M3 Pro vs M2 Pro MacBook Pro 14” base models benchmarked

[Dav1d_B]

Just picked up my new 14” MacBook Pro M3 and ran some benchmarks comparing it to my 14” MacBook Pro M2 (both base models, specs below).

14” MacBook Pro with Apple M3 Pro
18GB Unified Memory
11-Core CPU, 14-Core GPU
512 GB SSD Storage

VS

14” MacBook Pro with Apple M2 Pro
16GB Unified Memory
10-Core CPU, 16-Core GPU
512 GB SSD Storage

Geekbench 6:
M3 Pro CPU Test: 14239 (Multi-Core), 3127 (Single-Core)
M2 Pro CPU Test: 12421 (Multi-Core), 2699 (Single-Core)

M3 Pro GPU Test (Metal): 68950
M2 Pro GPU Test (Metal): 74910

M3 Pro GPU Test (Open CL): 43023
M2 Pro GPU Test (OpenCL): 45804


Cinebench 2024:

M3 Pro CPU Test: 851 (Multi-Core), 138 (Single-Core)
M2 Pro CPU Test: 801 (Multi-Core), 121 (Single-Core)

M3 Pro GPU Test: 5528
M2 Pro GPU Test: 2876

GPU tests were a bit surprising (at lest Geekbench 6): M2 Pro was actually beating the M3 Pro in the Geekbench 6 GPU tests (ran it twice, no material change), while the M3 Pro was getting almost 2x the points over the M2 Pro in the Cinebench 2024 test.

PS: Color and charging cable in space black really are awesome!
PPS: I am new here, so hopefully this is useful and not a double post.

 

[mavere]

Any storage speed tests you can run? Maybe with BlackMagic app or something.

 

[Dav1d_B]

Good idea - here you go: M3 Pro reading speeds considerably (!) faster across the board.

 

[ovbacon]

My regular M2 is faster than the M2 pro in the above test.... that does not seem to make any sense.

 

[a.r40]

this test is against the binned 11 core variant of the m3 pro

 

[CWallace]

GPU tests were a bit surprising (at lest Geekbench 6): M2 Pro was actually beating the M3 Pro in the Geekbench 6 GPU tests (ran it twice, no material change), while the M3 Pro was getting almost 2x the points over the M2 Pro in the Cinebench 2024 test.

I have read that Geekbench is not reflecting the ray tracing and memory allocation benefits of the M3 GPU so in real-world application use, the M3 GPU can be considerably faster than the M2 GPU.

 

[Mcdevidr]

These are both the entry bin models. Thank you for comparing. Seems to be an ok upgrade there. I’m really looking forward to trying out that single core later in my game. Anything to make the ticks faster. This may be faster than the i9 laptop I had and infinitely quieter to boot in that regard.

 

[jonnyr78]

Posted my 16" M3 Pro scores here: https://forums.macrumors.com/threads/my-macbook-pro-16-m3-pro-benchmark-scores.2410184/ not sure I put that thread in the right place! :/

 

[FrozenDarkness]

interesting. i'd trust cinebench just a tad bit more. but looks liek a solid improvement

 

[OrenLindsey]

Good idea - here you go: M3 Pro reading speeds considerably (!) faster across the board.

Apple probably switched back from the single NAND chip that people complained about last year to the double chip they were using before that got higher speeds.

 

[Analog Kid]

Geekbench 6:
M3 Pro GPU Test (Metal): 68950
M2 Pro GPU Test (Metal): 74910

M3 Pro GPU Test (Open CL): 43023
M2 Pro GPU Test (OpenCL): 45804


Cinebench 2024:
M3 Pro GPU Test: 5528
M2 Pro GPU Test: 2876

Great results for anyone who loves endless repetitive debates in the forums.

1. M3 GPU sucks, look at the Geekbench results.
2. Geekbench isn't well optimized, look at how much better it does in Cinebench
3. The M3 CPU in Cinebench is only 6% faster
4. Cinebench isn't well optimized, look at how much better it does in Geekbench
5. Goto: 1

 

[TechnoMonk]

Great results for anyone who loves endless repetitive debates in the forums.

