M3 Max Performance disappointing?

I have upgraded from a MacBook M1 (not Pro) to a MacBook M3 Max.

The M1 has 8 cores, 16 GB memory.
The M3 Max has 14 cores and 30 GPU cores, 36 GB memory

I am using Davince Resolve 18.6 under MacOS Sonoma.

I am kind of disappointed of the Performance of the M3 Max.

I used a simple mov-File, length 1min, 30fps, 4k. It is me speaking in front of a backgroud. I am using the Magic Mask to mask myself.

On my MacBook M1 it takes 1:55 min (at around 7fps) no fan-noise

On my MacBook M3 Max in Low Power Mode it takes 1.34 min (at around 9 fps), fan jumps in after the process

On my MacBook M3 Max in High Power Mode it takes 1.10 min (at around 12 fps), fan jumps starts in the mid of the process.

Maybe it is a difficult task to process but I kind of hoped that the M3 Max was significantly faster (more Cores, more GPUs, double RAM). Did I make a mistake?

Hi Breeze:

I have tried Magic Mask on my M1Max 10 core CPU 32 Core GPU. Performance was to slow for me so I infrequently use. My impression is that Magic Mask is for computers that have far more processing power that is available in any laptop.

I was considering a M3Max and wondered about Magic Mask performance. In your comment are you using 14" or 16" M3Max MBP?

Don

I have the 14" version. What I didn't expect is that the fan are switching on quite fast. I've never heard the fans on my M1 notebook. And yes, it's a 40% increase in performance, but I expected way more from a notebook with twice as much RAM and GPU cores.

Hi VB:

Watch this video comparing the 14 to the 16" M3MBP with respect to throttling and fan noise.

Don

I have the 14" version. What I didn't expect is that the fan are switching on quite fast. I've never heard the fans on my M1 notebook. And yes, it's a 40% increase in performance, but I expected way more from a notebook with twice as much RAM and GPU cores.

First, one can't judge performance improvements of hardware based on single software function. For start code itself could be poorly written not taking full advantage of hardware advancements. Also, in all my decades of experience dealing with performance of systems I have never seen hardware changes improve all tests equally. Some improve more, some improve less, some don't at all, and some might need rewrite as they slow down.

Second, one could say going from 1:55 to 1:10 is not 40% improvement, it is 65% improvement. That is not 100% but it isn't far from it either. It would be unrealistic to expect liner scaling just by doubling number of cores, that happens next to never. So 65% is pretty good.

Next, it is unrealistic to expect more performance and get almost double performance yet not want any more fan activity. Just like with cars, higher output requires more power, more power requires more energy consumed, more energy consumed means more heat generated, more heat generated means cooling will work harder.

Last but not least, your question did you make mistake ... only you will know is 65% improvement in that test worth it to you or not and that depends on how much time you spend in that function and how much that time is worth to you.

You are doing the wrong test here.

You gave both machines an easy task to do, so the only improvement that you can see here is at the level of the hardware encoder of the machine.

You said you tested an M1 vs M3 Max... then start adding some color grading and more demanding effects, that's when the power of the M3 Max's GPU will start to show.

Look for "Candle test"

These are scores from my benchmark Standard Candle UHD: Rocket Science

MacBook Pro M1 Max

09 Blur: 32 fps
18 Blur: 16 fps
30 Blur: 10 fps
66 Blur: 5 fps

1 TNR: 49 fps
2 TNR: 24 fps
4 TNR: 12 fps
6 TNR: 8 fps

MacBook Pro M3 Max

09 blur = 39.5
18 blur = 20.25
30 blur = 12.25
66 blur = 5.5

01 TNR = 63.5
02 TNR = 30
04 TNR = 14.5
06 TNR = 9.5

Faster, but not "I have to run out and buy a new M3 Max at once"-faster, if Resovle is your main focus.

In comparison, my relatively basic 2019 Mac Pro with 12c and two Vega Pro II:

09 blur = 78
18 blur = 43
30 blur = 27
66 blur = 13

01 TNR = 64
02 TNR = 35.5
04 TNR = 18.5
06 TNR = 13

But I'm very happy with my M3 Max. I do more Blender than Resolve and in Blender that benchmark score jumped from around 1000 to 3440. My Mac Pro scores around 850 per card, so 1700 total for two GPUs with 32GB Ram each.
That is thanks to the new raytracing and general architecture, and it was the reason why I moved from M1 Max to M3 Max.

First, one can't judge performance improvements of hardware based on single software function. For start code itself could be poorly written not taking full advantage of hardware advancements. Also, in all my decades of experience dealing with performance of systems I have never seen hardware changes improve all tests equally. Some improve more, some improve less, some don't at all, and some might need rewrite as they slow down...

Thanks for you response.

Firstly, I didn't judge the overall performance improvements of the M3 Max. The other benchmarks I tried show quite impressive results (such as Blender - about 2.5x or Xcode almost 3x). I just recently used Magic Mask on my M1 and was hoping to achieve more impressive results (subjectively speaking) with the M3 Max. So it was asides from benchmarks the first real use-case to test my new notebook.

Secondly, as I wrote, I expected an even higher performance, but I've learned that Magic Mask is based on Machine Learning, and the results here aren't as impressive as other benchmarks; again subjectively speaking from the perspective of a user. As Apple themselves states " ... Neural Engine is up to 60 percent faster than in the M1 family of chips ...". And this is consistent with my test.

Thirdly, yes, there is a higher density of transistors in the M3 chip and hence it generates more heat with higher load. But as other reviewers wrote, naturally the 14" casing can't absorb the heat as good as the 16" casing and hence the fans kick in earlier. The fans on full speed are annoying.

I am satisfied so far with the M3 Max, and will do further results in my everyday settings (e.g. in my case running a few VMs and Docker containers). But this has nothing to do with DaVinci Resolve and hence not a topic for this forum.

So again, thanks to all of you for your answers and have a nice Sunday.

I don't think it's the M3. I compared my old Intel, with the M3. It's faster but I have huge performance problems with exrs anyway. Operation like channel boolean is super slow. I mainly use Fusion, so I can defiantly say, it affects all version of Fusion after release 15.
No idea why, but my PC flies in different.
Also Resolve seem to be faster now on PC. Which is really sad to see. I have no idea why this happen, but it must have to do with updates. Especially on my PC and old Mac, I haven't changed anything. No OS updates, no hardware change since at least 3 years. Only thing I upgraded was Resolve and Fusion.

Time to drop Blackmagic ..

For a while, people complained that BM cares more for Macs
Anyway, if you give a demanding task to BM, make sure to have enough RAM on your Mac too. Those GPU cores can grab close to 2 GB each. Just watch with the Activity Monitor if the machine doesn’t start to swap. It won’t crash, but get really slow.

Thank you for the benchmarks related to Resolve. Really answered the question on if doing a trade-in and spending a bunch would provide a boost over my existing M1 Max MacBook Pro 64GB.

Apparently not significant enough to warrant the increase in cost.

Second that.