M5 Pro and Max unveiled

[Jimmyjames]

Let me be clear: it's not your lack of mathematical knowledge, quite the opposite, what you want is the correct way to do things - rather it's your mathematical knowledge confusing you! They're fundamentally doing it wrong. I understand why, or at least I have a reasonable hypothesis for why, they are doing it wrong, but they are doing it backwards because they assume most people won't understand why the better number is negative. They should either report that -17% faster/20% slower or +20% more performant/-17% less performant but instead they report +17 faster/ -20 slower.

Perhaps mathematical knowledge is the wrong way to put it. Although I have much to learn I’m sure. A lack of understanding or comprehension I suppose. It’s been a tough year.

 

[dada_dave]

We could take the top score. Still a nice uplift.
View attachment 38350

Do you still have your python script to read the opendata JSON file? If you were of a mind to do it (no need or pressure from my end), you could probably get the exact number yourself. Again, why opendata doesn't do this right for CPUs, especially when they do it right for GPUs, I do not understand.

Perhaps mathematical knowledge is the wrong way to put it. Although I have much to learn I’m sure. A lack of understanding or comprehension I suppose. It’s been a tough year.

I more than understand.

 

[Jimmyjames]

Do you still have your python script to read the opendata JSON file? If you were of a mind to do it (no need or pressure from my end), you could probably get the right number yourself.

I’ll see if I can find it.

I more than understand.

Thanks.

 

[dada_dave]

Thanks.

I just realized I screwed up my description and used faster/slower when I should've written more time/less time! Edited the post.

They should either report that the M4 completed it in -17% less time than the M5 and the M5 took +20% longer than the M4 or the M4 is +20% more performant(faster) and the M5 is -17% less performant(slower), but instead they report +17% for the M4 with the M5 as the baseline and -20 for the M5 with the M4 as the baseline.

 

[leman]

Blender 5.0

M5 Max GPU being within the striking distance of 5070 Ti is quite incredible.

 

[dada_dave]

M5 Max GPU being within the striking distance of 5070 Ti is quite incredible.

Blender 4.5 has the binned M5 Max GPU (I wonder which reviewer got that or have devices started to reach customers?). Also very impressive, faster than full M4 Max.

Blender - Open Data

Blender Open Data is a platform to collect, display and query the results of hardware and software performance tests - provided by the public.

opendata.blender.org

 

[RockRock8]

Blender 4.5 has the binned M5 Max GPU (I wonder which reviewer got that or have devices started to reach customers?). Also very impressive, faster than full M4 Max.

Blender - Open Data

Blender Open Data is a platform to collect, display and query the results of hardware and software performance tests - provided by the public.

opendata.blender.org

I don't know the difference between 4.5 and 5 Blender, but it's 95% of the performance of the 80 core M3, and that has ray tracing too.

 

[Jimmyjames]

Holy **** the binned M5 Pro.

 

[Jimmyjames]

So the M5 Max is the second best Blender laptop around. If it had faster memory I think it could be the first, or at least close.

 

[Jimmyjames]

Do you still have your python script to read the opendata JSON file? If you were of a mind to do it (no need or pressure from my end), you could probably get the exact number yourself. Again, why opendata doesn't do this right for CPUs, especially when they do it right for GPUs, I do not understand.

Turns out if you group by “none”, you can see individual scores. For the total score at least. Sub-scores are still hidden in the json.

 

[exoticspice1]

got my m5 pro 16”

The new M5 Pro struggles in Cinebench 2024 due to core scheduling issues.

Power is 45watts in Cinebench 2024 because of this. Clocks are low.

 

[Jimmyjames]

got my m5 pro 16”
View attachment 38355

The new M5 Pro struggles in Cinebench 2024 due to core scheduling issues.
View attachment 38356
View attachment 38357

Power is 45watts in Cinebench 2024 because of this. Clocks are low.

Enjoy it!

 

[exoticspice1]

In Cinebench 2026, the M5 Pro doesn’t have any issue and is able to maintain it clocks and use up to 88 watts in the cpu test

The Cinebench 2026 GPU test uses 64 watts.

 

[dada_dave]

got my m5 pro 16”
View attachment 38355

The new M5 Pro struggles in Cinebench 2024 due to core scheduling issues.
View attachment 38356
View attachment 38357

Power is 45watts in Cinebench 2024 because of this. Clocks are low.

Amazing! Thanks so much for sharing. Do you mind if I put your data in my charts? Your results confirm both NotebookCheck's score is right and that NotebookCheck's power measurement has to be wrong.

Not to impose too much on you, what happens if you turn "high power" mode on?

I do wonder if the M5 Pro CPU is down clocked or doesn't have the same "TDP" as the Max "CPU"? And that's how they are binning it instead of turning off cores?

View attachment 38359
View attachment 38360View attachment 38358

In Cinebench 2026, the M5 Pro doesn’t have any issue and is able to maintain it clocks and use up to 88 watts in the cpu testView attachment 38361

The Cinebench 2026 GPU test uses 64 watts.

Hmmmm ... maybe not if it can hit 88W. Still I think the M5 Max can hit 9500 in CB R26, yours hit 9100, yes?

 

[exoticspice1]

Do you mind if I put your data in my charts?

I don’t mind. Go ahead.

Not to impose too much on you, what happens if you turn "high power" mode on?

I tested in high power mode

 

[exoticspice1]

maybe not if it can hit 88W. Still I think the M5 Max can hit 9500 in CB R26, yours hit 9100, yes?

Yep hit around 9000 on the first run. But I wasn’t able to do it again. These new cpu cores are real finicky.

I haven’t ever seen this sort of score variation before.

 

[dada_dave]

I don’t mind. Go ahead.



I tested in high power mode

Thanks! You know given the massive variation between your result and NBC's, I might just leave the M5 Pro off the chart for now. (though I do think their 100W average can't be right)

Edit: Wait what was your final CBR24 score? I thought it was 2059 but looking again I see that was earlier in the run and a later screenshot shows ~1800.

Yep hit around 9000 on the first run. But I wasn’t able to do it again. These new cpu cores are real finicky.

I haven’t ever seen this sort of score variation before.

Sounds like Apple has some work to do - I assume indexing was finished and all that? Just trying to think of other things that could cause it to be more variable.

 

[RockRock8]

I don’t mind. Go ahead.



I tested in high power mode

Can you test the same test using automatic power mode, charged 100% (like you did before)?

 

[RockRock8]

Thanks! You know given the massive variation between your result and NBC's, I might just leave the M5 Pro off the chart for now. (though I do think their 100W average can't be right)

Edit: Wait what was your final CBR24 score? I thought it was 2059 but looking again I see that was earlier in the run and a later screenshot shows ~1800.



Sounds like Apple has some work to do - I assume indexing was finished and all that? Just trying to think of other things that could cause it to be more variable.

Indexing can take up to a week. Also are you adding high power mode to this chart? Because that dramatically increases power consumption usually, and usually it doesn't actually add performance but rather sustains it.

 

[RockRock8]

Yep hit around 9000 on the first run. But I wasn’t able to do it again. These new cpu cores are real finicky.

I haven’t ever seen this sort of score variation before.

It's indexing most likely. There are fluctuations in early GB tests.