1. M3 GPU sucks, look at the Geekbench results.
2. Geekbench isn't well optimized, look at how much better it does in Cinebench
3. The M3 CPU in Cinebench is only 6% faster
4. Cinebench isn't well optimized, look at how much better it does in Geekbench
5. Goto: 1

That’s what you get when you have an MBA running the company, not an engineer.
/s…

 

[haysoos123]

M3 Pro GPU Test (Metal): 68950
M2 Pro GPU Test (Metal): 74910

M3 Pro GPU Test (Open CL): 43023
M2 Pro GPU Test (OpenCL): 45804

vs.

M3 Pro GPU Test: 5528
M2 Pro GPU Test: 2876


There's something seriously wrong with one of these GPU benches when one has a 2x performance difference in favor the M3, and the other shows the M3 actually behind.

 

[diamond.g]

M3 Pro GPU Test (Metal): 68950
M2 Pro GPU Test (Metal): 74910

M3 Pro GPU Test (Open CL): 43023
M2 Pro GPU Test (OpenCL): 45804

vs.

M3 Pro GPU Test: 5528
M2 Pro GPU Test: 2876


There's something seriously wrong with one of these GPU benches when one has a 2x performance difference in favor the M3, and the other shows the M3 actually behind.

Nah the results are probably fine, GB is testing GPGPU compute while CB is testing rendering. For GPGPU compute the number of ALU's how fast they run matters a lot more than mesh shaders or ray tracing.

 

[teknikal90]

i'd be interested to see what the regular M3 does, as that will be closest in price (once you add on the 16gb option) to a refurb base model m2 pro

much longer battery life + slightly brighter screen vs. potentially tangibly faster CPU/GPU, more ports & two fans.

 

[Faize]

Good idea - here you go: M3 Pro reading speeds considerably (!) faster across the board.

Definitely better than last gen which was bad even by the standards of PCIe 3.0 SSDs, but the numbers still leave much to be desired - those are SN770 speeds, and the SN770 is a $50 budget/midrange SSD...

 

[darkmatter343]

Good idea - here you go: M3 Pro reading speeds considerably (!) faster across the board.

Was Apple using a single Nand chip for the 512GB version of the base M2 MBP? Seeing these SSD speed results would lead me to believe they were? and now we have 2x256GB but I haven’t seen any pictures of the guts so we’ll have to wait and see. That’s a pretty significant speed bump so I find it unlikely Apple just used a higher end single 512GB NAND... unless of course I missed the memo where they increased the PCI-e speed?

 

[Kronsteen]

Great results for anyone who loves endless repetitive debates in the forums.

1. M3 GPU sucks, look at the Geekbench results.
2. Geekbench isn't well optimized, look at how much better it does in Cinebench
3. The M3 CPU in Cinebench is only 6% faster
4. Cinebench isn't well optimized, look at how much better it does in Geekbench
5. Goto: 1

A goto statement. How fantastic! None of this new-fangled structured programming nonsense.

(A comefrom statement at line 0 might also have worked.)

 

[aeronatis]

There's something seriously wrong with one of these GPU benches when one has a 2x performance difference in favor the M3, and the other shows the M3 actually behind.

Geekbench GPU score is slightly lower due to M3 Pro having fewer cores (not sure if lower memory bandwidth affects is as well).

Cinebench GPU test, on the other hand, performs render test using Redshift which probably utilizes hardware ray tracing, which lines-up well with Apple's x2 claims for rendering ray traced scenes.

 

[macduke]

PPS: I am new here, so hopefully this is useful and not a double post.

Welcome! I am new here too, or at least that's what I like to tell myself to cope with aging.

This sort of CPU improvement reminds me of the Intel days. Seems pretty clear Apple is trying to further differentiate the Max from the Pro, in order to drive more upsells to the Max, and improve their margins. I mean, historically I have considered myself to be somewhat of a power user, and yet I would've been completely fine with the M3 Pro if it wasn't limited in other ways, such as RAM.

This is a form of inflation, as you get less than you used to, Apple can claim they aren't raising prices (they did something similar by basically keeping the price of the iPhone 15 Pro Max the same, but you couldn't get it in the cheaper 128GB size) and they even had the audacity to say they were dropping the price down to $1699, when really this is the replacement for the old basic M2 13" MacBook Pro that cost several hundred less. But you do get a bigger, nicer display and I think better camera/speakers. But it's not a price drop, lol.

Great results for anyone who loves endless repetitive debates in the forums.

