attachment 2309027

View attachment 2309029

It doesn't make sense for the M2 to be 25 percent worse than the M1. Is there something off with this benchmark software? Elsewise Logic needs an optimisation update.

 

[xraydoc]

It doesn't make sense for the M2 to be 25 percent worse than the M1. Is there something off with this benchmark software? Elsewise Logic needs an optimisation update.

Or it's very SSD-dependent and the M2 they're using has a slower SSD config than the M1.

 

[avkills]

Logic likes a lot of RAM from what I have seen from other posts about it.

Although the metal score makes me think that the M3 Max just might be neck and neck with a W6800X which is very impressive for a laptop GPU; and the power envelope it is using.

 

[3SQ Machine]

This is a great thread and sorry I found it so late. Logic only uses performance cores and M3 pro has less of them.

In my own personal use case, my M3 Pro 14" 18gb ram (12 core) had less CPU usage for my same projects than my M1 Pro 16" 16gb ram. However, my projects tend to spike on a single core. I'm guessing the much higher single core performance of the M3 pro makes a big difference for my use case. Nevertheless, I think there could be many other projects that need more muti-threaded performance across the performance cores that would find this a problem on the M3 Pro.