1. M3 GPU sucks, look at the Geekbench results.
2. Geekbench isn't well optimized, look at how much better it does in Cinebench
3. The M3 CPU in Cinebench is only 6% faster
4. Cinebench isn't well optimized, look at how much better it does in Geekbench
5. Goto: 1

This is the way...

There's something seriously wrong with one of these GPU benches when one has a 2x performance difference in favor the M3, and the other shows the M3 actually behind.

They are new chips, hopefully GeekBench can be recalibrated to account for this, but bigger changes like this are expected to bring unexpected results since this is the first Apple GPU with raytracing. In the real world it's probably better, but it seems clear that the M3 Pro is not where it's at unless you're already coming from an older Intel system. I'm coming from Intel and my M3 Max is going to blow my face off, can't wait.

 

[ggCloud]

M3 Pro GPU Test (Metal): 68950
M2 Pro GPU Test (Metal): 74910

M3 Pro GPU Test (Open CL): 43023
M2 Pro GPU Test (OpenCL): 45804

vs.

M3 Pro GPU Test: 5528
M2 Pro GPU Test: 2876


There's something seriously wrong with one of these GPU benches when one has a 2x performance difference in favor the M3, and the other shows the M3 actually behind.

Mac15,7 - Geekbench

Benchmark results for a Mac15,7 with an Apple M3 Pro processor.

browser.geekbench.com

Geekbench Search - Geekbench

This shows that M3 Pro GB6 Metal is better than M2 Pro. I believe the OP test was throttled because of background tasks maybe.

 

[galad]

M3 Pro GPU Test: 5528
M2 Pro GPU Test: 2876


There's something seriously wrong with one of these GPU benches when one has a 2x performance difference in favor the M3, and the other shows the M3 actually behind.

Cinebench is using hardware raytracing, so it will get a 2x speed up compared to the M2 that had no hw raytracing.

 

[rwhg]

Thank you for taking the time to do this. It’s a comparison I think a lot of people will appreciate seeing.

 

[rwhg]

Was Apple using a single Nand chip for the 512GB version of the base M2 MBP? Seeing these SSD speed results would lead me to believe they were? and now we have 2x256GB but I haven’t seen any pictures of the guts so we’ll have to wait and see. That’s a pretty significant speed bump so I find it unlikely Apple just used a higher end single 512GB NAND... unless of course I missed the memo where they increased the PCI-e speed?

They were using a single NAND chip for the M2 Pro MB Pro, much to everyone’s chagrin. Glad to see that seems to be corrected for the M3 Pro MB Pro.

 

[Populus]

Just picked up my new 14” MacBook Pro M3 and ran some benchmarks comparing it to my 14” MacBook Pro M2 (both base models, specs below).

14” MacBook Pro with Apple M3 Pro
18GB Unified Memory
11-Core CPU, 14-Core GPU
512 GB SSD Storage

VS

14” MacBook Pro with Apple M2 Pro
16GB Unified Memory
10-Core CPU, 16-Core GPU
512 GB SSD Storage

Geekbench 6:
M3 Pro CPU Test: 14239 (Multi-Core), 3127 (Single-Core)
M2 Pro CPU Test: 12421 (Multi-Core), 2699 (Single-Core)

M3 Pro GPU Test (Metal): 68950
M2 Pro GPU Test (Metal): 74910

M3 Pro GPU Test (Open CL): 43023
M2 Pro GPU Test (OpenCL): 45804


Cinebench 2024:

M3 Pro CPU Test: 851 (Multi-Core), 138 (Single-Core)
M2 Pro CPU Test: 801 (Multi-Core), 121 (Single-Core)

M3 Pro GPU Test: 5528
M2 Pro GPU Test: 2876

GPU tests were a bit surprising (at lest Geekbench 6): M2 Pro was actually beating the M3 Pro in the Geekbench 6 GPU tests (ran it twice, no material change), while the M3 Pro was getting almost 2x the points over the M2 Pro in the Cinebench 2024 test.

PS: Color and charging cable in space black really are awesome!
PPS: I am new here, so hopefully this is useful and not a double post.

Many thanks! This was, indeed, a very useful comparison, between both base configurations of the M2 Pro and M3 Pro.

By the way, maybe it’s just the way the picture was taken, but it seems like the M2 Pro screen (left) is actually whiter and more uniform than the new M3 Pro screen (right) which seems to have a yellowish area towards the left edge and a reddish/pinkish area towards the right edge. Is that true in real life?